Half of US adults now use AI chatbots, more than double the share in 2023. One in four use them every day. And more of those same adults expect AI to hurt them than help them.
Both findings come from Pew Research Center's 2026 data. Most commentary picks one of them and ignores the other. I think the contradiction is the most useful thing a marketer can study right now.
The numbers do not agree with each other
Look at usage first. Sensor Tower data shows ChatGPT reached one billion monthly users by May 2026, and Gemini logged 950 million users in July. OECD figures say more than a third of adults across 38 wealthy democracies used a generative AI tool within a three-month window.
Now look at sentiment. A Stanford report found that more than half of respondents worldwide say AI products make them nervous. A Gallup poll in May found 71 percent of US adults would oppose a new AI data centre in their area, compared with 53 percent who would oppose a nuclear plant. An NBC poll in March ranked AI as less popular than the internal combustion engine.
Read cold, that looks like hypocrisy. It is not. It is the same pattern we see with social media, cheap flights and fast food. People use what is useful and still dislike the system that delivers it.
People dislike the push, not the tool
Here is my reading, from watching how clients and their customers actually behave. People like the moment AI helps them. They dislike almost everything around that moment.
They dislike the AI button added to software that worked fine before. They dislike the chatbot that blocks the route to a human. They dislike the announcement that a company is "AI-first" in the same press release that cuts jobs. They dislike being told the future is inevitable by people who will profit from it.
The tool solves a problem. The marketing creates one. I wrote about this when I argued that selling AI as replacement kills trust. The data now says it at population scale.
There is a second layer. AI is the first technology many people use daily while being told, at the same time, that it may take their job, flood their feeds and fake their news.
Usage is private and practical. Opinion is formed in public, from headlines. The two run on different inputs, so they move in different directions.
What this means for anyone selling an AI product
Stop leading with the letters. "AI-powered" was a differentiator in 2023. In 2026 it is, for a large share of buyers, a mild warning label.
Lead with the outcome instead. Not "AI-powered scheduling", but "your calendar fixes itself before Monday". Not "AI insights", but "the three accounts likely to churn this month". The customer wants the result, and the result is what they are happy to pay for.
Make the off-switch obvious. Products that let people choose when AI acts, and show them what it did, earn more trust than products that act silently. Control is the thing the anxious majority is asking for.
Keep the human route open, and make it visible. A chatbot that hands over to a person in one step reads as service. A chatbot that loops is the single strongest generator of the resentment in those polls.
What this means for everyone else
Even if you do not sell AI, you use it, and your customers can tell. The same voters who dislike data centres can spot an automated reply, a generic image and a summary nobody checked.
That puts a premium on visible human judgement: named authors, specific opinions, real photographs, and answers that clearly came from someone who understood the question. I covered the playbook for this in how to market to people who distrust marketing, and AI has made that audience much larger.
Use AI heavily in the work behind the scenes: research, drafts, analysis, operations. Be careful about where it shows up in front of the customer. The public tolerates AI as a tool far more than AI as the face of your brand.
The gap will not close soon
Some people expect sentiment to catch up with usage, the way it eventually did with smartphones. I am not so sure. Smartphones did not arrive with a public debate about replacing their owners.
The more likely future is a long period in which both lines keep rising: more usage, more scepticism. That is a stable state for a market, not a temporary glitch. Brands that plan for it will message differently from brands that wait for the public to come around.
The same paradox runs inside your own company. Employees use AI tools daily and still distrust the rollout memo that announced them. Change management that leads with efficiency targets reads like a threat. Change management that leads with a specific task people hate doing reads like help, and adoption follows the second one.
Your customers already decided how they feel about AI. They are just not going to let it stop them using it.