Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
80 74 86 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
Insufficient record: 4 predictions resolved, 5 needed for a score.
-
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
85 74 94 -
Guidance
A company's own figures for its results over a period, checked against what it then reported.
Insufficient record: 0 guidance claims resolved, 5 needed for a score.
All types combined: 81 (likely 76 to 86, high confidence), from 86 claims: 70 factual claims, 16 promises. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Mark Zuckerberg's claims
4 claims
| Verdict | Kind | Said | Claim | Topic |
|---|---|---|---|---|
| Mixed | Prediction | 10 Jan 2025 | Requiring more confidence for takedowns will dramatically reduce the number of people whose Meta accounts are removed by mistake. Prediction · 10 Jan 2025 | Their own work or organisation |
| Mostly true | Prediction | 10 Jan 2025 | Changing Meta's content filters to require higher confidence and precision will reduce the vast majority of Meta's censorship mistakes. Prediction · 10 Jan 2025 | Their own work or organisation |
| Resolved true | Prediction | 18 Apr 2024 | A 1GW data center will be built eventually, but not in 2025 (next year). Prediction · 18 Apr 2024 | AI compute, hardware and data |
| Mostly true | Prediction | 28 Sep 2023 | Zuckerberg said he thinks Quest 3 is going to be the first mixed reality device that millions of people get. Prediction · 28 Sep 2023 | Technology |
“changing the content filters to have to require higher confidence and precision is actually going to be the thing that reduces the vast majority of the censorship mistakes that we make.”
Mark Zuckerberg, Meta CEO 10 January 2025 Transcript of Mark Zuckerberg on Joe Rogan Podcast (The Singju Post)
Changing Meta's content filters to require higher confidence and precision will reduce the vast majority of Meta's censorship mistakes.
Mostly true True in substance, wrong in a detail.
The evidence is the person's or their organisation's own records.
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
This is a prediction with no stated horizon. Judge it on evidence available by about one year after it was made (2026-01-10), using Meta's own enforcement metrics as published in its Community Standards Enforcement / Integrity Reports (enforcement mistakes, wrongful removals, appeals and restorations), together with independent analyses. True: after Meta raised its confidence thresholds for automated filters, reported enforcement mistakes (wrongful takedowns, or appeals that succeed or lead to restored content) fell by a large majority, roughly 50% or more, and the evidence links the fall mainly to the filter/threshold change. Mostly true: mistakes fell by a large majority but other policy changes made at the same time (such as narrower policies or ending fact-checking) contributed noticeably, or the fall was slightly under a majority while still large and clearly caused by the threshold change. Mixed: mistakes fell meaningfully but by well under a majority (roughly 25–50%), or the data is mixed across categories. Mostly false: mistakes fell only a little (under about 25%), or the fall is credited mainly to other changes. False: mistakes did not fall, or rose. If no measurement of mistakes exists, the claim is unresolvable.
Evidence
- Integrity Reports, First Quarter 2025 | Transparency Center Meta Platforms (Transparency Center) · 29 May 2025 Own record
“following the changes announced in January we’ve cut enforcement mistakes in the U.S. in half”
- Integrity Reports, First Quarter 2025 | Transparency Center Meta Platforms (Transparency Center) · 29 May 2025 Own record
“Requiring more confidence that content violates our policies before we remove it — by both raising the thresholds of our automated systems and in some cases requiring multiple layers of review before removal.”
- Meta 3 October 2025 Mr. Zuckerberg, I have the honour to ... Office of the United Nations High Commissioner for Human Rights (UN Special Rapporteur on freedom of opinion and expression) · 3 October 2025
“It claims a 75% reduction in enforcement mistakes on platforms in the United States compared to the previous quarter”
- Integrity Reports, H1 2026 | Transparency Center Meta Platforms (Transparency Center) · 19 March 2026 Own record
“Our rate of mistakes remained consistent during Q4 2025 with the exception of the number of false positives, which increased over 100% during a very brief period of the quarter due to a bug.”
Reasoning
Meta's Q1 2025 Integrity Report says that after the January changes it cut U.S. enforcement mistakes in half (E1). A UN Special Rapporteur letter from October 2025 notes that Meta claimed a 75% drop in U.S. enforcement mistakes compared with the previous quarter (E3), presenting it as Meta's own figure. Meta describes requiring more confidence before removal, both by raising automated thresholds and in some cases adding layers of review (E2); the evidence does not separate how much of the fall each step caused, so it does not credit the fall mainly to the threshold change. That fits "mostly true": Meta reports that mistakes fell by about half or more after the January changes (E1), but the evidence does not credit the fall mainly to the threshold change. The figures come only from Meta's own published reports, and E4 was published after the judgment date.
Cite this claim
Mark Zuckerberg, “Changing Meta's content filters to require higher confidence and precision will reduce the vast majority of Meta's censorship mistakes.” (said 10 January 2025). Verdict: mostly true. Reliability API, claim c_41e8bfe6622c, method 0.1.0+3f84d14d. https://epistacklabs.com/claims/c_41e8bfe6622c
Claims without a verdict
- Unresolvable 144
- Not yet due 17
- Own records only 41
- Not researched 223
- Not checkable 1,060
Is a verdict here wrong, or do you speak for Mark Zuckerberg? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.