Of the AI audit’s 147 problems, 25 change a main result in the papers.
Number of problems an anonymous AI audit listed in Saez and Zucman’s research, by our verdict. We checked all 147.
Number of problems
Why it matters. It shows which of the audit’s claims hold up, so readers can weigh the papers fairly.
Cyan: the problems that change a main result, a number in a paper’s summary or main tables. A judgment call is a choice of method that reasonable experts make differently.
Source: SaneCite check of the AI audit of Saez and Zucman’s research. The full check. Data as of 9 Nov 2026.
Source: SaneCite check of the AI audit of Saez and Zucman’s research, data as of 9 Nov 2026 · sanecite.com/charts
Show the numbers
| Verdict | Value |
|---|---|
| Judgment calls experts disagree on | 94 |
| Change a main result | 21 |
| Change no main result | 20 |
| Could not check | 8 |
| Change a main result if you use the standard method | 4 |
Free to share, with credit.
Download the image, paste the caption, or link to this page. Please credit SaneCite (sanecite.com). The numbers come from our published verdicts; nothing here is picked by hand.
- Where the numbers come from.Each figure is printed in the paper, its appendix or the authors’ own files, or recomputed from the authors’ own data. The build checks every one against our published verdict file.
- What “the standard method” means.The way government statistics offices, or at least two peer-reviewed studies by other researchers, measure the same thing. It does not mean the paper’s result is wrong, and it says nothing about anyone’s intentions.
- Check it yourself.This chart's data · the full check · every verdict (JSON) · the AI prompts.