In an exclusive video interview with Alison Corfield, brand director of the Jamie Oliver Group and Rebecca Lewis, cofounder of Baby Teeth we discussed how the series came together, why they chose the format and the business thinking behind the partnership.

This article builds on my coverage and analysis of how I think vertical drama is establishing new forms of distribution.

The full interview and my analysis are on Forbes.

Read: Jamie Oliver’s New Vertical Comedy Shows What Brands Are Buying Now

Photo by Denis Cherkashin on Unsplash

On August 5 we found out where some of that creator content will come from: TikTok.

The interesting question is not what Disney is adding to the feed, it is what Disney is trying to change about the way people use Disney+. And, what the creators supplying that material receive in return.

I wrote about the arrangement, the engagement problem behind it and the still-undisclosed creator terms for Forbes.

Read the full piece: Disney+ Turns To TikTok For More Daily Engagement With Fan-Made Shorts

-Maureen

Photo by Solen Feyissa on Unsplash

A New Test for Human Writing

Substack has added an AI detection feature that lets readers scan longer posts, notes, replies and comments to see an estimate of that which is written by a person. Writers can also scan their own drafts, explain how they use AI, if they do, and challenge results they believe are wrong.

Soon afterwards, Menlo Ventures led a $9 million investment in Pangram, the company providing the detection technology.

The timing interested me. A publishing platform decided that readers wanted a way to check whether a piece of writing was human. Investors then backed the company supplying that check. Ahem!

OpenAI Goes Direct to Researchers

At around the same time, OpenAI announced a programme giving academic researchers free access to its most advanced generative AI models, support integrating them into their day-to-day work and a default training opt-out. It begins with 10,000 researchers and is intended to reach 100,000.

That has implications for specialist publishers such as Elsevier, Wiley, and Springer Nature.

These publishers had begun to position themselves between researchers and AI systems, offering tools, access, and licensed connections to scholarly material. OpenAI is now going directly to the researchers.

If a researcher can use an advanced model for free, with support included, surely there is less reason to reach that model through a publisher.

Photo by Saung Digital on Unsplash

What Publishers Still Control and Where They Can Show Value

This does not make the publishers irrelevant, but it changes that for which they may be able to charge.

The negotiation becomes less about selling researchers an AI productivity tool and more about the terms on which AI companies can use the research those publishers control.

That gives the publishers a different source of power and relevance. A publisher may struggle to charge for access to a model when the model provider is giving that access away. But it still has journals, archives, metadata, rights and institutional relationships that the model provider does not own.

What Remains Valuable?

Put this beside Pangram, and the change happening in front of us becomes clearer.

OpenAI is making generation easier to access and Pangram is being paid to judge where the resulting language came from.

I am not convinced detection will provide a lasting answer. False positives will still matter, and detectors may eventually be defeated by tech models trained to avoid them.

But the commercial point remains relevant to this discussion. When access to powerful models becomes cheap or free, the most valuable questions change:

- Who controls the material used to train them?

- Who can make a credible claim about what they produced?

For specialist information providers such as Elsevier, Wiley and Springer Nature, the future may depend less on bringing AI to researchers and more on negotiating over the research AI needs.

-Maureen

Two hours after the short premiered on YouTube, its director, Zack London—the filmmaker known online as Gossip Goblin—appeared on an AI Live Q&A and said something still surprisingly rare in AI filmmaking: the machines are nowhere near the human voice, and he has no intention of waiting for them to catch up.

Watch Pomegranate and you can hear why.

Its protagonist is a jaded immortal cycling through simulated lives, speaking in a register I can only describe as Noël Coward after a very long century: clipped, weary and wonderfully pithy.

“I’m not one to impugn anyone’s recreational preferences,” he sighs, stepping over the debauchery outside his dealer’s shop.

It is among the best writing I have heard in an AI film, but the words are only half the achievement. What gives them their precision is a human performance.

Photo by Daria on Unsplash

Writing To A Voice

The story of that voice is also the story of how London works.

The actor is British and based in Sweden—”probably the only professional British voice actor in Sweden,” London told the livestream audience. He had previously played a minor corporate character in London’s earlier short, The Patchwright. London liked the performance so much that he decided, in his words, to “center the next film around this character and that voice.” He already had a story in mind, he said. But he built the script out on the pairing.

Not casting to fit a finished script. Writing to a voice he already trusted.

The Same Casting Agency

London’s case for human performers is not merely sentimental. It is also aesthetic and economic.

AI voices, he argued, tend toward a “hyper-generic,” rounded-out accent. Because filmmakers are drawing from the same narrow range of synthetic performances, “it’s like we’re all kind of using the same casting agency.”

A human actor can arrive with something stranger and more particular.

For one role, an actor channeled a friend’s father from Lincoln. It was a voice rooted in a specific person, place and memory.

“I can’t prompt for this,” London said, “because I don’t think the AI even knows what this is.”

That distinction runs through the rest of Pomegranate.

He works with a Foley artist to heighten what the models only approximate, and cycled through several composers because a score, he said, “elevates or embellishes, or helps to steer the film.”

AI music, in his memorable shorthand, tends to deliver Eurovision.

Jazz, By The Throat

His relationship with generative tools is more combative than utopian. The line of the session came when he described the process as a form of constrained improvisation:

“AI is like jazz... You can guide it, but it’s not really going to give you what you want 100% of the time. Sometimes you have to grab it by the throat.”

The Signature Is The Scarcity

That may be the pattern worth watching.

As image generation becomes cheaper and more accessible, London’s work is increasingly defined by the elements he refuses to automate: writing, performance, music, sound design and editing.

These are not nostalgic additions applied after the “AI part” is finished. They are the parts that prevent the work from collapsing into the synthetic sameness of the feed.

When abundance becomes the default, specificity becomes scarce.

And the signature is the scarcity.

There is a larger industry story forming around London’s work. His first feature, Gods Don’t Give Gifts, produced by Fable, is scheduled to arrive in U.S. theaters on October 30.

I recently sat down with London over video for Forbes to discuss the film and its move from YouTube culture into theatrical distribution. You can read that story here.

For now, Pomegranate is on London’s YouTube channel.

Listen to the aristocrat.

Then consider which part of that performance a model could actually have given him.

-Maureen

Three shows to start: a romance anime, a Gen Z paranormal horror and a virtual-world survival thriller. All use AI somewhere in the production workflow.

But the important part happens when the episode ends.

The characters don’t disappear. You can open a chat with them.

As a vertical-drama launch, Character.ai is a late arrival to an increasingly crowded market. What c.ai Series reveals is more interesting: vertical drama may now have a fifth business model—and the first in which the durable asset isn’t the content.

The map so far

I’ve been arguing for a while that vertical drama isn’t one category. It’s a form factor—short, serialized episodes built for the phone—being used by at least four different businesses that happen to share a screen orientation.

The paywalled apps, ReelShort and DramaBox chief among them, run economics closer to mobile gaming than television. The episode is the product. Cliffhangers convert into paid unlocks. And as executives acknowledged at MIP London this year, as much as 90% of budgets can go to marketing and user acquisition rather than content.

The customer is effectively repurchased with every campaign.

The social platforms run a second model. TikTok’s deal with Issa Rae’s Hoorae isn’t really a content business. The shows exist to keep you in the feed, and the feed is the business.

Brands run a third. P&G’s The Golden Pear Affair is a 55-episode soap that exists to sell product: serialized marketing wearing drama’s clothes.

Then there are the streamers. Netflix, Disney+ and Prime Video aren’t becoming vertical-drama companies at all. They use vertical clips as discovery—a doorway into the horizontal catalog they already own.

The money confirms the market is real even as its business models diverge. Short-drama apps generated $2.98 billion in in-app purchase revenue in 2025, up 115% year over year, according to Sensor Tower. Deloitte expects in-app micro-series revenue to more than double to $7.8 billion in 2026.

Hollywood has moved past watching from the sidelines. Fox took an equity stake in Holywater. Peacock is licensing ReelShort microdramas. Disney placed DramaBox in its accelerator.

The fifth model

Character.ai fits none of those four boxes, which is why I think c.ai Series deserves more attention than a three-show slate would normally earn. The company starts with an asset the others would kill for: roughly 20 million monthly active users spending an average of about 75 minutes a day on the platform. That isn’t viewing time; it’s conversation time. And the bet underneath Series is that the character, not the episode, is the durable asset.

A viewer who finishes an episode and then spends an hour talking to its protagonist is not a completed transaction. That viewer is now inside a relationship the platform owns. CEO Karandeep Anand told me the chat integration builds fan engagement with the original characters “while simultaneously providing a feedback loop” for the company’s in-house team on what resonates. The first slate was already shaped by that logic, with genres drawn from the kinds of characters and worlds the Character.ai community gravitates toward. Development risk starts to look less like a commissioning problem and more like a data problem.

The return mechanism is different in kind, too. ReelShort brings you back with a cliffhanger it paid to put in front of you; Character.ai is proposing to bring you back with a relationship you formed on your own.

What all five models are really competing for

Here’s the through-line I keep returning to in this newsletter: AI tooling is pushing the production cost of this format toward a floor, and as episodes become cheaper to make, the contested ground moves elsewhere. Every one of these five business models is answering the same question: what do you own when the content itself becomes a commodity?

ReelShort owns an acquisition machine.

TikTok owns the feed.

P&G owns the shelf the drama points toward.

Netflix owns the catalog the clips lead into.

Character.ai is proposing to own something I’d argue is scarcer than any of them: the audience’s attachment, and the first-party data generated by that attachment.

Voice, taste and trust were always going to become more valuable once volume stopped being scarce. A character you can talk to is that thesis turned into a product.

What I’m watching

Three things will show whether the fifth model is actually a model.

First, pricing. Anand said Character.ai will test monetization “similar to traditional microdrama platforms.” But if the relationship is genuinely a different asset, it should eventually support a different revenue model from the episode unlock. If it can’t, the fifth model collapses back into the first.

Second, the creator tools. The plan is to give a small group of leading creators access to the production tooling, then open it to any creator, publisher or studio. That would move Series from a slate toward a format and put Character.ai where every microdrama platform wants to be: supplying the pipeline rather than filling it.

Third, conversion. Seventy-five minutes a day is engagement; the open question is whether attachment converts into revenue at better economics than the incumbents’ 90%-to-marketing machine.

The incumbents proved audiences will pay for vertical stories one cliffhanger at a time. Character.ai is testing whether the character can outlive the cliffhanger. And if it can, the most valuable thing in the next entertainment franchise may not be the story you watched. It may be the person who stayed behind when the episode ended.

-Maureen

Photo by Pavol Duracka on Unsplash

That last detail is the whole story, and most of the coverage has read straight past it.

The easy version is that this is a cost cut. A studio points generative tools at a production line item, and the losers are the editors, composers, agencies and post houses that used to make the spot. That version is true, but it is too small. It treats the ad as an expense to be reduced rather than a barrier to be removed, and it never asks what the barrier was holding back.

