The week after Google I/O, DuckDuckGo CEO Gabriel Weinberg issued a statement: “Google is force-feeding AI with no way to opt out.” His company then published data to back it. From May 20 to May 25, U.S. app installs climbed 18.1% week-over-week, peaking at 30.5% on May 25. On iOS the growth averaged 33% week-over-week and peaked at 69.9%. Apptopia confirmed a 29% increase in average daily U.S. downloads independently.
Equally telling was noai.duckduckgo.com — a stripped-down search endpoint with every AI feature explicitly disabled. Visits to that page grew 22.7% week-over-week, peaking at 27.7%.
Sundar Pichai announced at Google I/O that AI Mode had surpassed one billion monthly active users, and he’s probably right. But that framing misses a different signal: a billion users who opted in via Google’s defaults do not represent universal enthusiasm. Google is making conversational AI the primary search interface; DuckDuckGo’s numbers measure the subset who noticed that change, disliked it, and left.
This is different from the LLM adoption story we usually tell — about developers embracing coding agents, enterprises consuming tokens faster than they budgeted. Those are cases where someone chose the AI product. Search is different. Most of Google’s users didn’t go to google.com to interact with an AI; they went to find things. When the interface rearranges itself to center AI responses, a portion of users experiences it as a loss, not a gain. The distinction matters: opt-in adoption tells you about enthusiasm, opt-out behavior tells you about tolerance.
YouTube’s announcement on May 27 adds a second data point. Starting this month, YouTube will automatically detect photorealistic AI-generated video and apply disclosure labels even when creators don’t self-disclose. Labels now appear directly below the video player — not buried in the description — and as overlays on Shorts. Content created with YouTube’s own Veo or Dream Screen tools gets a permanent, non-disputable label.
YouTube isn’t banning AI-generated content; it’s building infrastructure for transparency. That’s a meaningful distinction, but it also reflects something real: AI-generated content is proliferating fast enough that manual self-disclosure alone can’t keep up, and audiences increasingly want to know what they’re looking at before deciding whether to trust it.
Together these two stories point at the same friction. AI in search is expanding fast enough that a measurable subset of users is switching search engines. AI-generated video is proliferating fast enough that YouTube needs automated detection to label it at scale. Neither is a crisis. Neither suggests AI adoption is reversing. But they do suggest that “make AI the default” is a strategy with costs that aren’t evenly distributed across the user base.
The opt-out market DuckDuckGo is capturing is small in absolute terms. But its rate of growth is unusually high, and it’s being driven by a deliberate design choice Google made — not by DuckDuckGo’s own improvements. That’s an unusual competitive position to be in. The question isn’t whether Google’s billion MAUs validate AI Mode; they do. The question is whether the users who leave because of it matter enough commercially to make the tradeoff worth revisiting.