Netflix is already using AI in hundreds of titles this year alone
Netflix’s Q2 earnings letter disclosed that roughly 300 titles used generative AI in 2026, across concept, previz, and post. The disclosure went to shareholders, where it reads as good news. Nobody put it in the credits. Here’s why almost every studio does the same thing, and the number that explains it.
Netflix used generative AI on roughly 300 titles this year.
The company said so itself, on July 16, in its second-quarter earnings letter. Not in a press release, not in a credit crawl. In the document written for people who own the stock.
That’s the whole story, and it’s a bigger one than Netflix.
What Netflix actually disclosed
Per Variety, Netflix told shareholders that generative AI usage now “expands across every level of a program’s production process, from its concept and pre-visualization to post-production and release.”
Read that twice, because a lot of coverage is calling this a post-production story and it isn’t. Concept and previz means AI is in the room while people are deciding what the thing will look like.
The company named names: Glory, an Indian sports thriller series, Brasil 70: A Saga do Tri, a Brazilian soccer miniseries, and The American Experiment, a docuseries about the American Revolution. AI built “highly complex sequences” in all three, enhanced crowd sizes, historical battle sequences, worldbuilding establishing shots.
The money line, straight from the letter: “We are increasingly leveraging these tools to deliver higher-quality output more quickly and at a lower cost than traditional methods.”
Ted Sarandos told the earnings call that one documentary sequence came in twice as fast at half the cost. Netflix added that some large-scale scenes might never have been finished at all without the tools.
The uses are boring, which is the point
Nobody’s writing scripts with this. Nothing in the disclosure describes AI performing, directing, or generating a story.
It’s crowds. Backgrounds. Battle extras. The establishing shot of a city that used to cost six weeks and a matte painter. The unglamorous, expensive middle of the pipeline where a line producer is trying to make a budget survive contact with an ambition.
Which is exactly why 300 of them slipped past everyone until an accountant wrote it down.
Netflix told the one audience that likes this
Here’s the part worth noticing.
Netflix didn’t hide it. Netflix bragged about it. But look where: a shareholder letter, on earnings day, alongside $12.56 billion in quarterly revenue and 97 billion hours watched.
In that room, “cheaper and faster” is the point. Investors hear 300 titles and hear margin.
Now imagine the same sentence in the opening titles of Glory. Same fact. Same technology. Completely different reception.
So the disclosure wasn’t a disclosure to you. It was a disclosure to them, and you happened to be able to read it.
The number that explains why nobody tells you: 53%
Here’s the receipt, and it’s the most important figure in this story.
Steam requires developers to disclose generative AI use. It’s mandatory. And per MobileSyrup, games that carry that disclosure have reportedly lost up to 53% of revenue once players see the tag.
Half the money. For telling the truth.
That’s not a culture war. That’s a market signal, and it’s the loudest one in this entire debate. If honesty costs you half your revenue, honesty becomes a luxury good, and only companies that don’t need the money can afford it.
Netflix can afford it. Netflix doesn’t sell you individual titles, it sells a subscription, and nobody cancels over the crowd extras in a Brazilian soccer miniseries they’ve never heard of.
A studio selling one $60 product on Steam cannot afford it. So they don’t.
This isn’t a Netflix problem, it’s the whole industry
The pattern’s been sitting in plain sight for months.
A JOGA survey of Japanese online game companies this month reported 100% generative-AI adoption among respondents, mostly for user-preference analysis and behavior prediction. A Harris Poll of 600-plus studios across the US, South Korea, and Europe put adoption at 87%.
And a survey of Japanese creative professionals found 59% of companies used AI, of which 71.4% did not actively disclose it.
Jack Buser, Google Cloud’s global director for games, said the quiet part in April: “I think what players don’t realise is that their favourite games right now were already built with AI. Those games have shipped.” His read on why survey numbers vary so wildly, from 40% to 100%? “That gap is basically the developers’ willingness to tell you.”
He sells AI tools for a living, so weigh that accordingly. He also appears to be right.
The bad equilibrium nobody’s naming
Strip out the yelling and there’s a genuine structural problem here, and it should bother both camps.
If you’re pro-AI, the 53% number is your nightmare. It means the technology can never be normalized honestly, because the first company to be transparent gets punished for it. You’re stuck with an industry that uses your favorite tool and won’t defend it in public.
