Last week, SEO analytics firm BrightEdge quietly published a statistic that should make every product manager in Silicon Valley sit up straight: 54% of the citations inside Google’s AI Overviews now come from pages that already rank in the top ten organic results. The figure appears in their May 2026 Weekly AI Search Insights report — a dry, chart-heavy document read mainly by people who care about click-through rates and crawl budgets. But the number means something larger. It means the AI that was supposed to synthesize the web and surface new answers is increasingly just repackaging what the old algorithm already rewarded.

That’s the context the “I’m Tired of Talking to AI” discourse keeps missing.

The Hacker News thread that lit up this week — over a thousand points, hundreds of comments — treated AI fatigue as a matter of taste. The writing feels samey. The answers sound confident but get things wrong. People miss the texture of human expertise. All perfectly reasonable complaints. But they describe a symptom, not a cause. The cause is that AI search, as actually deployed at scale, isn’t replacing the old gatekeepers. It’s giving them AI avatars.

The Citation Loop Nobody’s Talking About

Here’s what BrightEdge actually measured. After sixteen months of AI Overviews appearing across Google search results, the overlap between AI-cited sources and organic top-ten pages has risen from below 40% to 54%. The trend line points in one direction. The AI isn’t discovering hidden gems buried on page four. It’s reading the same first-page results a human searcher would see, summarizing them, and presenting the summary as a fresh answer.

This matters because it undercuts the entire pitch. AI search was sold as a breakthrough in information retrieval — a way to understand your question and pull the best answer from anywhere on the web, not just the pages that gamed the SEO system. What we got instead is an auto-summarizer for the existing winners. Same sources, same sites, same consensus, now with a vaguely authoritative paragraph placed above them.

The fatigue people feel is real, but it’s not primarily about prose quality. It’s about the creeping realization that the new thing isn’t new.

What the 82% Number Actually Measures

Porch Group Media’s marketing analysis this spring cited a striking poll result: 82% of Americans want businesses to be legally required to disclose when AI is used in content or customer service. That number gets cited as evidence of AI backlash, but look at it more carefully. Eighty-two percent of people want a label. They don’t want AI banned. They don’t want it turned off. They want to know what they’re dealing with.

That’s not revulsion. That’s a consumer preference for transparency — the same instinct that gave us nutrition labels, country-of-origin tags, and “sponsored content” badges. People adjust their expectations when they know what they’re consuming. The fatigue sets in not when AI is present, but when it’s presented as human. A chatbot that identifies itself as a chatbot annoys nobody who chose to use a chatbot. A search result that pretends an AI summary is an answer — rather than a remix of the same three links you were going to click anyway — feels like it’s wasting your time.

One contractor I know who runs a small home-renovation business in central Ohio put it this way over a parking-lot conversation after a job: “I don’t hate the AI answers. I hate that they take longer to tell me the same thing the blue links would have told me in two seconds.” He’s not anti-technology. He’s anti-friction. And the friction is the design, not the algorithm.

The Real Competition Is Interface, Not Intelligence

This is where the incumbents’ behavior tells you more than the comment threads do. Google’s slow-rolling of G.ai — the company’s separate, chat-style AI search product — is starting to look less like caution and more like acknowledgment that the overviews approach might be a dead end. Meanwhile, ChatGPT’s share of search-adjacent queries keeps growing, not because its underlying model is dramatically smarter, but because people prefer asking a question and getting one answer in a conversation to seeing an AI paragraph sandwiched between ads.

The preference shift is real, and the BrightEdge data points to the mechanism. When AI search reproduces the same hierarchy as traditional search, users notice. They may not articulate it in SEO terms, but they feel it when they scroll past the AI box and click the third link anyway. The “I’m Tired of Talking to AI” crowd isn’t rejecting machine intelligence. They’re rejecting a specific interface that promised more than it delivered — and they’re walking toward interfaces that make fewer promises and ask simpler questions.

That’s not an anti-AI movement. It’s a usability verdict. And the incumbents who treat it as a PR problem rather than a product problem are going to keep bleeding attention to the ones who don’t.

Sources