Premium video advertising has never really been short of reach. Reach is the easy part. What has always been scarce is a good enough ad. Making one that can sit beside a film, a football match or a live event without embarrassing the brand takes money, time, a crew, a composer, an edit, and usually an agency to hold it together. That cost sorted advertisers into two groups: the ones who could afford broadcast-grade creative, and the far larger number who could buy search, social or retail media in an afternoon but could never make a thirty-second film.

The second group was locked out of premium video, not by the price of the media, but by the price of the creative.

That lockout did two things at once. It kept premium inventory scarce and expensive, which suited the people selling it. And it gave agencies their reason to exist, because someone had to make the thing most brands could not make for themselves.

Disney’s beta is aimed exactly at that locked-out group. Read it that way and it stops being a cost story and becomes an access story. Disney is not mainly trying to spend less on its own advertising. It is trying to let in the advertisers who were never able to run a proper video ad in the first place.

The generative tool is the key that opens the door. A business that lives on how much of its advertising inventory it can fill has every reason to open that door as wide as possible. The hours of attention are already there. The inventory exists. The problem is that too few advertisers can arrive with something fit to run in it.

That is the first shift: the platform lowers the creative threshold so more advertisers can enter premium video.

The second shift is subtle and larger. The company that sells the advertising space is now making the thing that goes in it.

For most of the history of this business, those were separate functions with separate loyalties. The brand and its agency made the ad. The broadcaster or platform sold the space. Keeping them apart was not incidental. It was the operating model.

Now Disney writes the ad, runs the ad and tells the advertiser whether the ad worked, all inside the same commercial machine.

That is why “agentic” is the right word here, even if it is overused. The change is not that AI can generate a picture or a line of copy. We have had that for a while. The change is that the system begins to act across the workflow. It does not just produce an asset. It moves the campaign forward.

You can see it in the Gordon Ramsay work for PolyAI. The ad does not simply sell AI. It dramatises the replacement of a broken service line with an agent that completes the task. Ramsay rages at the dead end of the phone tree, and the agent becomes the thing that listens, interprets and acts.

The product being advertised is agency itself: the ability to move from request to completion without handing the user back a maze.

Fox is moving the same idea into the transaction layer. Its new agentic advertising platform is not another dashboard. The pitch is end-to-end: planning, buying, activation, and transactions between buy-side and sell-side agents. It takes AI off the creative surface and puts it into the market structure underneath.

The old advertising stack was full of handoffs: strategy to creative, creative to media, media to trafficking, trafficking to measurement, measurement back to optimisation. Each handoff supported an industry. Agentic advertising treats the handoffs as latency, which is why the platform wants the whole chain. The more steps it absorbs, the more it can promise the thing advertisers actually buy when they buy automation: fewer meetings, fewer vendors, fewer delays, fewer excuses.

Alex Gough Cooper’s work points to the same pressure from the other end. The interesting use case is not a one-off AI-generated ad. It is the loop around the ad: research, concepting, variation, testing, learning, more variation. The creative unit is no longer the campaign. It is the loop.

Make, test, learn, remake.

The craft does not vanish, but it changes location. It moves upstream into the judgment that sets the frame and downstream into the judgment that reads what the machine has learned. The middle, the production of options, floods.

That is the crux, because premium video was built for scarcity: scarce slots, scarce production, scarce attention, scarce creative. The new system is built for abundance: more versions, more targeting, more surfaces, more advertisers, more automated decisions.

Disney’s beta, Fox’s buying layer, the Ramsay ad and the operator running a test loop instead of making one spot all belong to that world.

Put them next to each other and the pattern is hard to miss. Advertising is being pulled into the platform. Not only the media buy. Not only the measurement. The message itself.

That does not mean agencies disappear, but their default position weakens.

What is left is the argument that the platform should not mark its own homework. Independence becomes the product. There is a real business in telling a brand that the company selling the space should not also write the message and grade the result.

WPP and others are already moving that way with agentic buying standards, because if AI agents are going to negotiate with media-owner agents, someone has to define how those systems communicate, validate, escalate and prove what happened.

That is a serious role, but a more defensive one than agencies used to hold. It asks the brand to pay for distance at the exact moment the platform is offering to remove the cost of distance altogether.

And for many advertisers, especially the ones Disney is targeting, distance was never the product. Access was.

A small business that has never been able to afford television is not going to be precious about who held the camera. It wants the ad made, placed and measured. If the platform can do all three in an afternoon, that is the offer to beat.

Which is why “will the AI ads be good?” is less important than it sounds.

On day one, some will be good and many will be mediocre, carrying the faint, dead sheen of machine-made advertising that people recognise even when they cannot name it. But quality is not the first battle. Friction is.

The system that removes friction gets the budget. And a system with the budget improves.

The real question is not whether machine-made ads will be any good. It is what a brand becomes when the company it is paying to reach an audience is also the company that decides what it says.

If the platform writes the message, runs the message and scores the message, whose message is it?

-Maureen

Images: Photo by Kéoma Oran on Unsplash

I am not sure it survives contact with how Roku actually earns its money.

Roku is worth what it is worth because it is, to a useful approximation, nobody’s. Disney puts its app on Roku. So does Netflix, so does Warner, so does Amazon, so does every advertiser buying inventory against them. They do that because the platform sits underneath all of them without obviously belonging to any one of them. The home screen has value because it is not supposed to be tilted. Neutrality is not something Roku offers on top of the product. It is the product.

Which is the problem with selling it to Fox.

The moment a Roku home screen belongs to the company that also owns Tubi, Fox One, Fox Sports and Fox News, every other media company on the platform has to ask a question it never had to ask before. When I search for football, whose match surfaces first? When I open the home screen with nothing in mind, whose free channel greets me? When my app competes with Tubi for the same tile, who sets the ranking, and do they work for my competitor?

Fox can answer every one of those questions honestly and fairly, each time, and it will still not be enough. The damage does not require Fox to actually tilt the screen. It only requires rivals to suspect it might. A neutral platform runs on a kind of trust that does not survive the discovery that the umpire owns one of the teams. Once the umpire owns a team, every close call becomes evidence.

This is the part the financial logic walks straight past. Fox is not buying a switchboard and inheriting its value intact. It is buying a switchboard whose value was always conditional on no single caller owning the exchange. Ownership does not transfer that value. It begins to spend it.

And here is the trap folded inside the trap. The harder Fox uses Roku the way the deal logic says it should, to push Tubi, to convert Fox One subscribers, to give Fox Sports the best placement through the World Cup, the faster it confirms the suspicion. Restraint protects the platform and wastes the reason for the purchase. Aggression realises the purchase and corrodes the platform. There is no setting on the dial that hands Fox both.

Fox and Roku already know this. They have promised Roku will stay open and partner-friendly. They had to. But a pledge of neutrality from an interested owner is a strange instrument. It asks competitors to trust that Fox will decline to use the very advantage it paid twenty-two billion dollars to acquire. Either the pledge is real, in which case the strategic premium was overpaid, or the premium was justified, in which case the pledge cannot fully hold.

That is what makes the deal stranger than a content-plus-distribution story. It is a bet that Fox can own the interface without making the interface feel owned.

I think this is the subtle lesson sitting under a lot of the deals being chased right now. Everyone has worked out that the interface is the scarce position, that content is becoming abundant and the surface that organises it is becoming rare. What fewer people have worked out is that the most valuable surfaces earned their value by belonging to no one.

So the question I cannot get past is not whether Fox can afford Roku. It is whether anyone can own a thing whose entire worth was that nobody did.

-Maureen

Image: Photo by Glenn Carstens-Peters on Unsplash

I wrote about Fox’s two World Cup streaming products for Forbes last week. This piece is about the shift underneath them. Free ad-supported streaming has stopped being the discount tier of the business and started being a strategy in its own right, and live sport is where that change is most visible.

The decade that miscast free

For most of the streaming decade, free meant residual. Subscription was the business. Free was the funnel, the library dumping ground, the place catalogue went to earn something after the paying windows closed. The orthodoxy held that value lived in conversion, and that an audience you had not converted was an audience you had failed to monetise.

That framing was always borrowed from software, not television. Television spent seventy years proving the opposite: an enormous simultaneous audience, assembled free of charge, is one of the most monetisable assets in media. The subscription decade did not disprove broadcast economics. It just stopped practising them.

The Super Bowl was the test

Fox ran the experiment sixteen months ago. In February 2025 it became the first Super Bowl broadcaster to put its live feed out free, simulcasting the game on Tubi in 4K. The result was the most streamed Super Bowl in history: an average minute audience of 13.6 million, a peak above 15 million concurrent viewers, and 24 million unique viewers across game day programming.

The number that settled the argument was not a viewing figure. Fox reported more than $800 million in gross advertising revenue across the broadcast and the Tubi stream combined. Giving the game away did not cannibalise the economics. It extended them. The free stream added reach, the reach added inventory, and the inventory sold at Super Bowl prices.

The World Cup decision follows from that result. Fox is not placing the opener on Tubi as a loss leader. It is repeating a proven trade: the biggest possible audience, assembled at zero price, sold to advertisers at premium rates.

Photo by Norbert Braun on Unsplash

The numbers underneath

The category data says this is no longer a remnant business. Nielsen put ad-supported streaming at 45.6 percent of all ad-supported television viewing in the final quarter of last year. Within that, Tubi held a 6.2 percent share of ad-supported streaming viewing, which Nielsen’s ad-supported measure puts ahead of Netflix, Peacock, Pluto TV and Paramount+.

Tubi’s 18 to 49 audience has grown roughly tenfold in five years. On Fox’s most recent earnings call, Lachlan Murdoch reported Tubi revenue growth of 23 percent in the quarter.

Read those numbers against the platforms still defending paywalls around live sport and the contrast is sharp. The free services are compounding around audiences advertisers still struggle to reach elsewhere. The paid services are spending heavily on rights to slow churn among audiences they have already monetised once.

Why sport belongs to reach

Advertising compounds on simultaneous mass attention, and live sport is the last product in media that reliably produces it. Everything else has fragmented into personal queues and asynchronous viewing. A World Cup group game still gathers millions of people at the same moment, watching the same thing, through the same commercial breaks.

A paywall shrinks precisely the asset that makes sport valuable to advertisers. Every conversion gate in front of a live match trades audience scale, the thing sport uniquely supplies, for subscription revenue, the thing every other content category is already chasing. That trade can still make sense for a platform building a direct customer relationship. It is no longer obviously the best use of the biggest rights.

What the free streamers are building is not new. It is broadcast television, reassembled with better measurement, addressable inventory and no spectrum constraint. The irony of the streaming decade is that its most modern business model is its oldest one.

What gets repriced

If reach is becoming the scarce asset again, the rights market reprices around it. Leagues and federations sell to whoever can deliver the audience their sponsors and their growth ambitions require. For two decades that pointed first to pay television, then to subscription streamers. The Super Bowl result and the World Cup configuration point somewhere else: to hybrid structures where free distribution carries the biggest moments and paid products carry depth.

