Next free-source huntWe check for new free data every weekday.

Posts

'Peer review' papers that name AI tools: 3.4% in 2023, 13.5% in 2026 so far

Of OpenAlex works with 'peer review' in the title, 3.4% in 2023 also named large language models, ChatGPT, generative AI or AI-generated content. In 2026 so far it is 13.5% (175 of 1,292). Keyword matches, not measured AI use.

ResearchAIScholarly data
Bar chart: share of OpenAlex works with 'peer review' in the title that also name AI tools, from 0.0% in 2021 to 13.5% in 2026.
Data: Publication years 2021 to 2026 (2026 to 18 November). Figures re-checked against the source on 2026-11-19.

Research about peer review is talking about AI far more than it did three years ago. We counted works in OpenAlex, the free index of scholarly papers, whose title matches "peer review" (articles, reviews and preprints). Then we counted how many of them also mention "large language model", ChatGPT, "generative AI" or "AI-generated" in the title or abstract.

In 2023, 18 of 525 such works matched (3.4%). In 2024 it was 43 of 573 (7.5%) and in 2025 85 of 956 (8.9%). In 2026 up to 18 November, 175 of 1,292 matched (13.5%). Going back further, 2021 had 0 of 912 and 2022 had 1 of 565.

ChatGPT was released at the end of November 2022, and 2023 is the first year with more than one match. Since then the share has risen every year, and 2026 is not a complete year yet. It can still move as OpenAlex adds late records.

OpenAlex is a free, open index of scholarly works, authors, sources and institutions, and its API can count works by year without downloading them. That makes a question like this one cheap to re-run: two queries, grouped by publication year, and the raw responses are small enough to keep. If you try other wording, such as adding the bare acronym LLM, read the caveats below first. The counts move noticeably.

What to keep in mind

  • These are keyword matches, not measured AI use. A paper can name ChatGPT to criticise it, to study it, or in passing.
  • OpenAlex's title search stems words, so it also returns titles such as "peer-reviewed", and some non-English titles (for example Portuguese "avaliação por pares").
  • The term list does not include the bare acronym "LLM". Adding LLM and LLMs raises the 2026 count to 216 (16.7%) and 2025 to 93 (9.7%). We kept the published figures on the original, narrower list.
  • The 2026 figures count works published up to 18 November (175 of 1,292, 13.5%). OpenAlex also lists one work dated after that day, which is left out.

How we got this

  1. Queried the OpenAlex works API with filter title.search:"peer review" and type:article|review|preprint, grouped by publication_year. This gives the totals.
  2. Ran the same query with title_and_abstract.search:"large language model" OR ChatGPT OR "generative AI" OR "AI-generated" added. This gives the matches.
  3. Divided matches by totals for each year and rounded to one decimal. We saved the raw API responses with SHA-256 checksums.

Source and terms

OpenAlex works API

Provider page · Signal Lab listing

Terms: CC0 (public domain). OpenAlex releases every record under CC0. Read the terms

Information and education only. Not financial, investment, legal or tax advice. Listings can be inaccurate, incomplete or out of date and are provided "as is". Free tiers, limits and licenses belong to each provider and can change, so check the provider's own terms before you rely on them. Terms & Disclaimer

All posts · RSS