Netflix Used AI in 300 Titles. Here's What That

Before we do the discourse thing — the "AI is destroying creativity" vs. "AI is just a tool" argument that both camps have been running since 2022 — let's actually look at what Netflix said, because the specifics matter more than the framing.

In its Q2 2026 shareholder letter, Netflix disclosed that generative AI workflows had been used in "roughly 300" of its titles this year, with the largest concentration in post-production. That's the quote, per Engadget and The Verge. Variety notes the AI use spans "across the production process" more broadly, and Crypto Briefing flagged that the number was "bigger than most people expected."

Okay. So: 300 titles. Post-production heavy. Disclosed in an earnings letter. What do we actually know?

What 300 titles looks like in practice

Post-production is a big tent. It covers everything from color grading and visual effects to sound mixing, subtitle generation, localization dubbing, and the increasingly AI-assisted work of cleaning up footage — removing background objects, de-aging faces, filling in incomplete shots. When Netflix says "the largest concentration of work" happened in post-production, that description is compatible with something as unobtrusive as AI-assisted subtitle translation on a Korean drama and something as structurally significant as AI-generated background environments in a prestige sci-fi series.

We don't know which 300 titles. We don't know the depth of AI involvement in any of them. Netflix didn't release that breakdown, and nothing in the sourcing I've seen suggests they plan to.

That ambiguity is worth sitting with, because 300 sounds like a lot — and is a lot, by any absolute measure — but its meaning scales differently depending on what "used generative AI" actually covered in each case. A film where AI cleaned up a few frames of a practical effect is a different thing from a film where AI generated half the visual backgrounds. Both count. Dataconomy and IBTimes UK both note the range of stages involved, but neither has the internal granularity either.

The historical comparison that actually holds up

Every new production technology gets the same treatment: first fear, then absorption, then the fear moves somewhere else. The talkies were going to destroy the art of visual performance. CGI was going to replace real stunt performers and make movies feel fake. (Both concerns had merit! And movies kept getting made.)

But here's where the AI moment diverges from the CGI moment in a way worth naming: CGI gave directors new tools to realize visions they already had. The decisions — what story to tell, what a scene needs emotionally, what a character's face should do — those stayed with humans. What GenAI does, increasingly, is reach into those decisions. It's not just executing a vision; it's capable of generating options, suggesting directions, filling in creative gaps. That's not inherently sinister, but it does mean the pressure falls somewhere new — on whoever decides which AI output is "good enough" and which needs a human to go further. When CGI showed up, it changed what was possible. AI is starting to change who decides.

That's a distinction that tends to get lost in the "AI is just a tool" defense, which is true but incomplete. A hammer is also just a tool, and yet the invention of pneumatic nail guns changed who gets hired at construction sites.

What Netflix actually got out of saying this

Here's the thing about the Q2 earnings letter: Netflix chose to include this disclosure. That choice is doing work.

They're being transparent — but they're being transparent to Wall Street, which is a different thing than being transparent to audiences or to the workers whose crafts are being supplemented or replaced. Investors want to see AI ROI. They want to hear that a company with Netflix's content budget is finding efficiencies. The 300-title figure, dropped into a shareholder letter, is a signal to that audience: we are modernizing, we are not being left behind, this is a growth story.

IBTimes UK headlined the story with a "slash production costs in half" framing — I haven't seen the methodology behind that specific figure, so I'd hold it loosely — but it illustrates what investors are being invited to imagine when they read "roughly 300 titles." Cost reduction. Scale. Margin expansion.

What the earnings letter doesn't tell audiences: which shows they watched were AI-assisted, what parts, and what — if anything — would have looked or sounded different without it. There's no industry standard requiring that information. Nobody has to say. Right now, no streaming platform, studio, or network is required to tell audiences anything about AI use in production. The information exists, clearly — Netflix just disclosed it to shareholders — but it flows upward to capital, not outward to the people actually watching.

The things that are genuinely unsettled

A few tensions this story surfaces that I don't think have clean answers yet:

The quality question is still open. We have no data on whether AI-assisted titles performed better or worse with audiences. Netflix hasn't released that correlation, and it may not exist cleanly — production AI is too embedded in each title's specifics to isolate as a variable. We don't know if viewers noticed, cared, or preferred it.
The labor question is live. The 2023 SAG-AFTRA and WGA strikes were partly fought over exactly this territory. The contracts that came out of those negotiations included some AI guardrails, but "roughly 300 titles" suggests the pace of adoption has moved well past what those agreements anticipated. What happens to the VFX artists, the colorists, the localization specialists whose workflows AI is eating into? The earnings letter doesn't mention them.
The authorship question has no consensus. If an AI system generates a shot that ends up in a film, who created it? The director who approved it? The engineer who trained the model? The model? Courts are still working through copyright questions on AI-generated content, and the entertainment industry is not waiting for those answers before deploying the tools.
The IBTimes cost figure needs scrutiny. IBTimes UK framed this as AI slashing production costs in half. Maybe. I genuinely don't know what's behind that number — it could be accurate for specific workflows; it could be a best-case projection; it could be something a PR person said once. I'd want to see it sourced more granularly before treating it as a headline claim.

Netflix is genuinely ahead of most of its competitors on willingness to put a number on this. That's worth something. But "ahead on disclosure to shareholders" and "ahead on transparency to audiences or workers" are not the same finish line — and conflating them is probably what the earnings letter is banking on.

The real question isn't whether 300 is a big number. It's who gets to decide what that number means, and for whom. Right now, the answer is: the company, in a letter to investors, on a quarterly cadence. That might eventually feel insufficient. It might already.

Zara Chen is a tech and politics correspondent for Buzzrag.