The platforms most exposed are the ones that bought sport purely as a conversion tool. The platforms best positioned are the ones that can do both jobs at once, which is the structure Fox takes into this tournament.

The next major rights package to come to market will be the real test. The old question was which service could convert the audience. The new question is simpler: who can still assemble one?

-Maureen

The easy reading is that production is getting cheaper, so power is moving to the outsider: the filmmaker without a studio, the fan with a remix tool, the creator with a phone, the comic with a public-access set.

Sometimes, that is true. But as a theory of media power, it is incomplete. Cheap production does not make control disappear. More often, it changes where control sits.

The useful precedent is the soap opera. In the early days of radio, Procter & Gamble saw that low-cost daytime serials aimed at women at home could sell household goods. The programmes were inexpensive. That was the point. But the value was not only in the programme. It was in the sponsor, the slot, the audience habit, and the relationship between the story and the product being sold.

The content was cheap. The control was not.

The rhyme is not exact, but it is useful. AI is lowering the cost floor again. Microdrama is compressing serial storytelling into one- or two-minute vertical episodes. Fan tools are turning listeners into remixers. Public access can briefly look like the anti-network.

But the old question keeps returning. Who owns the slot?

The cost floor drops again

The headline number attached to Dreams of Violets, the AI-generated feature set to premiere at Tribeca on 10 June, is the reported production cost: around $2,000. Rolling Stone reported that Ash Koosha made the 75-minute film for roughly that amount, while the company announcement presents it as a feature-length, live-action film completely generated by AI and accepted by a major festival.

It is a striking figure. It invites the most optimistic reading of AI production: if a feature can be made for the price of a laptop, then the studio system loses some of its power. The barrier falls. More people get to make things.

But the number also proves something less romantic. The cost of making the film may have collapsed. The cost of being seen has not.

A $2,000 film still needs selection, attention, distribution, legitimacy and audience access. In this case, Tribeca supplies one of the scarce things: institutional validation. The festival does not make the film, but it helps decide whether the film enters the conversation.

That is the pattern beneath the week’s stories. When creation gets cheaper, value does not vanish. It moves to whatever remains scarce: permission, placement, trust, discovery and the audience relationship.

Peacock rebuilds the old model in a new format

Peacock may be the cleanest example.

NBCUniversal is bringing microdramas into the Peacock mobile app with two kinds of supply. One is built around Bravo talent and owned reality IP, including Southern Charm and Real Housewives connections. The other is a licensed slate of scripted material from ReelShort. Peacock’s own announcement says the format is coming to its mobile app this summer, while Business Insider reported that the ReelShort package includes ten microdramas.

That route in is telling. Peacock is not simply discovering microdrama as a disruptive outside force. It is absorbing the format on terms it can control.

It owns some of the IP. It licenses some of the supply. It controls the app. It owns the viewer relationship.

Strip away the vertical screen and the accelerated pacing, and the logic is familiar. The serial is cheap on purpose. The platform does not need every episode to be a masterpiece. It needs repeat viewing, habit, retention and a reason for the audience to stay inside Peacock.

This is where the soap-opera comparison earns its keep. The value is not just in the show. It is in the place where the show lives, the audience it gathers, and the commercial structure wrapped around it.

Spotify and Universal price the permission

The music industry is moving from another angle, but the same principle applies.

Spotify and Universal Music Group’s new agreement on fan-made AI covers and remixes is framed around consent, credit and compensation. Spotify says the tool will launch as a paid add-on for Premium users, with participating artists and songwriters sharing in the value created by licensed AI covers and remixes.

The structural move is the interesting part. When AI makes musical variation cheap, the rights holder does not disappear. The rights holder becomes more important.

The creation may be easy. The permission is not.

I have written separately about the Spotify and Universal deal for Forbes, so I will not re-run the full argument here. But it fits this week’s broader pattern. The industry is not only asking whether AI music can be made. It is asking who gets to authorise it, price it, label it and monetise it.

Colbert proves even the joke has an owner

The smallest story may be the sharpest.

After his final Late Show, Stephen Colbert returned to Only in Monroe, the Michigan public-access show he had hosted in 2015 before taking over from David Letterman. Jack White, Jeff Daniels, Steve Buscemi, Eminem and Byron Allen all turned up.

It was a deliberately low-cost public-access frame for one of the most expensive figures in American late night.

Then CBS and Paramount issued copyright takedowns on unofficial uploads, before suspending enforcement after the backlash.

That is almost too neat as a media parable. The set was cheap. The setting was local. The joke was about leaving the network machine. But the clip still had an owner.

Even public access was not outside the rights system.

Cheap content does not mean weak control

Put the examples together and the supposed disruption starts to look more complicated.

A film can be generated for around $2,000, but it still needs a festival, a distributor and an audience pathway.

A microdrama can be made quickly and watched vertically, but inside Peacock it becomes part of a major streamer’s retention and IP strategy.

A fan can create an AI remix, but only inside a licensed structure where the platform and rights holder set the terms.

A public-access comedy segment can feel like escape from the network, but the takedown notices remind everyone that rights still travel with the content.

This is not an argument against cheap production. Lower costs do change who can make things. They open doors. They allow more experiments, more formats, more outsiders and more strange work to exist.

But media history suggests that cheaper creation does not automatically produce a more open system. Often, it creates a larger supply of things that still need to pass through scarce channels: distribution, discovery, licensing, recommendation, legitimacy and payment.

The soap opera was never really about the soap, and it was never only about the programme. It was about who paid for the slot, who owned the audience habit, and who turned cheap serial content into commercial leverage.

AI lowers the production floor further than radio ever did. Vertical video lowers the friction of serial storytelling. Streaming platforms can absorb formats that began elsewhere. Music platforms can turn fan creation into licensed product.

So the question is not whether more people can make things now. They can.

The question is who owns the slot, now that the thing costs almost nothing to make.

-Maureen

Photo by Anthony Lim on Unsplash

It is the wrong comparison.

Netflix is not trying to become TikTok. It is trying to own the few minutes before a member decides what to watch.

That moment is the contested one. Someone opens Netflix without a plan. They have a few minutes. They are not ready to commit to a series or a film. They browse, hesitate, and either find something quickly or leave for YouTube, TikTok or Instagram.

Clips is built for that moment.

It is not really about short video as entertainment in itself. It is about stopping the user from leaving before they choose.

Netflix has been circling this for years

The feature looks new, but the idea is not.

Netflix has been testing vertical video on mobile for years. Mobile previews arrived in 2018. Fast Laughs followed in 2021. Those products were narrow. One was closer to a trailer strip. The other was built around comedy clips. Both sat near the main viewing experience rather than changing it.

Clips is different.

It makes vertical video a broader discovery layer for Netflix. The format is familiar. The ambition is bigger.

That distinction is the point. Netflix has not moved its main product into vertical video. The big shows, films, documentaries, sports and live events are still made for horizontal viewing, usually on the biggest screen in the home.

The phone is being treated differently. It is not just a smaller television. It is a place where people sample, decide, save and share.

Discover vertically, watch horizontally

The design says a lot.

Netflix has added a vertical feed without making the whole app feel like one. The row-based home screen is still the main door into Netflix. Clips sits in the bottom navigation as a separate tab. You have to choose to enter it.

That is not how TikTok works. On TikTok, the feed is the front door. On Netflix, the feed is a room you walk into.

That difference is the strategy.

Each clip has a job. It can push you to add a title to your list. It can make you share something with a friend. It can send you into a related collection. In other words, the feed is not meant to keep you there forever. It is meant to move you somewhere else.

Netflix wants you to discover vertically. It still wants you to watch horizontally.

The phrase Netflix keeps coming back to

Elizabeth Stone, Netflix’s chief product and technology officer, made the point before launch. Speaking at TechCrunch Disrupt in October, she said the vertical feed was not meant to copy or chase TikTok. Netflix, she said, was not trying to be “all things at every moment.” It was focused on what the company calls “moments of truth.”

The phrase is doing work. It has now become part of Netflix’s own language. In its Q1 2026 shareholder letter, Netflix listed “winning the most valuable moments of truth” as one of its priorities.

Clips fits that line exactly.

Netflix says Clips is for “the moments in between”: when someone wants to discover a title, watch a quick scene, or get a quick laugh.

I wrote about this product in Forbes earlier this week. The framing I kept returning to was the tension between intent and form. Netflix says Clips can help people decide what to watch without endless scrolling. But Clips is also a vertical, swipeable feed. And vertical, swipeable feeds come with a strong habit attached to them. Users have already learned that grammar from TikTok, YouTube Shorts and Instagram Reels.

Netflix may have a different intention. The format still carries other people’s habits.

Netflix is strong on the sofa. The fight is on the phone

Netflix does not have a living-room problem in the same way many entertainment companies do. In December, it captured 9.0% of total US TV viewing, the highest share Nielsen has recorded for the company.

That is a strong position.

But that figure measures television screens. It does not solve the phone.

The phone is where idle minutes go. It is where people open apps without a plan. It is where YouTube Shorts, TikTok and Instagram Reels compete all day, in small bursts, between other things.

Netflix knows this. Its shareholder letter said the lines between entertainment on TV and mobile are blurring. It also said video podcasts on Netflix are already over-indexing on daytime viewing and mobile devices.

Read alongside Clips, the pattern is clear. Netflix is treating the phone as its own surface, not as a smaller version of the living room.

The company already has people watching its shows on the television. Now it wants those same people opening Netflix when they are bored, between tasks, or not quite ready to choose a full episode.

This is about frequency

That is why Clips is more than a discovery feature.

For years, Netflix was mainly judged like a subscription business. Keep people subscribed. Reduce cancellations. Raise prices when the service is strong enough to carry them.

That logic still holds.

But advertising adds another logic.

An ad business does not only care whether people stay subscribed. It also cares how often they open the app, how many surfaces they touch, and how many ad opportunities the product can create without making the experience worse.

Netflix’s advertising business is no longer a side project. The company has said 2026 ad revenue is on track to reach about $3 billion. In countries where the ad-supported plan is available, more than 60% of new sign-ups now choose it. Programmatic advertising is also expected to account for more than half of non-live ad inventory this year.

That changes how to read Clips.

Clips is a discovery layer for a Netflix that increasingly benefits from frequency. It gives the app a reason to be opened when the user is not ready to watch a full episode or film. It creates another place for Netflix’s titles to circulate. It makes the phone more important to the business.

The question is not whether Netflix is becoming TikTok. It is whether Netflix can make the phone part of its own system without importing the worst habits of the feed.

-Maureen

Photo by Sayan Ghosh on Unsplash

Netflix launched Clips on April 30, bringing a vertical video feed into its redesigned mobile app.

Much of the coverage has treated the move as Netflix joining the TikTok wave—late, defensive, and format-chasing. I read it differently.

