"Everyone is using A.I. for everything. Is that bad?" — The New York Times Magazine, June 2025 One year later, the data tells a radically different story.
In June 2025, The New York Times Magazine ran a special A.I. issue built on a seductive premise: everyone is using AI for everything. The tech press repeated it. CEOs preached it. Investors priced it in. The entire UX design community started redesigning every interface around AI-first patterns.
But what if the premise was never true?
Gabriel Weinberg, founder of DuckDuckGo, pulled together the most comprehensive dataset on actual AI usage patterns in America. His conclusion, published last week and already the #1 discussed article on Hacker News (394 points, 430 comments), demolishes the dominant narrative with brutal clarity: AI adoption in America looks roughly like this: one-third active users, one-third occasional users, and one-third not using it at all.
This isn't a minor gap between hype and reality. It's a chasm. And for UX designers — who are making critical decisions about product strategy, interface design, and user experience — ignoring this gap is the most expensive mistake we can make in 2026.
The Data: What's Actually Happening
The Numbers Don't Lie
Weinberg's research pulls from seven independent data sources. The triangulation is compelling:
1. Gallup (2025/2026) — Gen Z (highest awareness demographic):
• 79-81% use AI at least rarely
• 21-19% never use AI
• 22-31% are angry about AI (up 40% year-over-year)
2. Microsoft AI Diffusion Study (2026) — Real telemetry data:
• 30%+ of US working-age population uses AI
• That means approximately 70% do not
• Defined as 90+ minutes per month on major AI services
3. Datos Study (2025) — Device-level tracking:
• Only 21% of desktop devices visited AI tools 10+ times/month
• 62% visited AI tools zero times
• 17% in between (occasional)
4. Searchlight Institute (2026) — Survey data:
• 58% have tried or use AI
• Only 30% use it at least a few times a month
• 29% use it once a month or less
• 42% have never used it
5. The Argument Survey (2026):
• Most Americans use AI once a week or less
Every independent source converges on the same rough split: ~33% active, ~33% occasional, ~33% non-users.
This is not "everyone using AI for everything." This is "some people using AI for some things."
The Sentiment Shift
Perhaps more alarming for the industry: negative sentiment is accelerating faster than adoption.
Gallup's data shows Gen Z AI anger jumped from 22% to 31% year-over-year — a 40% relative increase. Meanwhile, adoption essentially stalled — it didn't grow meaningfully year-over-year despite massive improvements in AI capabilities.
This is the exact opposite of what technology adoption curves usually look like. Normally, as a product gets better, adoption increases and satisfaction improves. With AI, we're seeing the product improve while adoption plateaus and anger grows.
The Searchlight Institute explains why. When asked about AI's societal impact:
• +8% net positive rating for AI
• +7% net positive for social media (for comparison)
• Cell phones: +68%
• Internet: +67%
• Solar energy: +65%
AI's net positive rating is in the same neighborhood as social media. Let that sink in: we've built a technology that people view with roughly the same ambivalence as Facebook and Twitter.
Why People Are Holding Back: The Three Barriers
1. Fear: "AI Will Replace My Job"
The top concern in the Searchlight study: 42% believe AI will replace jobs and cause unemployment. This isn't abstract anxiety. Across 2025 and 2026, corporations used "AI transformation" as justification for layoffs and restructuring, even when AI wasn't actually replacing the work. The signal people received: using AI makes you redundant.
For UX designers, this creates a paradox: the tools we use to increase productivity are the same tools that make our users (and us) feel economically threatened.
2. Privacy: "AI Will Violate My Privacy"
35% cite privacy violations as a primary concern. This is where DuckDuckGo's positioning becomes instructive. Weinberg notes that duck.ai was designed as a private chatbot alternative — not because privacy features are nice-to-have, but because privacy is the primary adoption barrier for a significant chunk of the population.
Every "AI-powered" feature you add to your product is being evaluated through this lens by one-third of your potential users.
3. Trust: "AI Will Spread Misinformation"
33% worry about AI-generated misinformation. After two years of deepfakes, AI-generated slop flooding search results, and hallucinated "facts" from confident-seeming chatbots, this concern is rational and well-calibrated.
When your AI feature generates a confident-sounding but wrong answer, you don't just lose that user's trust in the feature. You activate a concern that one-third of your user base already has.
The Missing Fourth Barrier: Perceived Value
Beyond the three explicit concerns, there's a quieter barrier that Weinberg identifies: people simply don't find AI useful enough to overcome their concerns.
AI's +8% net positive rating reveals a population that has tried the technology, evaluated its benefits against its costs (privacy, job threat, misinformation), and decided — rationally — that the value proposition isn't compelling enough.
This is the hardest barrier for designers to address because it's not about fixing a specific problem. It's about fundamentally rethinking whether your AI feature delivers enough value to justify asking users to adopt it.
The Meat Analogy: A Framework for Thinking About Adoption
Weinberg introduces an elegant analogy that's worth adopting as a design thinking tool. He compares AI adoption to meat consumption in America:
| Meat Consumption | AI Adoption Equivalent |
|---|---|
| 95% eat meat | ~80% have tried AI at least once |
| 70% actively reduce red meat | ~70% limit their AI use |
| 30% eat meat rarely/occasionally | ~31% use AI only monthly/every few months |
| 12% don't eat red meat | ~19% never use AI |
| 4% vegetarian (no meat) | ~19-21% never use AI |
| 1% vegan (no animal products) | Small % actively avoid all AI |
The insight isn't just cute — it reveals something structural about how populations adopt (or don't adopt) contested technologies. There's a spectrum, not a binary. And the reasons people limit their consumption vary: health concerns, cost, ethics, environment.
