The 2026 AI Comfort Survey: what people really think about AI at work
We asked 124 people across 15 countries how comfortable they feel about AI in 30 everyday work and life situations. The results challenge some common assumptions about who trusts AI, and why.
124
people surveyed
15
countries
30
AI use cases rated
What we found
People are not rejecting AI. They are comfortable using it for research and first drafts, and deeply uncomfortable when it replaces a human in decisions about people. The biggest factor in their comfort was whether a real person had checked the work.
"My responses assume no human oversight. If there was, my answers would mostly flip from uncomfortable to comfortable."
5 findings that matter
1. Research and drafting are safe ground
Every group we looked at ranked research and creating first drafts as the most comfortable uses of AI. This held across age, location, industry, employment status, disability and neurodivergence.
4. Comfort rose with age, not fell
People aged 20 to 39 were the least comfortable group, averaging 2.37 out of 4. People aged 60 and over were the most comfortable, at 2.73. Younger respondents raised wider worries about creativity, the job market and the economy. Older respondents were brief and practical: a useful tool, just check it.
2. People draw the line at decisions about people
Job interviews, performance reviews, complaints and whistleblowing were the least comfortable uses in every group. Concern about AI in hiring came up again and again, without being asked about, including AI-written applications flooding employers and opaque scoring tools.
5. People can spot AI writing, and it costs trust
Respondents named clear signs of AI-written content: em dashes, "it's not X, it's Y" phrasing, overly polished or generic wording, and a lack of personal voice. One in four named the em dash first.
3. A human in the loop changes everything
Many respondents said their answers would flip from uncomfortable to comfortable if they knew a human had reviewed what AI produced. Comfort is conditional. People want to know a person shaped the result.
Who feels what?
Age
Comfort rose with each age band: lowest for 20 to 39 (2.37), highest for 60 and over (2.73)
Location
The smallest effect we found. Every country ranked the same uses at the top and bottom
Neurodivergent people
Slightly less comfortable overall (0.12 lower), with the biggest gap in customer-facing work and presentations. Many described AI as reducing overwhelm, organising their thoughts and checking tone
Industry
The five largest sectors sat within a narrow 0.32-point band
Comfort scores are averages on a four-point scale, where 1 is "not at all comfortable" and 4 is "very comfortable".
People with disability
The lowest comfort of any group, averaging 2.15 compared with 2.67 for others, spread evenly across uses. A small group, and a signal worth more research
Employment status
Full-time employees and the self-employed (109 of 124 respondents) showed only modest differences
What to do with these findings
Keep a human in the loop, and say so. Never publish AI output unedited. Where trust matters, tell people a person has reviewed it.
Use AI freely for research and drafting. This is the one area every group is comfortable with.
Keep AI out of decisions about people. Interviews, performance reviews, complaints and whistleblowing are a near-universal no.
Get ahead of concerns about AI in hiring. Publish a clear, plain-language policy on how you use AI in recruitment.
Edit out the tells. A final human edit for voice, punctuation and phrasing makes a real difference to credibility.
Don't assume your younger staff are your early adopters. Ask your people what they think, rather than guessing.
Involve neurodivergent people and people with disability. They are already finding value in places most AI rollout plans don't consider.
How we ran the survey
The Cuttlefish Communications AI Comfort Survey ran in June 2026.
124 people from 15 countries rated how comfortable they felt about AI in 30 work and personal situations, on a four-point scale from "not at all comfortable" to "very comfortable". Respondents could also add comments, which we used to understand the reasons behind their ratings. Most respondents were from Australia, with others from Canada, New Zealand, the Philippines and elsewhere.
The survey used a voluntary, self-selected sample, so results show patterns within this group rather than the whole population.
The final report includes a literature review to verify, challenge, or add to our survey findings.
FAQs
-
The 2026 AI Comfort Survey is original research by Cuttlefish Communications. It asked 124 people in 15 countries how comfortable they are with AI across 30 everyday work and personal uses.
-
Research and creating first drafts. Every group in the survey rated these as the most comfortable uses.
-
AI replacing humans in decisions about people: job interviews, performance reviews, complaints and whistleblowing.
-
Not in this survey. Comfort rose with age. People aged 20 to 39 were the least comfortable and people aged 60 and over the most.
-
Keep a human reviewing AI output and tell people they are doing it. Many respondents said this would change their answer from uncomfortable to comfortable.
-
Yes. Please credit "Cuttlefish Communications 2026 AI Comfort Survey" and link to this page.
Want to know what your people think?
We can run a briefing on these findings for your leadership team, or help you survey your own people before an AI rollout.
Fill out the form to request a briefing, and click the button below to download the full report.