In my latest Forbes piece, I argue that Clips is less about copying TikTok than about how Netflix is repositioning mobile within its wider streaming business.

Read the full piece on Forbes: https://www.forbes.com/sites/maureenkerr/2026/05/13/netflix-clips-wont-replace-tiktok-but-will-influence-viewers/

-Maureen

Image created by Shyam Mishra (Unsplash)

Snap’s revenue range for the second quarter assumes no contribution from Perplexity. The $400 million one-year cash-and-equity arrangement announced in November, framed as a large-scale integration of Perplexity’s answer engine into Snapchat, ended in the first quarter. Management described the wind-down as amicable. Neither side has fully explained what changed.

That makes Snap and Perplexity an early public test of AI distribution economics. It is a more useful story than a single earnings footnote suggests, particularly for media boards.

What the deal was supposed to do

The Perplexity arrangement was structured to put an external AI answer engine directly inside Snapchat’s Chat interface. Perplexity would gain access to Snapchat’s young, mobile-native audience. Snap would receive $400 million over one year, in cash and equity, as global rollout was achieved. Snap described the partnership at announcement as the first large-scale integration of an external AI partner directly into the app, signalling that this was a template for how AI companies would reach Snapchat’s audience, not just a single commercial agreement.

The strategic logic on both sides was sound. Snap has been trying to diversify beyond advertising, especially while large North American advertisers remain a headwind. Perplexity needed consumer distribution beyond search, browser, and enterprise channels. Snapchat offered a scaled surface where AI search could become part of everyday messaging behavior.

The commercial logic was less tested. A $400 million price depends on assumptions about adoption, query volume, retention, monetization, and product fit. It also depends on where value actually accrues: to the platform that owns the audience, to the AI product that answers the query, or to whichever surface the user returns to next.

None of those variables had much public evidence at the scale the deal implied. By February, Snap was telling investors the two sides had yet to agree on a path to broader rollout. By the end of the first quarter, the relationship was over.

Why the wind-down is the more useful data point

Coverage will read this as a Snap problem. The Perplexity contribution is now absent from Q2 guidance. Other Revenue, which grew 87 percent year-on-year to $285 million in Q1, has to compound from a different base. That is real, but not existential. Snap’s subscription business, including Snapchat+, Memories Storage, and Lens+, is already doing the diversification work the Perplexity arrangement was meant to accelerate.

The more useful point is what the wind-down suggests about AI commercial agreements across the market. Three observations stand out.

The first is that headline values in AI distribution deals are being set ahead of proven unit economics. In traditional content, distribution, or licensing agreements, a number of this size usually sits alongside volume commitments, performance gates, rollout triggers, termination rights, and revenue-sharing mechanics. The public version of the Snap-Perplexity deal was a headline number tied to rollout. The fact that both sides could walk away amicably within a quarter suggests the early evidence did not justify pushing the relationship forward.

The second is that consumer surfaces remain difficult places to monetize generative AI. Perplexity is one of the better-positioned independent AI companies in product quality and consumer awareness. Snapchat is one of the few platforms with a large, young, habitual messaging audience. If that combination could not quickly produce a rollout path both sides wanted to continue, the assumption that AI products will compound simply by being placed inside social or messaging surfaces deserves more scrutiny.

The third is governance. Boards approving large AI commercial commitments are being asked to underwrite forecasts that remain thinly tested. Snap and Perplexity have now produced a public early-mover example. The arrangement was announced in November. By February, Snap was already telling investors there was no agreed path to broader rollout. By the end of the first quarter, it was gone from guidance. That is a five-month arc from announcement to disappearance, and it is information.

The reframe for media companies

Most of the AI commercial commitments now in the market involve media companies on one side. Studios are licensing libraries to model providers. Publishers are signing distribution or training agreements with AI companies. Music rights holders are negotiating new compensation structures for synthetic media and model development.

Each of these deals involves a counterparty whose ability to turn AI capability into durable audience behavior is still being tested. That does not mean the deals are wrong. It means they should be structured as if the early forecasts might be wrong.

The Snap-Perplexity case is therefore a precedent worth reading carefully. It argues for floor commitments rather than headline values. It argues for performance gates that adjust economics as adoption becomes visible. It argues for termination language that makes renegotiation normal rather than reputationally awkward. It also argues for disclosure language that does not force companies to defend continuity after the commercial rationale has changed.

Media boards approving large AI deals in the second half of 2026 will now be doing so with a visible example of how quickly an early-mover commitment can unwind. They should ask their counterparties how their deal differs.

The cleanest lesson is not that Snap failed, or that Perplexity failed. It is that distribution alone is not yet proof of AI monetization. The market is still learning what these products are worth when they leave the demo environment and enter someone else’s daily interface.

-Maureen

Image created using OpenAI.

The framing is elegant. It is also, on closer reading, the inverse of what is actually being built.

iQiyi is not decentralising. It is consolidating. The platform is taking ownership of the entire production stack, from foundation models to digital assets to distribution to monetisation. What it is decentralising is the financing, the editorial judgement, and the commercial risk. Those are now the creator’s problem.

This is worth understanding clearly, because iQiyi is the first Netflix-scale streamer to commit publicly to AI-generated content as the future of its catalogue, and the model it is building is the most aggressive version of a move the industry will spend the next eighteen months evaluating.

What “open” means in this context

The standard reading of a creator-economy move is that the platform gives up control in exchange for volume. YouTube gave up editorial control over what got uploaded and won the long tail. TikTok gave up control over what the algorithm surfaced and won short-form attention. The creator economy thesis, in its classic form, is that platforms cede the production systems to individuals and capture value at the distribution and discovery levels.

iQiyi is doing something different. The production systems is not being ceded. It is being absorbed.

Nadou Pro is the production engine. Roughly seventy AI agents covering every stage of film and television production, integrated into a single workflow. The IP library, the digital asset library and the talent database are all iQiyi’s. The distribution and monetisation systems are iQiyi’s. The performance-tied revenue share is iQiyi’s mechanism.

What the creator brings is the idea, the labour, and the financing. What the creator owns, in iQiyi’s framing, is the IP, the audience relationship, and a share of the upside. What the creator carries is the risk that the project does not perform.

The move is being described as opening the platform. Mechanically, it is closing the stack.

The risk transfer

The financial logic underneath this is precise. Traditional Chinese long-form streaming, on iQiyi’s previous model, ran on the same economics as Netflix or any other commissioned-content business. The platform paid for development. The platform paid for production. The platform absorbed the loss when a title underperformed and captured the upside when it did not. Editorial judgement was centralised because the capital was centralised.

The new model breaks that link. The capital obligation moves outward to creators. The editorial judgement, accordingly, moves outward too. iQiyi keeps the parts of the value chain that scale: the production tooling, the asset library, the distribution surface, the monetisation system. It exits the parts that do not: development capital, editorial selection, individual project risk.

A 20% subsidy for AI-generated mid-form series through year-end 2026 is not a generous incentive. It is a transition cost iQiyi has decided is cheaper than continuing to commission long-form drama under the old model. The Chinese box office is soft. The subscription business is strained. Pushing financing risk to creators in exchange for a platform fee is a more attractive economic position than commissioning the catalogue itself.

This is not a creator-empowerment story. It is a margin-protection story.

The Netflix counter-example

The contrast that makes this clearest is Netflix. In March, Netflix acquired InterPositive, the AI filmmaking technology company founded by Ben Affleck, for up to $600 million. Co-CEO Ted Sarandos described the technology as “proprietary technology created specifically for filmmakers and filmmaking.” Netflix is bringing AI production tooling in-house. The use case is its own commissioned content, made by filmmakers Netflix pays under conventional production economics.

The tooling is roughly equivalent to what iQiyi is offering through Nadou Pro. The financing model is the inverse. Netflix is using AI to lower the cost of content it commissions. iQiyi is using AI to exit commissioning altogether and push the financing risk to creators. Same tooling, opposite economic position.

That comparison reveals that iQiyi’s move is a strategic choice, not a technology necessity. The capability does not require the platform model. iQiyi is selecting the model that transfers risk outward because the alternative, the Netflix model, requires continuing to absorb development capital and individual project risk on a balance sheet that is already strained. The AI tooling is the enabler. The strategic question is who pays for the content and who keeps the upside, and on that question Netflix and iQiyi have answered in opposite directions.

What the rights structure reveals

The clearest signal of what iQiyi is actually building is the actor talent database, and the response to it.

iQiyi announced that more than one hundred actors had joined the Nadou Pro talent database. Within hours, the studios of Zhang Ruoyun and Yu Hewei publicly stated that the actors had not signed any AI-related authorisation, and that their legal teams were following up. iQiyi clarified the next day that inclusion in the database indicated only a willingness to discuss potential collaborations. Each project would require separate negotiation and a signed agreement before any use of a performer’s image or voice.

The clarification is telling. iQiyi is centralising the talent section, in the sense that the database is its asset and is positioned as a feature of the production platform. But it is doing so without first securing the contractual permissions that ownership of that layer would normally require. The platform wants the optionality of presenting talent as a feature inside Nadou Pro without taking on the rights infrastructure that would make that optionality real.

That is structurally similar to what is happening on the creator side. The platform is extending its surface area without extending the obligations that traditionally come with that extension. Volume without commissioning, talent connectivity without licensing, distribution without curation. The model works only if the parties on the other side accept asymmetric terms. The actor studios are already pushing back.

Why this travels beyond China

The temptation is to read this as a Chinese streaming story shaped by domestic conditions. Soft box office, strained subscriptions, regulatory tailwinds for AI deployment. All of those are real. But the structural move iQiyi is making is one any large content platform will examine.

Every commissioned-content business is exposed to the same arithmetic. Long-form production is expensive, slow, and structurally lossy at the per-title level. The streaming-era assumption was that scale, retention and subscription pricing would absorb that loss. That assumption is now under pressure across the industry. The strategic split is becoming visible. Some platforms, Netflix among them, are using AI tooling to make their own commissioned content cheaper while retaining the financing model. Others, iQiyi most clearly, are using AI tooling to leave the financing model behind. The first group is defending the existing economics. The second is rebuilding them.

iQiyi has gone first on the most aggressive version of the move. Take the production stack in-house through AI tooling. Open the surface to outside creators who fund their own projects. Extract a platform fee. Position the model as decentralisation while the architecture concentrates control of every layer that scales.

If the model works, the long-form content business changes shape. The category survives, but it stops being a commissioned-content business and becomes a platform-fee business. The companies that own the production tooling, the asset libraries and the distribution surfaces capture the value. The companies that previously sat in those positions as commissioning entities, but did not make the AI tooling investment, lose the layer they used to own.

Gong Yu put the framing precisely. AI is not changing distribution this time. It is “directly changing content itself.” That is correct, and it is also the part of the sentence that does the most work. The question this Substack has been asking for the past year is who captures the value when content production becomes orders of magnitude cheaper. iQiyi has now offered a clear answer. The platform that owns the stack captures it. The creator absorbs the rest.

