Fact check
“OpenAI also included scores from AI benchmarks showing similar performance from watermarked and unwatermarked text.”
Reasoning
The provided evidence contains an explicit statement that OpenAI presented AI benchmark scores indicating comparable performance for watermarked and unwatermarked text. This directly confirms the claim.
On confidence: The claim is directly supported by the cited excerpt from The Verge article, which states that OpenAI included benchmark scores showing similar performance between watermarked and unwatermarked text.
Important context
The evidence comes from a reputable technology news source (The Verge) reporting on OpenAI's rollout of text watermarking and its accompanying benchmark results.
Evidence
Supporting (2)
- Tier 1 — Primary sourceOpenAI is adding text watermarking in ChatGPT and Codex | The Verge
and AI. An invisible, machine-readable watermark in text output is rolling out to ChatGPT and Codex, but only for users in the European Union at first. OpenAI says its textGrain watermarking “matched or exceeded” other approaches like…
- Tier 1 — Primary sourceOpenAI is adding text watermarking in ChatGPT and Codex | The Verge
unwatermarked text. But it notes that textGrain “does not guarantee reliable detection,” and says text watermarks don’t verify accuracy, determine who owns the text, measure how much a human contributed, or prove human authorship. Rolling…
Limitations
The evidence is limited to a single source excerpt; no additional independent verification or detailed benchmark data is provided in the excerpt. No independent origin corroborated this claim, so it cannot be verified and confidence cannot be high.
- Last verified:
- Oct 5, 2026, 3:45 PM CDT
- Pipeline:
- 2.1.0
- Claim type:
- Factual
Where this claim appeared
OpenAI is adding text watermarking in ChatGPT and CodexThe Verge · Stevie Bonifield