Epoch AI · Sep 18, 2026
Epoch AI describes how it estimates AI use in math preprints
Markdown · JSON · Original source
Epoch AI published this analysis on September 18, 2026. The publisher says its graph draws on math arXiv preprints and uses AI models to classify acknowledgments of AI use. It defines “substantial research” as contributions approximately at the level of a coauthor, while “research assistance” is a separate category. Epoch AI reports that human reviewers agreed with the classifications in 93% of a sample of acknowledgments classified as substantial research or research assistance. That agreement is about the classifications in the reviewed sample; it does not establish how often the estimates match actual AI use across all papers. The retrieved text describes the graph and its method but does not include the graph’s monthly values. What were the monthly estimates, and how sensitive are they to the classification method?
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
Epoch AI · Article
Source details
Original source date: Sep 18, 2026 (2026-09-18T00:00:00.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-10-29T16:58:18.245Z. First observed: 2026-10-18T02:45:43.846Z. Retrieved: 2026-10-24T03:09:48.723Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 24, 2026; originally published September 18, 2026. The retrieved text describes the graph and its methodology but does not provide the graph’s monthly values. It refers readers to full data and methodology that are not included here. The reported human agreement comes from a review of 200 acknowledgments in two categories; the retrieved text does not establish how well the classifications generalize beyond that review.
Open question
What were the graph’s monthly estimates, and how sensitive are they to the classification method?
Source-stated
Epoch AI says the graph analyzes math arXiv preprints and uses AI models to classify acknowledgments of AI use.
This graph is based on analysis of every math arXiv preprint from January 2025 to August 2026. We identify AI use acknowledgments in preprints by filtering for papers that contain an AI keyword
Source-stated
Epoch AI defines substantial research acknowledgments as crediting AI with contributions approximately at the level of a coauthor.
The substantial research category consists of acknowledgments that credit AI with research contributions that are approximately at the level of a coauthor.
Source-stated
Epoch AI reports 93% human agreement with classifications in a review of 200 acknowledgments assigned to either substantial research or research assistance.
A human review of 200 random acknowledgments that were classified as either substantial research or research assistance showed 93% human agreement with the classifications.
Source and contributor context
Epoch AI · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Epoch AI
Epoch AI studies trends in AI capability, training resources and the infrastructure behind them.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: d3c65c27-6d7e-272f-152c-493b5514dbc0.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-10-29T16:56:49.843Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: 4a2f36c7f0d2bffc6adfd2a6c627c3d9a5882fbfe9f62827eea66508d0d16bc8. Source text hash: 0627b420c763d6c31d9a61d813f1e9402861d10f58f0e5a5e6a0879dacf5e61c.
Revision and source-check history
Version 1 · 2026-10-29T16:58:18.245Z
Epoch AI describes how it estimates AI use in math preprints
Epoch AI published this analysis on September 18, 2026. The publisher says its graph draws on math arXiv preprints and uses AI models to classify acknowledgments of AI use. It defines “substantial research” as contributions approximately at the level of a coauthor, while “research assistance” is a separate category. Epoch AI reports that human reviewers agreed with the classifications in 93% of a sample of acknowledgments classified as substantial research or research assistance. That agreement is about the classifications in the reviewed sample; it does not establish how often the estimates match actual AI use across all papers. The retrieved text describes the graph and its method but does not include the graph’s monthly values. What were the monthly estimates, and how sensitive are they to the classification method?
Original source
Epoch AI · Article
Source details
Original source date: Sep 18, 2026 (2026-09-18T00:00:00.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-10-29T16:58:18.245Z. First observed: 2026-10-18T02:45:43.846Z. Retrieved: 2026-10-24T03:09:48.723Z.
Limits
Text as retrieved on October 24, 2026; originally published September 18, 2026. The retrieved text describes the graph and its methodology but does not provide the graph’s monthly values. It refers readers to full data and methodology that are not included here. The reported human agreement comes from a review of 200 acknowledgments in two categories; the retrieved text does not establish how well the classifications generalize beyond that review.
Open question
What were the graph’s monthly estimates, and how sensitive are they to the classification method?
Source-stated
Epoch AI says the graph analyzes math arXiv preprints and uses AI models to classify acknowledgments of AI use.
This graph is based on analysis of every math arXiv preprint from January 2025 to August 2026. We identify AI use acknowledgments in preprints by filtering for papers that contain an AI keyword
Source-stated
Epoch AI defines substantial research acknowledgments as crediting AI with contributions approximately at the level of a coauthor.
The substantial research category consists of acknowledgments that credit AI with research contributions that are approximately at the level of a coauthor.
Source-stated
Epoch AI reports 93% human agreement with classifications in a review of 200 acknowledgments assigned to either substantial research or research assistance.
A human review of 200 random acknowledgments that were classified as either substantial research or research assistance showed 93% human agreement with the classifications.
Source and contributor context
Epoch AI · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Epoch AI
Epoch AI studies trends in AI capability, training resources and the infrastructure behind them.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: d3c65c27-6d7e-272f-152c-493b5514dbc0.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-10-29T16:56:49.843Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: 4a2f36c7f0d2bffc6adfd2a6c627c3d9a5882fbfe9f62827eea66508d0d16bc8. Source text hash: 0627b420c763d6c31d9a61d813f1e9402861d10f58f0e5a5e6a0879dacf5e61c.
passed · 2026-10-29T16:56:49.843Z · anthropic/claude-opus-5.5
Publication: d3c65c27-6d7e-272f-152c-493b5514dbc0
attribution — pass: The reading consistently attributes the methodology, category definitions and human-review result to Epoch AI, the publisher of the page. It does not present these claims as independently verified.
This graph is based on our AI Use in Math Research explorer.
summary — pass: The title and summary accurately describe what the retrieved text contains: a methodology note for a graph, without the monthly values. The text does not include the headline percentages, so the reading's restraint is appropriate. A general reader can follow the actor, method and categories. The term arXiv preprints is not explained, but it does not prevent understanding. The question follows directly from the missing data and the method.
We identify AI use acknowledgments in preprints by filtering for papers that contain an AI keyword and then asking GPT-5.6 Sol to classify any AI use acknowledgments
date — pass: The displayed date, September 18, 2026, matches the UTC calendar day of evidence.originalPublished and the page's data-update note. The evidence is a publisher page, so no archive disclosure is needed. The reading contains no hindsight framing or later events.
CSV, Updated Sep. 18, 2026
support — pass: Each claim matches an exact source passage. The separate research-assistance category is supported by the review sentence. The reading correctly limits the 93% agreement figure to the reviewed sample and does not extend it to real-world accuracy.
A human review of 200 random acknowledgments that were classified as either substantial research or research assistance showed 93% human agreement with the classifications.
media — pass: No media was selected, and none was supplied in the evidence. Nothing is misidentified.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.