-Maureen

(Listen to The Media Beat with Maureen & Claire for discussion around this topic)

Image created with OpenAI.

Netflix is usually described as the winner of streaming. That is true, but incomplete. Netflix has won by refusing to remain in the business it had just won.

Roughly every five to seven years, the company has changed its own category. DVD-by-mail became streaming. Streaming became original content. Original content became global subscription scale. Subscription scale became advertising and paid sharing. Now Netflix may be changing again, from a retention business into a yield business.

That distinction is what the market has not resolved. It understands the old streaming model: add subscribers, reduce churn, raise prices, spend on content. It is less clear how to value the next model: more visits, more ad inventory, more formats, more surfaces, and more monetisation per unit of attention.

Netflix’s history is useful here because its major pivots tend to look inevitable only after they have worked.

The pattern

The first Netflix was not a streaming company. It was a DVD-by-mail company. The innovation was convenience: no late fees, a deeper catalogue than the local video store, and discs delivered to the door. Netflix did not initially destroy the rental model. It made the existing model easier.

Then, in 2007, Netflix launched streaming. At first, it was a free add-on for DVD subscribers. It did not immediately replace the old business. It sat beside it. But Netflix understood earlier than most of the industry that the delivery system was changing. The internet would replace the postbox.

The Qwikster episode in 2011 is useful because it separates strategic direction from execution. Netflix tried to split the DVD and streaming businesses into two brands. Customers hated it. The stock was punished. Netflix reversed the decision. But the underlying call was right. DVDs were legacy. Streaming was the future. The execution failed; the direction held.

The next pivot was content. Early Netflix streamed other people’s shows and films. That worked while studios treated Netflix as incremental revenue. Once streaming became the future, those same suppliers became competitors. Disney, Warner, NBCUniversal and Paramount all wanted more control over their own libraries. Netflix had to become a producer. House of Cards in 2013 was the symbolic moment. Netflix was no longer only a distributor. It was becoming a studio.

That created the Netflix model of the late 2010s: spend heavily on content, drive viewing, reduce churn, add subscribers, and raise prices over time. Netflix became a retention business. Hours watched were not the goal in themselves. They were a proxy for habit. The more people watched, the less likely they were to cancel. Content spend created engagement. Engagement reduced churn. Low churn supported pricing power. Pricing power justified more content spend.

Then came the 2022 shock. Netflix lost subscribers, the stock collapsed, and the company was forced to do two things it had spent years saying it did not want to do. It launched advertising. It cracked down on password sharing. Both moves changed the model. Advertising allowed Netflix to serve price-sensitive users without discounting the whole product. Paid sharing turned non-paying users into paying customers, or pushed them off the system. Growth returned.

Those moves did more than repair the subscriber story. They prepared the next pivot.

From retention to yield

Netflix is no longer only asking how to keep people subscribed. It is asking how much value can be generated each time someone enters the product.

That is the shift from retention to yield.

Retention is about whether the member stays. Yield is about how much economic value Netflix can create from the relationship while the member is there.

The redesigned mobile experience points directly at this problem. The hardest moment in streaming is often the first one: the user opens the app without knowing what to watch. The catalogue is large. The recommendation system is sophisticated. The user still bounces because choosing feels like work.

Netflix’s answer is vertical discovery: short clips, autoplay, and a feed-like way to move through scenes before committing to a title. The company says it is redesigning the mobile experience and launching vertical video, while also expanding video podcasts, live events, games, and AI-enabled tooling.

That does not mean Netflix is becoming TikTok. Netflix is using vertical video as a discovery surface, not replacing its core viewing format with short-form vertical drama. Vertical is where Netflix wants viewers to find something. Horizontal is where it wants them to watch.

In the old model, a successful session was a long viewing session. In the new model, success begins earlier. Did the user open the app? Did they find something quickly? Did Netflix keep the attention inside its own environment? Did the visit create monetisable inventory? Did it increase the odds of another visit tomorrow?

That is why advertising is no longer a side project. Netflix says its ad plan represents more than 60% of Q1 sign-ups in countries where the ad tier is available, that its advertiser base grew more than 70% year over year, and that it expects roughly $3 billion in advertising revenue in 2026, about double 2025 levels.

At that point, advertising is not merely a new price tier. It changes the company Netflix is becoming.

What a yield business needs

A pure subscription company wants viewing because viewing reduces churn.

An advertising-supported company wants viewing too, but it also wants frequency, targeting, inventory, ad pricing, and measurable outcomes. It wants more reasons for users to open the app, more surfaces on which to sell attention, and more formats that can bring people back without requiring a full film or series commitment.

That explains the seemingly scattered moves: live events, video podcasts, games, vertical discovery, mobile redesign, and AI-enabled production tools.

The InterPositive acquisition was framed as creator tooling. That is fair. But inside this broader shift, it also looks like a supply-chain investment. If Netflix is creating more surfaces, formats, entry points, and visit occasions, it needs a cheaper, faster, more flexible content system. Reuters reported that Netflix acquired Ben Affleck’s AI film-tech company InterPositive as part of a wider set of growth initiatives after the Warner Bros. deal fell away.

This does not mean Netflix is abandoning premium television or film. The opposite is more likely. Premium content remains the anchor. But it is no longer the whole system.

The old Netflix question was: what show keeps you subscribed?

The new Netflix question is broader: what combination of shows, films, clips, live events, podcasts, games, recommendations and ads creates the most value from each member relationship?

That is not a streaming question. It is an attention-yield question.

What the Q1 reaction revealed

The headline numbers were strong. Netflix said Q1 revenue grew 16% year over year. Operating income grew 18%. Operating margin was 32.3%. The company reiterated full-year revenue guidance of $50.7 billion to $51.7 billion and an operating margin target of 31.5%.

Net income rose sharply, but the number was flattered by a $2.8 billion Warner Bros.-related termination fee, recognised in interest and other income. That made the quarter optically stronger than the underlying operating performance.

The market reaction was negative anyway. Part of that was the Q2 guide. Part was earnings quality. Part was Reed Hastings saying he will not stand for re-election to the board when his term expires at the annual meeting in June.

But the deeper issue is category confusion.

If Netflix is a mature subscription video business with an advertising tailwind, the valuation looks demanding. Mature subscription businesses eventually slow. There are only so many households, so many price increases, and so many hours in the day.

If Netflix is becoming an attention-yield business, the comparison set changes. The important metrics are not only subscriber growth and churn. They are ad load, ad pricing, visit frequency, discovery efficiency, live-event acquisition, gaming engagement, ad-tier mix, content-supply cost, and monetisation per member relationship.

That is a different business.

The Hastings transition is symbolically neat but operationally less important than it looks. Hastings stepped down as co-CEO in 2023. Netflix has already been operating under Ted Sarandos and Greg Peters. His departure from the chair role closes the original Netflix chapter, but it does not create the new strategy. The strategy is already visible in the product.

The pattern, again

That is usually how Netflix changes. The product moves before the market has agreed on the story.

The DVD company became a streaming company before the economics of streaming were obvious. The streaming company became a studio before the industry understood the scale of original content. The subscription company moved into advertising after years of resisting it. Now the streaming winner may be trying to become something broader: a system for monetising attention across formats, surfaces and price points.

The question is no longer just whether Netflix won streaming.

The question is whether streaming is still the right category.

If Netflix is a subscription business, the stock is expensive. If Netflix is a yield business, the market may still be early in understanding what it has become.

Each previous pivot took years to absorb. This one will not be obvious any faster.

-Maureen

Image created with OpenAI

(Listen to the discussion on The Media Beat with Maureen & Claire)

On April 15, a federal jury in Manhattan found that Live Nation and its Ticketmaster subsidiary illegally monopolized concert ticketing and live event promotion. A jury has now settled the liability question.

The market reacted, but only within limits. Live Nation shares fell 6.3% on the day of the verdict. A week later, they were still trading in roughly the same range. Rival platforms moved higher. Analysts mostly stayed constructive. The striking fact was not that the stock fell. It was that a definitive monopoly verdict produced nothing like a collapse.

That gap is the story. The verdict was legally decisive. The market’s response was conditional.

What the jury actually decided

The case, brought by 33 states and the District of Columbia, argued that Live Nation used the integration of its promotion business and its ticketing business to suppress competition. Ticketmaster controls roughly 86% of ticketing at major US concert venues. Live Nation’s promotion arm controls access to the artists and tours those venues depend on. The states argued that the company used the leverage of one business to reinforce the other. The jury agreed.

This converts a long-contested claim into a legal finding. For years, Live Nation’s dominance was debated as a matter of market structure and political rhetoric. It is now a matter of liability. That changes the company’s legal position even before any remedy is imposed.

Some of the evidence was vivid. One episode involved Barclays Center’s decision to leave Ticketmaster for SeatGeek. A former Barclays executive testified that Live Nation CEO Michael Rapino reacted furiously, and that the venue understood subsequent tour-routing decisions, including Billie Eilish skipping Barclays for UBS Arena after a cancelled tour returned, as retaliation. The government presented that as evidence of coercive conduct. The jury accepted the broader theory.

The broader finding, though, was not about a single venue or a single tour. It was about the integrated structure itself. The jury found that Live Nation used that structure in unlawful ways to preserve market power. That is a major legal loss. It is not yet a final answer to what the business will look like from here.

Why the stock barely moved

A six percent decline after a monopoly verdict is not the market shrugging. It is the market separating liability from remedy.

That distinction is now what drives everything.

Judge Arun Subramanian will hold a separate proceeding to determine what, if anything, should be done to change the structure or conduct of the business. The states have asked for serious remedies, including separating Ticketmaster from Live Nation, limiting how much of an artist’s touring Live Nation can promote, and barring Ticketmaster from participating in ticket resale. The judge has the power to impose meaningful relief. Any serious order will also be appealed.

That is why equity investors priced the verdict as a liability event, not a structural one. Three separate things still have to happen. First, the judge has to order a remedy that goes beyond symbolic or incremental conduct changes. Second, that remedy has to survive appeal. Third, it has to alter the economics of the business more than the March DOJ settlement already did.

The federal settlement already imposed meaningful constraints, including a 15% cap on service fees at Live Nation-owned venues, the divestiture of a limited number of exclusive booking contracts, and looser restrictions on non-Ticketmaster ticketing within Live Nation venues. That does not exhaust the remedy space. But it does narrow the range of additional outcomes that would materially surprise the market.

The stock’s reaction reflects that. Investors are not saying the verdict is trivial. They are saying the path from liability to structural change is long, uncertain, and likely to be more incremental than a monopoly headline suggests.

Will ticket prices come down

This is the question most readers will care about. The honest answer is probably not in any direct way.

