AI Was Supposed to Win People Over by Now — It Hasn't

Fifty-two percent of Americans now say they are "more concerned than excited" about AI — up from 37 percent in 2021 — and last week Anthropic CEO Dario Amodei became the first prominent AI executive to say clearly what that polling has been signaling for years: the industry hasn't delivered on its promises. TechCrunch's Sarah Perez documented the moment on August 19, 2026, tracing how the trust curve bent downward even as investment curves bent sharply up. The numbers don't describe a technology on the verge of mainstream acceptance; they describe a credibility gap that has been widening since the commercial AI boom began.

The Claim

Amodei's admission is worth sitting with, because it's unusually specific for a sitting CEO. His framing was that the "by far most accurate criticism" of AI companies — including his own — is that they haven't yet delivered on their largest promises. He called the resulting situation a "crisis of trust." That's not spin or a tactical pivot; it places the cause of public skepticism squarely on product delivery rather than on an uninformed public that doesn't understand the technology yet.

Airbnb's Brian Chesky, speaking as an outside observer with no AI product to defend, supplied a companion argument about what delivery should actually look like. His framing centered on access democratization: give people the equivalent of a doctor on demand, the kind of expert guidance that is normally gatekept by cost, geography, and scheduling. That's a fundamentally different value proposition than what the AI industry has mostly shipped, where the pitch has centered on productivity features, enterprise integrations, and capability demonstrations rather than expanding access to things people couldn't otherwise get.

The broader polling landscape sharpens both points. An Economist/YouGov survey found more than 70 percent of Americans believe AI is advancing too fast, and a separate CNBC poll of adults aged 18 to 34 — the demographic historically most sympathetic to new technology — found a majority don't trust the industry's top leaders to act responsibly. Taken together, the data doesn't suggest confusion about AI's capabilities; it suggests a settled, negative judgment about whether the people building it can be trusted to do so wisely.

What We See

The 15-point shift in Pew's concern measure deserves emphasis on its own terms. The share of Americans who are "more concerned than excited" moved from 37 percent in 2021 — when most consumers hadn't directly interacted with an AI product — to 52 percent in 2026, after years of product launches, media coverage, and multiple rounds of claimed breakthroughs. The concern didn't precede the products; it followed them. That's the inverse of what a successful technology rollout looks like.

The Queen Zone's July 2026 reporting, drawing on a Quinnipiac University poll from March of that year, adds a layer that makes the situation stranger: adoption and trust are moving in opposite directions simultaneously. Quinnipiac found 76 percent of Americans trust AI-generated information "hardly ever or only some of the time." Yet Pew separately reports that 44 percent of U.S. adults have now used ChatGPT — up from 18 percent in 2023 — and 38 percent of employed adults use AI for work tasks. Brookings analysis of 2024 data put generative AI usage among working-age adults at 39.6 percent. People are using the tools; they are just not believing in them. Quinnipiac's Chetan Jaiswal, an associate professor of computer science, described the split plainly to reporters: Americans are adopting AI with deep hesitation rather than deep trust. Multiple sources corroborate that framing across different polling methodologies.

Our read is that this hesitation-without-rejection pattern is the most important signal in all of this data, and it explains why the industry's instinct to "educate" the public is the wrong prescription. Someone who uses AI daily to draft emails while refusing to trust it with a medical question isn't confused — they've made a calibrated judgment based on experience. Closing that gap requires the tools to be right in the high-stakes moments, not just functional in the low-stakes ones. That's a much harder problem than shipping more features.

The cultural signals running alongside the polling add texture that most reports haven't fully connected. The same period that produced these trust numbers also produced a measurable boom in retro technology — dumbphones, cassette players, algorithm-free music devices selling at premiums on resale platforms, in-person social clubs that explicitly exclude apps. Among younger adults especially, there appears to be deliberate construction of AI-free leisure space. When a demographic defines part of its social identity by the absence of a technology, that technology has a relationship problem that product updates alone are unlikely to solve.

Where It Falls Short

Both Amodei's and Chesky's framings, honest as they are by industry standards, leave something out. Amodei treats the problem as primarily a delivery gap — the promises were right, the products just haven't caught up yet. But 70 percent of the public believing the pace is already too fast isn't a delivery complaint; it's a legitimacy concern. A meaningful portion of people aren't saying "I'll trust this when it works better." They're saying they're uncertain whether they want it to keep accelerating at all, regardless of quality. That distinction matters enormously for how companies should respond, and neither CEO addressed it directly.

The infrastructure economics add a ground-level signal worth noting. Reports indicate AI companies are now offering community inducements — job guarantees and teacher bonuses reportedly reaching $50,000 — to secure local approval for new data center builds. If communities require financial incentives to accept the physical footprint that AI needs to scale, the trust deficit has moved out of the opinion-polling layer and into municipal negotiations. That's a different category of problem. Chesky's vision of democratized access is compelling as product-design theory, but the industry's track record of sustained commitment to unglamorous, high-value problems — rather than to novel capabilities that attract press attention — remains unproven. The polls suggest people are watching for exactly that, and so far they are not seeing enough of it.

Sources