Surveys find people spot AI in some places, not others

Start here

Before recommending a setting or a habit, it helps to know what people typically get wrong about how these tools work, not guess. Two government-adjacent and independent surveys measured this directly rather than assuming it.

What the documents say

Both are surveys with stated methods, not lab experiments. Pew Research Center's February 2023 report, based on its American Trends Panel with 11,004 US adults surveyed December 12 to 18, 2022, asked people to identify whether AI was involved in six everyday activities, from spam filtering to product recommendations to fitness trackers. Only 30 percent identified all six correctly; individual items ranged from 51 percent correct for spam filtering up to 68 percent for fitness trackers analyzing exercise and sleep. Pew states these six questions are not an exhaustive list and that the survey measured whether people could spot AI's presence in an example, not deeper understanding of how the system works. Separately, the UK's Centre for Data Ethics and Innovation tracker survey, a nationally representative UK survey of 4,200 people fielded in December 2021 and published March 2022, found 52 percent said they knew only a little or nothing about how data about them is collected and used day to day.

Check this

A reader can treat these findings as pointing to two distinct, checkable gaps rather than one vague AI-literacy problem: whether someone can tell that a familiar product is using AI at all, and separately, whether they know what happens to information they type into it. The concrete check is to open a tool's own privacy or help page, which both surveys imply most people have not done, rather than assume either that AI is absent from a familiar tool or that its data practices match another product's.

What holds and what fails

Both findings hold as descriptions of self-reported awareness where and when each survey was fielded, the US in December 2022, the UK in December 2021, and neither claims to measure comprehension of how AI systems technically function, only whether respondents could identify AI's presence or describe their own data knowledge. It fails as a current-state claim: awareness of specific tools like ChatGPT has almost certainly shifted since either survey closed, and neither was repeated with the same questions to track a trend the way the usage surveys elsewhere in this series were. Treating a 2022-2023 gap as a permanent trait of 'the public' is editorial overreach neither document supports.

• Ask yourself whether a familiar app or device uses AI before assuming you would notice if it did.
• Open a tool's own privacy page once, rather than assuming its data practices match a different product you know better.
• Treat an awareness survey's percentage as a snapshot from its own field dates, not a permanent fact about the public.

Both surveys make the same underlying point in different ways: the gap is not that people know nothing about AI, but that specific, checkable pieces of knowledge, what counts as AI, and what a tool does with data, go unchecked far more often than people probably assume of themselves.