For the biggest tours, the face value of a ticket is constrained less by monopoly in the abstract than by willingness to pay. If fans will pay a thousand dollars on the secondary market, the primary seller already has a signal about where demand clears. That was always the strongest part of Live Nation’s argument. Breaking up the company does not suddenly make fans less willing to pay market-clearing prices for Beyoncé, Taylor Swift, or Kendrick Lamar. What it changes is who gets to capture the surplus.

The real consumer issue is fees. Service fees, venue fees, processing charges, parking, upgrades, and all the other charges that accumulate around the ticket are where market power is felt most directly. That is also where policy intervention has the clearest line of sight. The March settlement’s 15% service-fee cap at owned venues is important for exactly that reason. It addresses the part of the ticket-buying experience most plausibly shaped by market structure rather than pure demand.

The states may still win additional transparency rules or tighter constraints on resale and bundling. Those changes could improve the buying experience and reduce some of the most aggressive fee capture. But they are unlikely to produce a world in which major concert tickets suddenly look cheap.

That is also how the market is reading the verdict. Vivid Seats and StubHub rose not because investors expect lower prices for fans, but because they expect a greater share of fee revenue and transaction flow to escape the Live Nation system. That is not a consumer-price story. It is a market-share story.

Will Live Nation be broken up

Live Nation trades at roughly 15 times next-twelve-month EV/EBITDA. CTS Eventim, the closest public comparable in the European live-events business, trades at under 7 times. That gap is the premium the market assigns to the integrated promotion-plus-ticketing model. If investors believed a breakup was probable in any near-term sense, that premium would have compressed much more aggressively than it has. Benchmark’s post-verdict research note called a forced breakup “still quite unlikely.”

There are straightforward reasons for that caution. Liability is now established, but remedy is separate. Structural relief is harder to obtain than a verdict headline suggests. Any breakup order would trigger years of appeals. The DOJ is not pursuing divestiture after its March settlement. The states can still pursue major relief, but doing so without federal alignment makes the path narrower and more complex.

None of that means a breakup is impossible. It means the market is still assigning it a low probability.

What the verdict actually changes

The verdict does not change the structure of the business or the price of a ticket. It changes posture.

Live Nation no longer operates as a company whose dominance is alleged. It operates as a company a jury has found to have unlawfully maintained it. That affects every negotiation with venues, every conversation with independent promoters, and every future private case built on the same factual record. It changes the baseline from which the company now has to defend itself.

What it has not yet changed is the part of the story public markets care about most: durable cash flows under a new remedy regime.

The market’s current answer is that not much will change unless the remedy phase produces something stronger than investors currently expect. That answer is visible in a stock down only modestly after a historic loss and in a valuation that still prices the integrated structure as unusually valuable.

That answer could still turn out to be wrong. But for now, the jury decided liability. The economics are still waiting on the judge.

-Maureen

Image created using OpenAI

What that reaction did not reflect was a weakening operating business.

The quarter itself was strong. Netflix reported $12.25 billion of revenue, up 16% year over year, with operating margin at 32.3%, above the company’s own guidance. Management maintained its full-year framework of 12% to 14% revenue growth, 31.5% operating margin, and about $3 billion of advertising revenue.

The call added operating detail that the stock action obscured. Greg Peters said Netflix’s primary internal member-quality metric hit another all-time high in Q1 and that retention improved year over year in every region. He also said view hours grew at a similar rate to the second half of 2025 despite 17 days of Winter Olympics competition.

So the sell-off was not really about Q1. It was about how investors chose to compress three separate issues into one story.

The first was the Q2 guide. Netflix told investors that the second quarter will carry the highest year-over-year content amortization growth of the year before decelerating to mid-to-high single-digit growth in the second half. That is the key context for the weaker near-term margin outlook. It may be disappointing, but it is not the same as abandoning the full-year plan. Netflix did not change that plan.

The second was EPS optics. Q1 diluted EPS was boosted by the $2.8 billion Warner Bros. termination fee, which Netflix recognized in interest and other income. That made the headline quarter look cleaner than the underlying run rate and made Q2 look weaker by comparison. That is a real issue for the tape, but it is not evidence that the operating model has deteriorated. Reuters noted that the breakup fee materially lifted quarterly earnings.

The third was Hastings. Founders are not valued like ordinary directors, and markets do not price symbolic loss cleanly. But on the call, Ted Sarandos said Hastings’ departure had “absolutely” nothing to do with the failed Warner Bros. transaction, and the company has already been operating under the current leadership structure since Hastings stepped down as co-CEO in 2023. Reuters also noted that he will remain through his current term.

The more important story in the quarter sat underneath all of that: Netflix is changing what it monetizes against.

For years, the market’s rough model of Netflix was simple. Spend on content, drive hours, hold subscribers, raise price. Peters is now describing a more complex system. He said view hours are still relevant, but they are no longer the only or even primary way Netflix evaluates value creation. He pointed instead to a proprietary member-quality metric that management says correlates better with retention than raw hours do.

The implication is that newer formats no longer have to fit the old framework. On the call, Netflix said the World Baseball Classic reached 31.4 million viewers, became its most-watched program ever in Japan, drove its largest single sign-up day there, and helped make Japan the company’s strongest contributor to member growth in the quarter. Those outcomes are strategically significant even if they do not maximize hours in the way a global scripted hit might.

The ad business sits on the other side of that same reframe. Live events and podcasts produce premium inventory that scripted libraries structurally cannot: higher attention, appointment viewing, mobile reach. A format that generates fewer view hours but better ad attachment has a different monetization ceiling than raw engagement math suggests. That is the logic underneath the spend. This is already visible in advertising: Netflix reiterated its target of about $3 billion in ad revenue this year, and management said programmatic is on its way to becoming more than 50% of its non-live ads business. The advertiser base grew over 70% in 2025 to more than 4,000 advertisers.

That is why the quarter looks different depending on whether you view it through the stock or the business. NFLX traded on a light Q2 guide, messy comparisons, and founder symbolism. Netflix, the company, showed continued revenue growth, stable full-year discipline, stronger retention, and early signs that its next monetization layer is becoming real. It showed a business becoming more complicated to value than the market’s first reaction allowed.

-Maureen

In February, I argued that the fastest way to misunderstand vertical drama is to confuse a format with a business model. The confusion starts when people use “microdrama” as shorthand for the entire category, because the term often smuggles in assumptions about gating, payment mechanics, and user acquisition that do not apply uniformly across the market. I separated the category into three distinct markets: app-gated IAP platforms, platform-native social series, and streamer-surfaced discovery content. Format is not a business model. That was the argument.

PineDrama is now making that argument more forcefully than I did.

When TikTok launched PineDrama in January, much of the coverage framed it as a platform entry story: another player entering a crowded field, ByteDance bringing its distribution muscle to a format that has been growing quickly. That framing is not wrong. But it understates what is happening. TikTok has not simply entered one of the three buckets I described. It has taken the format and declined to choose a bucket at all. That is a different move, and it puts pressure on all three simultaneously.

What PineDrama actually is

The format credentials are unambiguous. PineDrama offers bite-sized fictional series told through one-minute vertical episodes in a feed that behaves like TikTok’s core product, except every video is a scripted installment in an ongoing story. Serialised, vertical, swipe-driven, cliffhanger-paced: by any functional definition, this is microdrama.

The divergence is in the business model. While competitors rely on pay-per-episode mechanics, PineDrama is free. Its objective is not conversion but immersion: keeping viewers binging complete series inside the ByteDance ecosystem. No paywall, no subscription prompt, no unlock funnel, none of the mechanics that define the IAP-gated half of the market.

That is not just a pricing decision, it is an architectural one. TikTok has taken the format intact and stripped out the business model assumptions that helped build the category. In February, I argued that app-gated microdrama behaves less like television than mobile gaming: performance marketing, conversion funnels, payer cohorts. PineDrama offers a similar content experience without asking the user to convert at all.

The more consequential signal is what sits alongside it. TikTok has filed a trademark for “TikTok Drama” covering everything from webisodes to short series, and it has begun casting actors for original soap-style content. This is not just a distribution move. It is the beginning of a content ownership play. TikTok is not only building a pipe for other people’s vertical drama. It is positioning to own what flows through it.

Why the free model is the structural threat

The incumbents have built real businesses. ReelShort generated $1.2 billion in gross consumer spending in 2025, up 119% year on year. DramaBox brought in $276 million, more than doubling its 2024 revenue. Those are real numbers. The IAP model works.

But it works under specific conditions: users must accept a paywall for content they cannot get elsewhere, and the cost of acquiring those users must remain manageable relative to lifetime value. PineDrama applies pressure to both conditions at once.

On the content side, the immediate question is not whether PineDrama is better than ReelShort. It is not. The question is whether free, algorithmically distributed vertical drama eventually erodes willingness to pay for unlock-based viewing when both experiences sit within the same discovery environment. Audiences do not make principled choices about monetisation architecture. They make choices about friction. PineDrama has almost none.

On the acquisition side, the pressure is even clearer. TikTok already accounts for 19% of paid user acquisition spend for leading microdrama platforms, second only to Facebook at 25%. Platforms that built meaningful businesses on TikTok traffic are now competing with TikTok for the same audience, on TikTok’s own infrastructure, in a format TikTok is now offering for free. The acquisition logic of the IAP model assumes a relatively stable advertising environment in which you pay to reach potential subscribers. That environment changes when the platform selling you reach is also your direct competitor.

In February, I described the fragile part of the app-gated model: customer acquisition costs that have to remain stable as scale increases, payback periods that have to remain tolerable under competitive auctions, and content costs that cannot outrun lifetime value. PineDrama does not break those mechanics outright. But it tightens all three constraints at once by introducing a free alternative at the top of the same funnel.

What this does to the taxonomy

The three-bucket framework I proposed in February was useful because it separated markets the industry kept conflating. PineDrama is the first significant move that cuts across all three and puts each under a different kind of pressure.

The IAP bucket assumed that platforms could control their own acquisition economics and sustain paywalls because content loyalty would outweigh switching costs. PineDrama tests both assumptions. The platform-native bucket assumed that social distribution and content ownership were separable: studios owned the shows, platforms owned the pipe. TikTok’s trademark filings and original casting activity begin to change that relationship. The streamer-surfaced bucket is less immediately threatened; Netflix and Disney+ are not competing with PineDrama for exactly the same user moment. But Netflix LATAM has already drawn an explicit distinction between short-form content and vertical-format adoption, and that position becomes harder to sustain if TikTok succeeds in building original vertical IP that audiences begin to expect across surfaces.

That does not make the taxonomy wrong. It makes it time-bound. It described the market accurately at a particular moment. That moment is now changing.

The Issa Rae signal

The clearest current evidence of that shift is Issa Rae. Her production company, Hoorae Media, has committed to co-develop multiple micro-series with TikTok on PineDrama, not as a one-off experiment but as a structured original slate and an ongoing creative relationship.

