What 41 people in downtown Toronto said about trusting AI
Asked on a downtown Toronto street how much they trust AI on a scale of 1 to 10, 41 people gave a number. The average was 5.6 and the median 6. Nobody said 10, two said zero, and the person who scored it highest said they still fact-check it.
What we asked
On 22 August 2025, people passing the corner of Bay, King and University in downtown Toronto were asked on camera one question: "On a scale of 1 to 10, how much do you trust AI?" 44 people were interviewed.
The numbers
- 41 of the 44 gave a numeric score.
- The average was 5.6 out of 10. The median was 6.
- Nobody said 10. The highest score was 9.5.
- Two people said zero, below the bottom of the scale they were offered.
- The other 3 answered without giving a number. They are not counted in the average.
Read 5.6 as a description of these 41 answers, not as a measure of Toronto or Canadian opinion. The people interviewed were whoever stopped, at one intersection, on one day.
A correction to an earlier figure
An earlier internal count gave an average of 4.9 across all 44 people interviewed. It had recorded the people who gave no number as scores. Recounted from the transcripts, with non-answers left out, the average is 5.6 from 41 scores. Any earlier use of 4.9 for this study is superseded by this note.
What people said around the number
The person who scored AI highest, at 9.5, said they still check what it tells them. Another described themselves as an AI optimist and added: "I will always verify." One of the people who would not give a number put the objection in terms of accountability rather than accuracy: "Human error is human error. If AI is erroring, that's totally different."
Why the index publishes this
Even people who trust AI say they check it. A Trust Index reading is built to be checked: every reading publishes its questions, the assistants asked, the number of runs and the spread, so a reader can see exactly what a number rests on.
Method and limits
- Sample: convenience sample at one downtown intersection on one day. Not random, not weighted, not representative.
- Measure: a self-rated answer to one question, on a 1 to 10 scale. Range answers (for example "seven or eight") were recorded at their midpoint.
- Source: transcripts of the filmed interviews. Five answers came from passages where the transcript was unclear; without them the average is 5.5. The figure is published to one decimal place for that reason.
- Who ran it: the interviews were filmed before the index launched.
- Not used in scoring: nothing from these interviews enters any Trust Index ranking or score.
- Later filming: interviews have since been filmed in New York and South Korea. They are not part of this note.
Notes are written and signed by a named Trust Index analyst. Corrections to a note are logged on the note with their date. To report an error, use the corrections page.
All notesRequest a correction