If you’re anti-AI, it’s worse. Non-disclosure means you’re already consuming the thing you object to and can’t tell which thing it is. You can’t boycott what nobody labels. The market can’t register your preference if the market can’t see it.
Both sides want the same thing for opposite reasons: labels. And the current arrangement guarantees nobody gets them, because the only companies that disclose are the ones with nothing to lose.
Netflix is spending real money on this
None of this is a phase. In March, Netflix acquired InterPositive, the AI startup founded by Ben Affleck, for up to $600 million, bringing a 16-person team in-house with Affleck himself as a senior advisor on filmmaker-facing tools.
The company is also building an AI-focused animation studio. Its 2025 series The Eternaut reportedly finished its AI-assisted VFX ten times faster than a traditional pipeline would allow, which is presumably how you get from one proof of concept to 300 titles in about a year.
Sarandos told Politico in March that AI “should be a creator tool,” comparable to other production tools that evolved over time, and that it could only make things “faster and cheaper” if the quality bar held.
The counterargument is equally real and it isn’t going away: the models trained on human work without permission or payment, the data centers have an environmental cost, and every crowd scene an algorithm fills is a crowd scene somebody used to get paid to build. VFX artists aren’t wrong to notice that “higher-quality output more quickly and at a lower cost” is a sentence about them.
What 300 actually means
Three hundred titles in roughly six months isn’t an experiment. It’s a standard operating procedure that arrived without an announcement, and the only reason anyone knows the number is that a public company had to file a document.
Every other studio in the business is doing some version of this. Most of them will not be putting a number on it, because the number costs money, and the audience has already demonstrated exactly how much.
You’ve probably already watched several of them. That’s not a warning. It’s just a fact, and the fact that it’s a fact is the story.
Nobody’s going to tell you which ones. Ask yourself why that is, and then look at the 53%.
Want More Clownfish TV?
This article was brought to you in part by The Reefers of more.clownfishtv.com. Free subscribers get articles like this one in their inbox. Paid subscribers get the full Clownfish TV podcast feed, livestreams, and members-only episodes that never hit YouTube.
D/REZZED is part of Clownfish TV. For more news, views, and rants on gaming, tech, and pop culture, watch @ClownfishTV on YouTube and find the podcast on Apple Podcasts, Spotify, and iHeart.
Article compiled and edited by Derek Gibbs (entertainment editor) and the Clownfish TV newsroom.
Hat Tips:
Variety (July 16, 2026), the primary reporting, verified Netflix’s second-quarter earnings disclosure that roughly 300 programs used generative AI across their production process so far this year, the shareholder letter’s statement that usage “expands across every level of a program’s production process, from its concept and pre-visualization to post-production and release,” the named titles Glory, Brasil 70: A Saga do Tri and The American Experiment, the use of the technology for “highly complex sequences” including enhanced crowd sizes and battle sequences, the quote that “we are increasingly leveraging these tools to deliver higher-quality output more quickly and at a lower cost than traditional methods,” Netflix’s March acquisition of Ben Affleck’s InterPositive, its AI animation studio, and Ted Sarandos’ March comments to Politico that AI “should be a creator tool”
ComingSoon, World of Reel, MobileSyrup, and Kotaku (July 16, 2026), verified the additional details — Ted Sarandos’ earnings-call statement that one documentary sequence was completed twice as fast at half the cost, Netflix’s note that some large-scale scenes may never have been finished without generative AI, the $12.56 billion quarterly revenue up 13.4% year over year, 97 billion hours watched in the first half of 2026, the reported 10x faster AI-assisted VFX on 2025’s The Eternaut, the InterPositive deal valued at up to $600 million for a 16-person team with Affleck as senior advisor, and MobileSyrup’s report that games disclosing generative AI use on Steam, where disclosure is mandatory, have lost up to 53% of revenue
Automaton West, VGC, and Mobilegamer.biz (April-July 2026), verified the wider pattern reported previously — the JOGA Online Game Market Research Report 2026 finding 100% generative-AI adoption among surveyed Japanese online game companies, a Harris Poll of more than 600 studios across the US, South Korea and Europe finding 87% adoption, a survey of Japanese creative professionals finding 59% of companies used AI with 71.4% not actively disclosing it, and Google Cloud games director Jack Buser’s statements that “their favourite games right now were already built with AI” and that the gap between survey figures “is basically the developers’ willingness to tell you”