What matters here is not celebrity validation. It is production alignment. Issa Rae is not appearing in somebody else’s vertical series. Her company is building original content inside a platform ecosystem through a sustained development relationship. That is a different arrangement altogether. It is the difference between a studio selling to a distributor and a studio operating, in effect, inside one. The creative infrastructure is being built within TikTok’s system, not merely distributed through it.

For independent vertical studios, especially those that currently rely on TikTok’s recommendation engine for discovery, that raises a new question. If the platform starts commissioning originals and attracting serious production partners, what position does the independent studio occupy in that ecosystem? A valued supplier, or a lower-priority source competing with the platform’s own slate?

That is not a crisis yet. But it is the shape of the structural question now facing both the IAP and platform-native models.

The ByteDance playbook

None of this is especially surprising if you have watched ByteDance in China. The company has already expanded into microdrama in Asia through platforms such as Melolo and Red Fruit, applying lessons from Douyin and other regional experiments to Western markets. In China, this consolidation has already happened: the dominant short-video platform formalised the category, built original content infrastructure, and became both the distribution layer and a meaningful content owner. PineDrama looks like that playbook exported westward.

The difference in the West is that the IAP platforms arrived first and built substantial revenue before TikTok moved decisively. That gives them more runway than some Chinese equivalents may have had. But it does not alter the direction of travel. When the largest distribution infrastructure in short-form video decides it wants the content layer as well, every other player in the category faces the same question: what is the defensible position once the pipe and the content are owned by the same company?

What the taxonomy needs now

In February, I argued that the most likely medium-term shape of the industry is hybrid distribution: IP designed to travel across multiple forms, supported by deal structures that assume movement across surfaces. I still think that is right. But PineDrama adds a fourth consideration that the original taxonomy did not fully account for: platform capture.

The three buckets described how content is monetised. They did not fully describe what happens when the dominant distribution platform vertically integrates into content ownership. For content companies, that is not a business model choice. It is a structural condition.

Format is not a business model. I still believe that. But distribution infrastructure, it turns out, is increasingly becoming one.

-Maureen

Image created with OpenAI

That reaction is understandable. It is also where the story becomes too neat.

The market is pricing Adobe as a legacy creative software company facing AI pressure on a seat-based model, while missing the more important strategic shift now underway: its move into the orchestration layer between AI models and enterprise content workflows.

The bear case is real

The bear case has real force. If AI makes creative work easier and cheaper, Adobe’s historic moat looks less secure. If designers become more productive, companies may need fewer seats. If simpler platforms keep spreading through teams that do not need the full complexity of Photoshop or Illustrator, Adobe can start to look less like essential infrastructure and more like an overbuilt incumbent. The failed $20 billion Figma acquisition still hangs over the story as a symbol of either strategic foresight or competitive anxiety, and the market has mostly chosen the darker reading. Reuters noted that investor skepticism has persisted even as Adobe continues to beat expectations, because the concern is not this quarter alone but whether Adobe’s AI strategy can offset longer-term pressure on the old subscription model.

The numbers do not look like collapse

The problem with that verdict is that Adobe’s actual numbers still tell a more resilient story. In Q1 FY2026, Adobe reported record revenue of $6.40 billion, up 12% year-on-year, with non-GAAP EPS of $6.06 and total ending ARR of $26.06 billion, up 10.9% year-on-year. Its trailing twelve-month free cash flow reached $10.3 billion. These are not the headline figures of a company already in visible financial decline. They are the figures of a large, profitable software business still growing at double digits.

The more interesting point is what is happening inside those numbers. Adobe disclosed that Firefly ending ARR exceeded $250 million, that generative credit consumption grew more than 45% quarter-on-quarter, and that its creative freemium user base crossed 80 million monthly active users, up more than 50% year-on-year. Annualized revenue from AI-first products more than tripled year-on-year. That does not look like a company simply waiting to be disrupted. It looks like one trying to turn AI into a new revenue layer inside the business.

The orchestration layer is the overlooked story

The underappreciated move is not only financial resilience. It is Adobe’s attempt to occupy the orchestration layer in enterprise AI: the middle position between the models enterprises want to use and the workflows they actually need to run. Adobe described its own strategy in a February SEC filing as building partnerships “across all four layers of the AI technology stack spanning data, models, agents and interfaces,” naming AWS, Google Gemini, Microsoft Azure, Nvidia and OpenAI among its partners. That is not the language of a company retreating into a legacy niche. It is the language of one trying to sit across the stack rather than inside a single point product.

The partnership map shows the strategy

The partnership map helps clarify the position Adobe is building. Adobe and Nvidia announced in March that they are working together on the next generation of Firefly models and on agentic workflows for creativity, productivity and marketing. Nvidia separately said Adobe would adopt its Agent Toolkit as a foundation for running long-running Adobe agents, placing Adobe inside the emerging enterprise agent infrastructure.

The WPP partnership points in the same direction. Adobe and WPP said in February that Firefly Foundry will be integrated into WPP Open, combining Adobe’s custom-model and content capabilities with WPP’s marketing and activation workflows. Adobe is moving closer to the operating layer through which large brands create, adapt, and deploy campaign content at scale.

Its OpenAI relationship reinforces the same strategic posture. Reuters reported in December that Adobe brought Photoshop, Express and Acrobat into ChatGPT, extending its tools into one of the most important consumer AI interfaces. Adobe also said in February that it was a pilot partner in OpenAI’s ads program. Taken together, those moves suggest Adobe is trying to be present wherever AI-mediated creative and marketing work is happening: inside its own products, inside enterprise orchestration, and inside external interfaces where users are increasingly starting work.

The business model is shifting with the stack

The simplest version of the negative thesis was always seat compression: fewer designers needed means fewer licences sold. But Adobe is moving away from pure seat economics toward a hybrid of seats and usage. When enterprises scale Firefly and related APIs, generative credits are consumed. As AI workflows move from experiment to production, revenue can scale with usage intensity rather than just headcount. Adobe’s more than 45% quarter-on-quarter rise in generative credit consumption is the revenue model shifting, not a product statistic.

The moat is deeper than the tools

The enterprise moat is also more structural than the public narrative often allows. Large organizations have built workflows, archives, approvals, and brand systems around Adobe formats and tools over decades. Adobe has widened that embedding by selling not just creative applications but a broader content and customer-experience stack. In its earnings materials, Adobe highlighted AI-powered capabilities across creativity, productivity and customer experience orchestration, and positioned GenStudio and Adobe Experience Platform as part of that broader system. Displacement is not just about whether another tool can generate an image. It is about whether another vendor can replace the workflow layer as well.

Adobe also retains a meaningful trust advantage in enterprise AI. The company has repeatedly emphasized commercially safe deployment and customer-specific model building through Firefly Foundry. Enterprise buyers are not choosing AI tools solely on output quality. They care about governance, brand consistency, workflow fit, and procurement comfort.

The risks are real

None of this makes the risks imaginary. The CEO transition is real uncertainty. Reuters reported there is no named successor yet, which leaves the market to wonder whether Adobe is entering a new phase of acceleration or a more defensive phase of adaptation. Competitive pressure is also real, especially over a longer horizon, as newer users build creative habits in lighter-weight tools. And the orchestration strategy itself is execution-heavy: maintaining a neutral position across multiple powerful partners only works as long as each sees Adobe as more useful inside the stack than outside it.

What the market may be missing

Still, the market is pricing Adobe primarily as a legacy creative-software company facing AI disruption to a seat-based model. That framing captures something real, but it misses the larger move in progress. What Adobe is actually building looks more like a brand-safe, workflow-embedded orchestration layer connecting models, agents, interfaces and enterprise content operations. The financials do not disprove that thesis. So far, they support it: $26.06 billion in ending ARR, $6.40 billion in quarterly revenue, more than $250 million in Firefly ending ARR, and sharp growth in usage-driven AI metrics.

Adobe Summit runs April 20 to 22, with preconference on April 19, and should offer a clearer view of how aggressively the company intends to push this AI-native platform story. The CEO selection process will matter too. But for now, the cleanest conclusion is this: Adobe does not look like a company already buried by AI. It looks like a company that saw the threat early, is repositioning around the middle of the stack, and is being judged by a market that has not yet fully caught up with what it is building.

-Maureen

Image created with OpenAI.

Maureen Kerr is the founder of Kerr Partners, a media industry strategy and research consultancy. This piece is analytical commentary and does not constitute investment advice.

On April 2, OpenAI announced it had acquired the show. Terms were not disclosed, though the Financial Times reported a price in the low hundreds of millions of dollars.

A lot of the coverage framed this as an unusual move: an AI company buying a media property. That description is true, but it misses the more important point.

What TBPN was actually worth

TBPN’s value was never its revenue. Thirty million dollars in advertising is immaterial next to OpenAI’s latest funding round. Access was the real asset, and so was the atmosphere that made that access possible.

TBPN had become a venue where some of the most powerful figures in tech would sit down and speak with a level of candour they rarely bring to traditional media. That candour depended on the show feeling like a peer conversation rather than a formal interview. Coogan and Hays are former founders; their guests were not walking into an adversarial press setting so much as joining a conversation among people who understood the room.

That dynamic is difficult to manufacture and harder still to scale. It is not something a communications team can simply recreate on command.

What OpenAI acquired, then, was not just a content business. It acquired a venue that had already become habituated as a place where influential people were willing to speak freely.

The editorial independence problem

OpenAI has been careful in how it describes the deal. TBPN, it says, will maintain editorial independence. An Editorial Independence Covenant will limit the company’s ability to influence coverage. The hosts will continue to choose their own guests and run their own programming.

Then there is Chris Lehane.

TBPN will report into Lehane, OpenAI’s chief political and communications strategist. That detail matters. Lehane is not a neutral steward of media pluralism; he is a veteran operator in reputation, messaging, and institutional positioning. His oversight does not prove direct interference. But it does make clear that this acquisition is strategic, not incidental.

The problem with editorial-independence covenants is not that they are meaningless. It is that they govern the wrong thing. The pressure that changes a media property is rarely an explicit instruction. More often it is the gradual shift in atmosphere that comes when everyone involved knows who owns the room. Guests become more selective. Hosts become more self-conscious. Certain topics acquire friction. Nothing needs to be said aloud for the editorial environment to change.

That is the difficulty for TBPN. Its value depended on an unusual degree of ease. Ownership does not automatically destroy that, but it does complicate it.

This movie has been made before

The most useful comparison is not Robinhood launching Sherwood or Plaid acquiring This Week in Fintech. It is Elon Musk buying Twitter.

When Musk acquired Twitter in 2022, the public framing was ideological: free speech, platform openness, a challenge to legacy moderation norms. Structurally, though, something simpler had happened. The owner of several companies that were themselves subjects of intense public scrutiny had acquired one of the main platforms through which journalists, commentators, and insiders distributed information and shaped perception about them.

The consequences did not arrive only through formal directives. They emerged through product changes, incentives, reinstatements, monetisation choices, and the altered behaviour of everyone using the platform once the owner’s interests became impossible to ignore.

