Bar chart
How many survey participants use AI on open-ended questions
Compares three published estimates of how many online research participants use large language models to answer open-ended survey questions: 45% showing copy and paste behaviour in Rilla and colleagues' 2025 pilot, 34% self-reporting it in Zhang, Xu and Alvero's 2025 study, and up to 24% of crowdworkers using one even when banned, from Veselovsky and colleagues. Use it when deciding how much weight to give free-text survey answers.
Where this appears
What's in this bar chart
| Participants using an LLM | |
|---|---|
| Rilla et al. 2025, copy-paste logged | 45% |
| Zhang, Xu and Alvero 2025, self-report | 34% |
| Veselovsky et al., despite a ban | 24% |
Different samples and methods, so the figures are not directly comparable. Veselovsky is an upper bound.
Use this image
Free to use in your docs, decks and posts under CC BY 4.0, as long as you credit Koji with a link to this page. To embed it, paste this into your page:
html
<a href="https://www.koji.so/visuals/llm-use-in-online-survey-responses-three-studies"><img src="https://www.koji.so/visuals/llm-use-in-online-survey-responses-three-studies/llm-use-in-online-survey-responses-three-studies.png" alt="Bar chart of survey participants using AI on open-ended questions: 45% copy and paste logged, 34% self-reported, up to 24% despite a ban." width="1200" height="630" /></a>
<p>Source: <a href="https://www.koji.so/visuals/llm-use-in-online-survey-responses-three-studies">How many survey participants use AI on open-ended questions</a>, Koji</p>