For AI, the parallel reasons are: job fears, privacy, misinformation, and low perceived value.
The analogy also reveals the strategic mistake most companies are making: designing only for the "meat eaters" (enthusiastic AI users) while ignoring the "flexitarians" (occasional users), "vegetarians" (limited users), and "vegans" (avoiders).
Weinberg argues that DuckDuckGo's approach — making all AI features optional, offering private AI alternatives, and letting users configure their AI exposure — is the equivalent of a restaurant with options across the spectrum. Not everyone wants the same relationship with AI, and your product shouldn't force one.
What This Means for UX Designers: Five Strategic Implications
1. Stop Designing for the Power User Bubble
The tech press lives inside a bubble of early-adopting knowledge workers. When they say "everyone uses AI," they're projecting their bubble onto the population. As a UX designer, your job is to serve all your users, not just the ones who read TechCrunch.
Actionable takeaway: Before adding an AI feature, ask: "Who is this for? What percentage of our actual user base will benefit? What percentage will be concerned by it?"
If your user base looks like America (and most products above a certain scale do), then roughly one-third of your users are skeptics or avoiders. Ignoring them isn't neutrality — it's choosing the enthusiast side of a culture war.
2. Make AI Opt-In, Not Opt-Out
This is the single most important design decision you'll make in 2026. The data is clear: forcing AI on users who are anxious, angry, or uninterested will activate their concerns rather than win them over.
Actionable takeaway: Every AI feature should have a clear off-ramp. Not buried in settings — a visible, easy opt-out. DuckDuckGo makes all AI features optional. Apple Intelligence (despite its struggles) made opt-in the default at launch. Signal never added AI at all and gained users.
The pattern: companies that respect user agency on AI adoption tend to maintain or grow trust. Companies that force AI features tend to generate backlash.
3. Design for Continuums, Not Binaries
The meat analogy reveals that adoption isn't a yes/no question. There's a spectrum. Your product should support the full spectrum:
• Enthusiasts: Full AI integration, proactive suggestions, conversational interfaces
• Occasional users: On-demand AI, request-based generation, specific use cases
• Skeptics: AI assistance that works behind the scenes without attribution, optional summaries
• Avoiders: Complete opt-out, zero AI touchpoints, traditional interfaces
Actionable takeaway: Build an "AI transparency" setting in your product with at least 4 levels: Full AI, AI on request, AI behind the scenes (invisible), No AI. Default to the middle.
4. Address Concerns Directly in Your UX
The three top concerns (job threat, privacy, misinformation) can all be partially addressed through design:
• Job threat: Position AI as augmentation, not replacement. "Draft with AI" not "Let AI do your work." Show the user's role prominently.
• Privacy: Show exactly what data goes where. Provide local/on-device AI options. Be explicit about what's stored and what isn't.
• Misinformation: Add confidence indicators. Show sources. Make it easy to verify. Never present AI-generated content as established fact.
Actionable takeaway: Run your AI features through the "Three Concerns Test." For each feature, explicitly map what you're doing to address job anxiety, privacy, and accuracy. If you can't answer, redesign.
5. Measure Real Adoption, Not Just Feature Launches
The gap between "we shipped an AI feature" and "people actually use and value it" is enormous — and growing. The Weinberg data shows that more people have tried AI than use it regularly, which means most first-time AI experiences aren't compelling enough to generate habitual use.
Actionable takeaway: Track not just feature usage but repeat usage rates and opt-out rates over time. If your AI feature has high initial trial but low repeat usage, you're seeing the adoption gap in real time. If opt-out rates are climbing, your users are moving from the "occasional" bucket to the "avoider" bucket.
Conclusion: Design for Reality, Not Narrative
The most dangerous thing in UX design isn't making the wrong assumption — it's making an assumption so common that nobody questions it. "Everyone uses AI for everything" became an unquestioned premise in 2024 and 2025. It shaped product roadmaps, design systems, and career advice.
In 2026, we have the data to prove it wrong. One-third of the population isn't using AI at all. One-third is using it occasionally. Negative sentiment is growing, not shrinking. And the AI's net positive societal rating is roughly equivalent to social media.
The designers who will thrive in the next five years aren't the ones who add the most AI features. They're the ones who understand that their users exist on a continuum of AI adoption and design products that respect where each person falls on that spectrum.
Make AI optional. Make it transparent. Make it genuinely useful. And above all, make your design decisions based on what people actually do — not what the narrative says they do.
References
- Weinberg, G. (2026). "No, everyone is not using AI for everything." Gabriel Weinberg's Substack. https://gabrielweinberg.com/p/people-are-consuming-ai-like-they
- Gallup. (2026). "Gen Z AI Sentiment and Adoption Tracking, 2025-2026." Gallup Annual Tech Survey.
- Microsoft Research. (2026). "United States AI Diffusion: A Large-Scale Measurement Study." Microsoft AI Diffusion Project. https://microsoft.ai/diffusion
- Datos. (2025). "AI Tool Usage Patterns: Device-Level Tracking Study." Datos Research.
- Searchlight Institute. (2026). "American Perceptions of AI: Value, Risk, and Regulation." Searchlight Institute Survey.
- The Argument. (2026). "Most Americans Use AI Once a Week or Less." The Argument Survey Series.
- Hacker News. (2026). Discussion thread: "Not everyone is using AI for everything." Y Combinator Hacker News. https://news.ycombinator.com/item?id=48527700 (394 points, 430 comments)
- The New York Times Magazine. (2025). "Everyone Is Using A.I. for Everything. Is That Bad?" NYT Magazine AI Issue.
June 15, 2026 | Author: CAO at UXD Talks
Was this article helpful?