The AI backlash is in full effect. And it is no longer confined to comment sections.
From college students booing the mere mention of AI at their own graduations to graffiti reading “we don’t have to accept this future” on subway ads for AI, it’s clear that sentiment has shifted.
Just three years ago, when the generative AI frenzy exploded, the public viewed these breakthroughs with wonder and excitement. Now, as Axios puts it, we’re in an “AI hate wave.” More than half of Americans have negative views of AI. Only 18% of young people ages 18 to 29 feel hopeful about the technology.
For years, people have warned that AI will extinguish millions of jobs, make it harder to know what is fact or fiction, replace human connection and seep into every part of our lives — whether we like it or not. In short, people feel like AI is being done to them, not being built for them.
Every day, people are encountering new, loud, flashy features that leave them wondering, “Who asked for this?” Think about the pop-up explaining a response was “generated by AI” for a task that didn’t need explaining. Or the sparkle icons stamped on every button, whether or not anything has changed.
To rebuild trust with the public, companies need to move away from the intrusive, in-your-face approach to AI. What’s needed instead is a commitment to AI that disappears — one that quietly makes people’s lives better without asking for attention or credit.
Grounding AI Strategy in What Users Want
Of course, this is easier said than done.
AI has attracted trillions of dollars in investment. Tech companies feel immense pressure to build full speed ahead and capitalize on this moment. They also feel pressure to be loud about their progress with social media blitzes, press releases and 50 mentions of AI on every earnings call.
But the danger is when that mentality begins to influence what is built — when features are shipped to justify valuations and secure front-page headlines instead of delivering what users need. That approach may impress investors. But users inevitably are turned off.
To avoid that trap, companies need to invest in robust research that understands and measures what users actually want. A lot of consumer-facing organizations make the mistake of relying on engagement metrics alone — how much people click on an AI feature and spend time with it. Real research answers a more fundamental question: Is the user better off now that the feature exists?
One example is Spotify’s Discover Weekly. Instead of forcing users into a loud chat sidebar to ask for recommendations, the app uses AI to examine your favorite songs, genres, artists — as well as listeners with similar tastes. And every Monday, Spotify drops 30 songs into a playlist that feels made just for you — to help you find your next favorites. Success is measured by a real outcome: Did people find new artists and songs they love?
Every AI feature has a line where it tips from helpful to intrusive. It’s found by watching people while they use your product, not tracking engagement — prioritizing real user reactions and outcomes over internal opinions.
Building AI That Disappears
In my experience over the years, people don’t want to know that features are AI. They don’t think visibility means value.
While Silicon Valley conferences and Slack channels can make it seem like AI is the center of the world, most of the world doesn’t think that way. They just want products to work well and stop noticing AI is there at all.
Think Google Maps quickly rerouting your trip after a wrong turn. Or live noise canceling on calls. Zoom, Teams, and AirPods strip out the dog barking or the café behind you. No banner, no sparkle — the call just sounds clean.
The best AI doesn’t announce itself. Rather, it solves a problem and then gets out of the way. So how should that core insight inform how a company should brand AI vs. hiding it — especially when transparency is a top priority that’s fundamental to earning user trust? Here are a couple of principles that guide me and my team:
- The default position is to hide AI — users want results, not to know how it works. It doesn’t need to be branded if AI is just making something they already do on our app work better.
- AI should be branded when users need to opt in, their consent is needed, or to make sure they understand the trade-off in a particular situation — think permission for access to their photos or anything that involves sensitive data.
- And whenever it’s a close call, it’s important to do a quick internal gut check and ask: Would a user feel betrayed finding out that AI did this? If the answer is yes, we show it.
Knowing When AI Doesn’t Belong
That gut check works for most decisions. But some questions go even deeper — like when to decide if AI features don’t belong at all.
AI promises quick fixes that take the friction out of life. But what if that means optimizing away what makes us human in the first place?
Stanford professors Bob Sutton and Huggy Rao have introduced this concept of “good friction” — how some processes that force people to slow down and reflect can lead to more enriching experiences. It may lead to harder moments, but the payoff down the road is better outcomes.
Good friction can be small features — like a long password requirement when signing up for an app to ensure a safer experience or a pop up note asking “are you sure?” before deleting thousands of emails.
Other times, good friction should shape the entire user journey. That’s something I think about a lot as the builder of a dating app. The ultimate goal is helping humans meet other humans. Sometimes those first interactions between two single people are awkward. Often, people aren’t sure what to say or how to say it. But that’s how real relationships are earned and grow over time. That’s good friction. AI should never automate the authenticity out of the experience.
How does that belief show up in everyday product decisions? One example is how people communicate on our platform. AI will not draft any of the conversations — that should be as human as possible. AI might offer some suggestions, but will never let technology substitute for human interaction.
Friction like this is worth protecting. It may make things harder, but it also makes the journey more real for people. And that’s what people want in this age of AI — experiences and technology they can trust.
So before every company answers the question “can AI do this?”, it should first answer “should AI do this?”
Final Thought
Trust in AI can’t be achieved with one product launch or one PR campaign. It’s earned with a series of small decisions that make people feel they are the priority — not the technology. That their well-being is front and center, and not some afterthought in the AI race.
The companies that earn this trust won’t be the ones with the loudest features. They’ll be the ones with AI that nobody talks about, but everybody loves — because it got out of the way and made lives easier, instead of disrupting them.
