AI is being used for more than editing emails in the workspace
Core finding
In a Verasight survey of 1,000 U.S. adults conducted August 12 to 18, 2026, 28.0% at least sometimes rely on AI for work requiring original thinking, judgment, synthesis or contextual understanding.
Another work question finds 33.7% at least sometimes rely on AI output they lack the expertise to check. Both percentages cover all adults, including those who are not employed or do not use AI at work.
AI is used for tasks requiring original thinking
Reliance on AI for this work is occasional for 18.0%, frequent for 6.5% and very frequent for 3.5%, totaling 28.0% who use it at least sometimes.
Another 14.5% rarely rely on it and 9.9% never do. The remaining responses are people who do not use AI for work (22.0%) or are not employed (25.6%).
Topline
28.0% of all Americans at least sometimes rely on AI for work requiring original thinking or judgment; 22.0% do not use AI for work and 25.6% are not employed.
n = 1,000 · Aug 12–18, 2026 · MoE ±3.2%
Some rely on output they cannot fully check
For tasks they lack the expertise to verify, 21.0% rely on AI output sometimes, 9.0% often and 3.7% very often.
This measures reported reliance, not whether the output was correct. It also does not establish that these are the same people who rely on AI for original thinking.
Topline
33.7% of all Americans at least sometimes rely on AI work output they lack the expertise to assess; 22.1% do not use AI for work and 23.1% are not employed.
n = 1,000 · Aug 12–18, 2026 · MoE ±3.2%
Editing work messages is another common use
Some 26.2% use AI daily or weekly to change the tone of work messages, including making an email sound nicer or more polite. Daily use is 12.8%, and weekly use is 13.4%.
Another 36.5% never use AI for this purpose, while 28.1% say it does not apply. Keeping those responses in the denominator avoids overstating use among employed adults.
Topline
26.2% of all Americans use AI daily (12.8%) or weekly (13.4%) to adjust the emotional tone of work messages; 28.1% say this use does not apply to them.
n = 1,000 · Aug 12–18, 2026 · MoE ±3.2%
Methodology
Full methodology →| Mode | Verasight panel recruited via random address-based sampling, random person-to-person text messaging, and dynamic online targeting |
|---|---|
| Population | US adults age 18+ |
| Field dates | Aug 12–18, 2026 |
| Base (unweighted) | 1,000 |
| Margin of error | ±3.2% |
| Module | AOM Survey #2026-152 |
| Sponsor | Verasight |
| Weight variable | weight |
| Weighting targets | age, race/ethnicity, sex, income, education, region, metropolitan status |
Sources
- reports.verasight.io/r/aom-2026
When using generative AI tools for work-related tasks, how often do you rely on AI-generated output for tasks that require original thinking, judgment, synthesis, or contextual interpretation?
- reports.verasight.io/r/aom-2026
When using generative AI tools such as ChatGPT, Microsoft Copilot, Gemini, or Claude for work-related tasks, how often do you rely on AI-generated output for tasks where you do not have enough expertise to confidently judge whether the output is accurate or appropriate?
- reports.verasight.io/r/aom-2026
For example, to make an email sound nicer or more polite.
Cite this topic
AOM Omnibus, fielded August 12-18, 2026, N=1,000 US adults age 18+, +/- 3.2%.