Megan Skiendel: Jeremy. Good — you're here, I need someone to be annoyed at this with me.
Jeremy Clarkson: Oh, I am already annoyed and I don't even know what it is yet.
Megan Skiendel: About 70% of the U.S. working-age population has never used AI at all — yet last year the New York Times Magazine ran a cover story declaring 'Everyone Is Using AI for Everything.'
Jeremy Clarkson: Right. So the New York Times Magazine — the publication that exists inside a building in midtown Manhattan, staffed entirely by people who use AI to draft their grocery lists — looked around at their immediate social radius and concluded: everyone.
Megan Skiendel: Honestly, I've seen this exact move before — when I was at Google, circa 2012, there was this internal assumption that because everyone in the building had an Android, smartphone saturation was basically a solved problem, and I mean — it wasn't, the numbers were embarrassing outside the coasts.
Jeremy Clarkson: And what your anecdote actually predicts — which is the important bit — is that this isn't a media failure, it's a class failure: the people writing about AI are, to a man, the people using AI, and they have simply confused their dinner party for the country.
Jeremy Clarkson: The New York Times Magazine ran an introduction last year. Titled, quite literally, 'Everyone Is Using A.I. for Everything.' And Gabriel Weinberg — founder of DuckDuckGo, so not exactly a disinterested party, but still — looked at that and said: no. Actually, roughly one-third of Americans have never used it at all.
Megan Skiendel: The Searchlight Institute survey he cites. 58% have ever tried it — and that's split almost evenly between about 30% who use it regularly and 29% who've touched it once or less. That's his data.
Jeremy Clarkson: Which means the Times was describing, at best, the cultural zip code of the people who write for the Times.
Megan Skiendel: Here's the thing — I don't think that's cynical. I think that's sincere. I've sat in editorial meetings where a senior editor was genuinely baffled that half their staff wasn't using ChatGPT. Not performing bafflement. Actually surprised.
Jeremy Clarkson: Oh, I'm sure they were surprised. That's rather the point, isn't it? The Hard Fork podcast — which is what that NYT Magazine intro was actually drawn from — is a podcast for people who already think AI is the only story. You're selecting for the most enthusiastic possible audience and then declaring that everyone agrees with you.
Megan Skiendel: But then Epoch AI and Ipsos ran a poll in April 2026. 50% of Americans used an AI service in the past week. That's not a bubble number.
Jeremy Clarkson: Now that's where the definitions matter enormously. Used an AI service. Does that include Google's AI search results you didn't ask for? Does it include autocomplete? Because Weinberg's point — and it's a good one — is that the infrequent cohort gets collapsed into 'users' by surveys that then imply daily engagement.
Megan Skiendel: Survey definitional variance. Yes. The WIN World AI Index is drawing on 40,000 respondents across 44 countries — that's a very different question population than Searchlight's U.S. sample.
Jeremy Clarkson: And yet WIN is showing a 24-point jump in daily and frequent users in a single year. That is not noise. That's something structural changing.
Megan Skiendel: Which is why Weinberg's meat analogy starts to crack under pressure. The idea is: some people embrace it, some limit it, some avoid it — like meat consumption. Stable, principled categories. But a 24-point jump year-over-year is not a stable market.
Jeremy Clarkson: The meat analogy breaks down the moment AI stops being a consumer choice and becomes infrastructure. You can avoid meat because someone will cook you something else. You cannot avoid the AI your mortgage lender is using to assess your application.
Megan Skiendel: That's — yes. And I've watched this happen inside organisations. You roll out an AI tool, you've got a third of the workforce who are in, a third who are cautious, a third who want nothing to do with it. And within about six months, the avoider category doesn't disappear. They go underground. They improvise in the shadows — reverting to old workflows while appearing compliant.
Jeremy Clarkson: So they don't become users. They become—
Megan Skiendel: Hidden resisters. Which is a management problem, not an adoption statistic.
Jeremy Clarkson: And nobody's counting them in the WIN survey.
Megan Skiendel: Right. But here's what actually surprised me — Anthropic ran the largest qualitative AI study ever done. 81,000 Claude users across 159 countries. And they found that hope and fear coexist in the same individuals. Active daily users who are also genuinely frightened of what they're using.
Jeremy Clarkson: Good lord. That completely demolishes the NYT framing from a different direction. Not because adoption is low — but because adoption doesn't mean acceptance. You've got people who are technically in Weinberg's 'embracer' bucket who are actually anxious pragmatists.
Megan Skiendel: Exactly. So the tripartite model — Weinberg's one-third, one-third, one-third — it's too clean. Even inside the active-user third, you've got people who are conflicted.
Jeremy Clarkson: But that doesn't rescue the Times. Look, the Weinberg motivations for avoidance are real — cost, ethics, environmental concerns, genuine anxiety about cognitive dependence. Those aren't transitional-phase reluctances. Those are principled positions. The problem is that principled positions don't survive mandatory rollouts.
Megan Skiendel: So you're saying Weinberg is right about the data and wrong about the forecast.
Jeremy Clarkson: I'm saying the Searchlight Institute surveys and The Argument data describe real, genuine stratification. But the moment AI becomes workplace infrastructure rather than a consumer app, the choice architecture collapses. We're already moving from a market of preference to something closer to a regime of compelled use. And the interesting question isn't how many people are using it. It's — when does the hidden resistance stop being hidden?
Megan Skiendel: And whether the people doing the compelled using are being counted as adopters by the surveys that are making everyone feel better about the narrative.
Jeremy Clarkson: Which brings us neatly back to the Times. Because that framing — 'everyone, for everything' — it doesn't just overstate adoption. It preemptively closes the conversation about whether any of this was chosen.
Megan Skiendel: And I'm still not sure that's malicious rather than myopic. The people writing that cover genuinely cannot imagine the one-third who never tried it. That's not a conspiracy. That's just a very specific kind of blindness.
Jeremy Clarkson: Well. Whether it's dishonest or merely oblivious, the effect is identical. Someone who opted out — for cost, for ethics, for whatever reason — picks up the Times and is told they don't exist. That seems, frankly, like a problem worth naming.
Megan Skiendel: And by 2027, when the workplace rollouts finish, will there even be enough of them left to complain?
Jeremy Clarkson: Well, that's the question, isn't it. Not whether people are using it. Whether the people who aren't get to keep saying no.
Megan Skiendel: Yeah. And honestly — I've watched that window close before, on other things, and the moment you realize it's shut is never when it's shutting. It's always later. You look back and think, oh, that was the last time anyone asked.
Jeremy Clarkson: Right. So the sharper question isn't how many people are using AI. It's: at what point does opting out stop being a preference and start being an act of defiance. I don't know the answer. I'm genuinely not sure anyone does yet.
