What if Consumers Hate AI Content?

Consumers love using AI, but they hate consuming it. What does that mean for our AI economy?

Last week, OpenAI announced they’d be discontinuing Sora, their video generation app. If you believe the press release, this is a strategic move to divert compute to more valuable revenue streams, like agentic AI.

That’s a cleaner public narrative than the truth: consumers hated Sora.

Sora peaked at 3.3 million downloads in November 2025. Three months later, that number had fallen 67%. OpenAI was burning roughly $1M per day running an app with 2.8 stars in the App Store. The Disney partnership collapsed and the term “AI slop” hit the zeitgeist.

Sora is the most expensive consumer product flop in AI so far (one that I predicted – sorry I can’t help myself). But it doesn’t look like it’ll be the last.

Over the past 3 years, AI has gotten exponentially better at replicating human content: from videos of Brad Pitt and Tom Cruise fighting to many of the articles on this app. It wasn’t hard to predict this trend. You could easily extrapolate from the 2023 and 2024 AI-generated videos of Will Smith eating pasta to a place in the not-too-distant future where AI and real life are almost impossible to distinguish.

This chart from ARK Invest made the rounds on Twitter recently. It’s my least favorite kind of chart – there’s no explanation of methodology (including on ARK’s website) - so it’s safe to assume it’s all garbage.

But the reaction and virality of the chart (Elon Musk retweeted it) hints at something we’re all feeling: AI content is everywhere and inescapable, and, even if it hasn’t outpaced human output yet, there’s a good chance it will soon.

The question isn’t whether AI can produce all this content (with enough compute, it can) or whether the quality is poised to improve (it clearly is). The question is whether anyone wants to consume it.

AI content still suffers from the uncanny valley effect, meaning consumers can kind of tell something’s off. But the typical argument is that there will be a point where the technology improves enough that the discomfort goes away.

I’m not sure I believe that. Yes, every new creative technology triggers an authenticity backlash. People hated auto-tune. They mocked early CGI. Synthesizers were going to kill real music. But those tools enhanced human creators, they didn’t replace them. Consumers can feel the difference between a technology that exists to serve them and one that exists to cut costs on them.

The uncanny valley stems from the fact that humans are uniquely good at pattern recognition and spotting when something isn’t quite right (there’s a well-researched neural mechanism for this). We’re also extremely sensitive to being duped. It’s a phenomenon known as the betrayal aversion.

In the early days, you might not have been so alert to AI content. But now, knowing it’s all around you – in the content you’re reading, in the ads you’re seeing, in the photos and videos on social media – your senses are probably a little heightened. Certain terminology and image effects trigger something instinctual, similar to the feeling when you spot someone you know from far away with their face mostly covered, but can still recognize them.

The backlash isn’t just aesthetic or that people think AI content looks weird. It’s that they feel deceived by it (back to that betrayal aversion). 52% of consumers say they’re concerned about brands posting AI-generated content without disclosure and 31% say AI in ads makes them less likely to choose that brand.

In other words, even if AI eventually can deceive us, it might not work. We have a natural allergic reaction to it, and the better AI gets at mimicking human work, the more betrayed people feel when they discover the deception.

The evidence isn’t anecdotal anymore – we see this response in everything from advertising to social media platforms.

There’s a staggering perception gap in advertising: 82% of ad execs believe young consumers feel positively about AI-generated ads but the real number is below 50%. And that gap is widening, not shrinking.

Coca-Cola has become the classic example. Their AI-generated Christmas ads in 2024 and 2025 were hammered as “soulless” “lifeless” “digital slop.”

In contrast, Le Creuset got ahead of AI allegations by tagging a creator and their process in a recent Instagram photo. Nearly every comment on the post is about the choice to use a real artist and not AI.

We’re seeing this in the creator economy as well. Consumer preference for AI-generated creator content has fallen off a cliff: from 60% in 2023 to 26% in 2026. Some of that initial 60% was probably novelty (“wow a fun new toy”). But the drop hasn’t stabilized; it’s still falling. And it’s showing up in behavior, not just surveys: brand partnerships with virtual influencers fell nearly 30% during 2025. Instagram’s Adam Mosseri faced direct backlash from creators claiming the platform “failed to protect them” against AI slop. In response, Instagram updated its algorithm to actively penalize highly polished synthetic content.

Similarly, Pinterest now lets users filter AI-generated content out of their feeds through a “GenAI interests” feature and added a per-pin “Show fewer AI Pins” option.

Two platforms built on visual content (the ones that arguably invented the influencer economy) decided that AI-generated content is bad for its ecosystem.

Meanwhile, creators are leaning hard into proving their work is handmade (the newly dubbed authenticity premium). “Messiness” is becoming essential – you need evidence of imperfection and some human touch. When the tools that democratize creation also destroy the signal that makes content valuable, craft becomes the last remaining differentiator. “Made by Humans“ is starting to behave like a quality label similar to “organic” in food.

You might wonder why this is important (we’ve been fine without AI content for many years), but there are important implications for AI economics, and by extension, our broader economy.

There are only three ways for large AI companies like OpenAI to generate the kinds of revenues needed to justify their staggering capital expenditures and valuations: (a) replacing human labor at enterprise scale, (b) consumer tools, and (c) winning consumer attention with AI-generated content.

The signals on the latter are increasingly dire.

OpenAI is currently running at about $25 billion in annualized revenue, with enterprise revenue growing fast enough to approach parity with consumer.

And that consumer revenue is almost entirely people paying $20/month to use ChatGPT as a tool – to draft emails, brainstorm, ask embarrassing questions, plan trips. It is not coming from people consuming AI-generated content. You could argue that OpenAI hasn’t seriously tried the content play beyond Sora because they never built an AI entertainment feed or an AI social platform. But Sora was the test. It was the most well-resourced attempt any AI company has made at a consumer content product, and it failed on retention, ratings, cultural reception. If OpenAI couldn’t make it work with that much capital and attention, it’s fair to ask who can.

There’s an asymmetry here that isn’t fully priced into projections about ChatGPT’s growth: consumers love using AI, but they hate consuming AI.

In other words, you are probably very happy to have ChatGPT write an email for you. But you don’t want to read an email that ChatGPT wrote for someone else.

I want to use AI to edit this post; but you’ll probably be less interested in reading it.

If that asymmetry holds, then at least one revenue path for AI companies narrows significantly. It funnels toward enterprise and labor replacement. The consumer content play – AI-generated ads, AI influencers, AI video platforms, AI art – starts to look less like a massive market and more like… a dead end.

And I haven’t heard a great thesis about what bridges the gap between the current ick-factor and mass consumer acceptance of AI content. The “it’ll get better” argument assumes the problem is quality. But the data increasingly says the problem is trust, authenticity, and a deep human preference for knowing that another human made the thing you’re experiencing. Those preferences don’t dissolve with better models.

This isn’t an anti-AI essay. In fact, I’m extremely long AI – as a platform for discovering and building things we haven’t imagined yet. There are plenty of real, massive revenue streams for OpenAI and Anthropic and Google that have nothing to do with generating content for consumers. But the assumption that AI will eventually win over consumers as a content medium — not just a productivity tool — is baked into a lot of growth projections and valuations right now. And the data is increasingly saying that we shouldn’t take that for granted.