DuckDuckGo’s no-AI search page saw nearly 28% more visits in the week after Google CEO Sundar Pichai sat down with the New York Times’ “Hard Fork” podcast on May 22 and insisted, with the placid confidence of a man whose compensation package just hit $692 million, that people genuinely love AI-generated search results.

The timing was not subtle. Pichai’s interview dropped. Within days, a measurable slice of search traffic decamped to noai.duckduckgo.com—a URL whose very existence is a shiv. You don’t accidentally type “noai” into a browser. You go there because you want the machine to know it.

The easy read on this is that Google has a product revolt on its hands—that ordinary users are rejecting AI slop and voting with their clicks. This is comforting if you already dislike AI search, but it’s almost certainly wrong. The 28% bump represents a rounding error against Google’s daily query volume. If you squint at Similarweb charts, you can make the spike disappear entirely.

So why does it matter? Because the people who typed noai.duckduckgo.com into their address bars are not a random sample of the search population. They are the multiplier users—the cohort whose tool choices don’t stay confined to their own machines.

The Multiplier Problem

Every product has multiplier users. They’re the ones who configure the family laptop. They set the default search engine in the IT procurement spec. They write the documentation, teach the workshop, answer the question in the group chat. When they switch, they don’t switch alone.

Google spent two decades capturing these users for a reason. Get the developer, and you get the default browser on every machine they touch. Get the tech-adjacent relative, and you get the Thanksgiving tech-support session that migrates three more households. The multiplier effect is why Chrome became the browser and Google became the verb—not because anyone ran a Super Bowl ad, but because the people who install browsers for other people chose Chrome.

What DuckDuckGo’s traffic bump suggests is that a nontrivial number of those multiplier users are now actively seeking an exit. Not trying DuckDuckGo out of curiosity. Specifically seeking the version without AI.

“I switched my parents over last weekend,” one software engineer told me in a Slack DM, referencing the no-AI page. “My mom just wants to know what time the pharmacy closes. She doesn’t need Gemini to write her a paragraph first.” This is not a mass movement. It’s a thousand small decisions made by the people whose decisions propagate.

The Metric That Won’t Alarm Anyone

Here is what will not happen: Google’s quarterly earnings will not show a search revenue decline attributable to AI-mode defectors. The traffic numbers are too large, the defectors too few. Pichai can point to engagement metrics—time on page, query refinement rates, whatever internal dashboard he prefers—and argue convincingly that AI mode is sticky for most users.

He will probably be right. Most users will accept whatever search experience sits in front of them, the same way most users accepted whatever homepage their ISP set in 1999. The median search user is not optimizing their toolchain. They’re typing “weather” into whatever box appears.

The problem for Google is that multiplier users don’t need to be numerous to be damaging. Apple’s market share in personal computing never needed to exceed single digits for the Mac to shape the entire software industry—because the people who built the software used Macs. Google’s multiplier users are the equivalent: small in headcount, enormous in downstream influence.

If they start setting DuckDuckGo as the default on devices they configure for others, Google loses distribution it cannot buy back with ad spend. No campaign budget replaces the nephew who sets up the new laptop.

What Google Actually Misunderstood

The conventional critique of Google’s AI mode is that it’s a worse product—hallucinations, bland prose, an answer engine that removes the need to visit websites, killing the ecosystem that made search valuable in the first place. All of that may be true. But it’s not why the multiplier users are leaving.

Multiplier users aren’t leaving because AI search is bad. They’re leaving because AI search is presumptuous. It assumes you want synthesis when you wanted a list of links. It assumes the question behind your question is more interesting than the question you actually typed. For a certain kind of user—the kind who knows exactly which documentation page they need and just wants the fastest route to it—this is maddening in a way that product managers who measure “engagement” cannot detect.

Google built AI mode for the median user and measured its success with median metrics. What it missed is that the users who matter most to long-term distribution are almost never median. They’re edge cases. And edge cases, when ignored long enough, find the exit.

Twenty-eight percent is not a crisis. It’s a warning. Google just isn’t organized to read it.

Sources