The TBPN deal is a smaller, more targeted version of the same logic. Twitter was a mass platform. TBPN is a precision instrument: a high-trust venue where the AI industry’s most important figures talk to one another in public.

That is also why the risk to TBPN is immediate. Its value depended on credible outsiders continuing to treat it as neutral enough to enter. Now every guest has to decide whether appearing on the show means appearing inside OpenAI’s communications perimeter. The erosion, if it comes, probably will not look dramatic. It will look like delayed replies, diary issues, softer bookings, and the slow thinning-out of the very access that made the asset valuable.

The other side of the squeeze

The pressure is not all coming from tech companies buying in. It is also coming from the financial fragility of the media businesses being bought.

That is what makes Vox interesting. Vox Media remains home to some of the most credible personalities in tech and business media, including Kara Swisher and Scott Galloway. Last year, its bankers explored options including a sale of the podcast network, and now Versant, the newly formed parent of CNBC, is reportedly in talks to acquire the company. Whether or not that deal closes, the broader pattern is visible enough: the independent middle is under pressure.

At one end, companies with enormous balance sheets are buying access, audience, and narrative leverage directly. At the other, media companies facing capital constraints are being pushed toward larger corporate structures, each with their own priorities, sensitivities, and incentives.

The result is not the disappearance of independent media. It is the shrinking number of high-access venues that remain both influential and structurally independent.

But isn’t this just a cycle?

There is an obvious objection. Every wave of consolidation produces a counter-wave of independence. Journalists and hosts leave institutions, build direct relationships with audiences, and publish on Substack, YouTube, podcasts, and other distribution platforms. In many ways, that world is healthier and more accessible than the old one.

But that is not really the issue here.

Distribution is no longer scarce. Access is.

An independent commentator can build a large audience outside legacy media. What is much harder is building the kind of trust that leads the most powerful executives in tech to speak candidly, repeatedly, and in ways that generate news. That kind of access is venue-specific, socially accumulated, and difficult to transplant.

So yes, independence will continue to flourish at the level of commentary and analysis. What is becoming harder to preserve is the small number of rooms where the people shaping the industry still feel comfortable speaking before the rest of the market has decided what to think.

Owning the room

That is why the TBPN acquisition reveals most clearly.

It does not sit in isolation. It fits a broader OpenAI pattern: owning more of the interface, more of the distribution, more of the direct relationship with users, and now a venue that helps shape elite conversation around the company and its sector.

That is what makes “media acquisition” too weak a description. This looks less like a content bet than a control-point move: acquiring a strategically important position between the product and the public story about the product.

And the timing makes the logic easier to see. OpenAI is moving toward the next phase of its corporate life under intense scrutiny around valuation, competition, regulation, and legitimacy. In that environment, narrative infrastructure is not peripheral. It matters directly to enterprise value.

The independent voices will keep publishing. New ones will emerge. Commentary will remain abundant.

But the rooms where powerful people speak candidly are becoming assets in their own right. And once those rooms are treated as assets, they begin to get valued, and acquired, accordingly.

-Maureen

Image created with OpenAI

Did the models train on copyrighted work without consent? Would rightsholders sue? Could a licensing framework be built that made generative music commercially legitimate rather than legally radioactive?

ElevenLabs is betting that the answer is yes. Its Music Marketplace, built on licensing deals with Merlin and Kobalt, is an attempt to do AI music the proper way: opt-in, compensated, guarded, legible. In an industry still dealing with the fallout of lawsuits against Suno and Udio, that alone is enough to make the launch significant.

But it also makes the launch more revealing.

Because once an AI music platform starts to solve the consent problem, the industry runs into the harder question underneath it: what happens when generative music is not just legally cleaner, but actually usable at scale?

That is the real significance of ElevenLabs’ move into music. Not that it has built another text-to-music model, but that it is trying to build a governable one.

Founded in 2022, ElevenLabs became one of the dominant names in AI voice by turning text-to-speech and voice cloning into infrastructure used across media, publishing, entertainment, and enterprise. Its move into music was presented as a natural extension of that broader audio identity, but it is also something more consequential than adjacent product expansion. It is an attempt to apply the company’s existing marketplace logic — already visible in its voice products, where creators can license synthetic voice replicas in exchange for royalties — to music generation itself.

The timeline has been unusually compressed. Eleven Music launched in August 2025. By March 2026, Music Marketplace was live as a commercial layer on top of the generation engine. This is not a speculative side project. It is capital flowing toward the construction of a licensed AI media stack.

The obvious comparison points are Suno and Udio, whose rise helped define AI music as a copyright confrontation. ElevenLabs is trying to arrive at a similar destination through a different route. Where others launched first and litigated later, it has licensed first and wrapped the model in contractual architecture.

That architecture is the story.

The most important deals are with Merlin and Kobalt. Merlin, which represents 30,000 independent labels and distributors, gives participating members an opt-in pathway to allow their works to be used in AI training in exchange for royalties. Kobalt brings publishing into the structure. Together, the two agreements create something the AI music sector has mostly lacked: a framework that treats music generation not simply as an extraction event, but as a rights market.

That does not mean the framework is comprehensive. It is not. But it does mean ElevenLabs has moved the conversation.

The Kobalt deal in particular contains two provisions that matter well beyond this product.

The first is parity. Revenue from Eleven Music is split 50/50 between compositions and recordings used to train the model. That may sound technical, but it cuts directly against one of the defining economic distortions of the streaming era, where publishers and songwriters have routinely ended up with worse terms than labels. AI licensing could easily have reproduced that imbalance. Instead, Kobalt appears to have forced a different principle into the center of the deal: that the song itself is not a secondary input into the model, but a co-equal source of value.

The second is the most-favoured-nation clause. If any recorded-music rightsholder later negotiates better terms than Kobalt, Kobalt’s deal automatically upgrades to match them. That is not just a defensive measure. It is a pressure point. It means any future dealmaking with larger rightsholders happens under the shadow of publishing parity.

And that is where this gets interesting.

For the past year, the music business has often treated AI as a threat when it arrived without permission and as an opportunity the moment it became licensable. Litigation and licensing have increasingly become part of the same negotiation cycle.

ElevenLabs, by contrast, does not yet have the majors inside its system. That absence is both a limitation and a signal.

It is a limitation because, without major-label catalogues, the training base is narrower and the product remains less powerful for mainstream commercial use cases. Users can prompt by style rather than artist name, but the absence of the biggest repertoires still matters. If you want to become the default infrastructure for commercially useful AI music, rights breadth matters.

But the majors’ absence is also a signal because it exposes the tension inside ElevenLabs’ approach. The company may have built a cleaner and more principled licensing architecture than some rivals, but precisely because of clauses like parity and MFN, it may also be offering recorded-music rightsholders less room to dominate the economics. A system designed to be more balanced may prove less attractive to the biggest players used to extracting imbalance.

That is why the legal structure matters so much. It does not just reduce risk. It redistributes leverage.

For independent labels, this is where the story is easiest to read as good news. Merlin members get a new revenue stream, opt-in participation, and a framework built around formal consent rather than retroactive damage control. For a sector that has spent the past two years watching AI companies build on copyrighted culture while insisting that licensing could come later, that is not nothing. It matters that one of the first scaled attempts to commercialise generative music through licensing has given independents a way in.

But good news at the level of rights holders is not the same thing as safety at the level of working musicians.

That is the contradiction that keeps surfacing across the AI music economy. A platform can become more ethical in the narrow sense of permission and compensation while still intensifying the broader conditions that make creative work more precarious.

The most obvious pressure point is saturation. Even if ElevenLabs is more carefully licensed than its rivals, it still contributes to a market in which music becomes dramatically easier and cheaper to produce. For independent artists already competing inside an oversupplied attention economy, a flood of generative music does not need to be artist-specific to be destructive. It only needs to be abundant.

The first part of the market likely to feel that pressure most acutely is production music: stock audio, background cues, commercial-use tracks, functional music made to specification rather than fandom. This is exactly the terrain ElevenLabs is targeting. Brand-safe background music, game soundtracks, social content, advertising — these are not marginal use cases. They are the practical center of the music licensing economy. They are also the categories where speed, price, and flexibility often matter more than authorship.

That exposure is not abstract. CISAC’s (The International Confederation of Societies of Authors and Composers) economic study projected that, under current conditions, 24% of music creators’ revenues could be at risk by 2028 as generative AI scales into streaming and music-library markets.

In other words, the most commercially plausible use cases for properly licensed AI music map almost perfectly onto the sectors where human composers have historically made a living by being reliable, adaptable, and fast.

If the benchmark is whether rightsholders consented, ElevenLabs is making a far more serious effort than much of the field. If the benchmark is whether songwriters get treated as equal contributors to model value, the Kobalt structure may even represent progress relative to streaming. But if the benchmark is whether musical labour as such is being protected, the picture is much less reassuring.

Session musicians are the clearest example. They often work on a work-for-hire basis and do not own copyright in the performances they create. That means even when labels and publishers negotiate compensation frameworks for AI training, a large class of working musicians remains outside the protection perimeter. Their labour may help make the recorded ecosystem valuable, while the rights structure routes payment elsewhere. The platform can be fully licensed and still leave them exposed.

For all the focus on whether AI companies are stealing from artists, the deeper shift may be that they are learning how to pay the people with legal rights while still destabilising the people whose labour sits underneath those rights. Once that happens, the argument changes. It is no longer mainly about infringement. It becomes a question of substitution, bargaining power, and which kinds of music work remain economically defensible once generated supply becomes cheap, licensable, and endless.

Even the copyright issue around AI-generated outputs points in that direction. Value accrues around the system, but not necessarily around a creator in the conventional sense.

That is why ElevenLabs matters: it forces a more mature version of the argument.

It is easy to denounce AI music when the companies involved appear cavalier about copyright. It is harder when one of them starts building the system critics said they wanted: licensed inputs, royalties, guardrails, restrictions on outputs, commercial clarity. Once that exists, the industry loses the comfort of treating ethics as a simple matter of consent.

Consent matters. Compensation matters. A licensed model is better than an unlicensed one. But a licensed model does not solve the economic shock of generative media. In some respects, it may accelerate it by removing the legal friction that kept mainstream commercial adoption slower and messier.

That is the hinge point here. ElevenLabs may be building the most legitimate version of AI music so far. It may even help establish precedents that songwriters and publishers should want to defend, especially if publishing parity holds. But legitimacy is not the same as protection.

Licensing may settle the copyright argument without settling the labour argument.

And once AI music can be licensed, monetised, and folded into the ordinary rights economy, the real question is no longer whether it belongs inside the market. The question is what that market now rewards, what it makes disposable, and whether “responsible AI” is really a safeguard or just a smoother path to the same underlying displacement.

ElevenLabs is the clearest sign yet that the argument is moving on, not away from copyright, exactly, but past the point where copyright alone can explain what is at stake.

-Maureen

Image created using OpenAI