Fact check AI news summaries and viral social claims
Last updated: July 26, 2026
To fact check an AI news summary or viral claim, extract each factual assertion and ground it against independent sources for a per-claim verdict. Mephistopheles flags unsupported claims and returns unverifiable for breaking news no authoritative source has confirmed yet, rather than guessing.
AI summaries misrepresent the news more than you would expect
AI assistants distort news at a rate large enough to have been measured by major broadcasters. In a study coordinated by the European Broadcasting Union and led by the BBC, released in October 2025, professional journalists across 22 public-service media organizations in 18 countries, working in 14 languages, evaluated over 3,000 responses from ChatGPT, Copilot, Gemini, and Perplexity. They found that 45% of AI answers contained at least one significant issue, 31% had serious sourcing problems, and 20% had major accuracy issues such as hallucinated details or outdated information (see the EBU release). It built on an earlier BBC study from February 2025 that first surfaced the problem.
The distortion was consistent across languages and markets, which points to a systemic issue rather than a quirk of one assistant.
How claim verification works for news
Paste an AI summary, a quote, or a viral claim into /chat. The steps:
- Claim extraction. Each factual assertion is isolated, including quotes attributed to named people. See claim extraction.
- Grounding. Each claim is checked against independent sources, including current web results and structured knowledge bases. See grounding.
- Verdict. Each returns supported, contradicted, disputed, or unverifiable, with the source used.
For a summary of a specific article, the check also catches the common failure mode the EBU study flagged: quotes and attributions that were altered or that never appeared in the cited source.
Why breaking news is often unverifiable
For events unfolding right now, the honest verdict is frequently unverifiable. If no authoritative independent source has confirmed a claim yet, Mephistopheles says so rather than treating early or unconfirmed reporting as fact.
Honest scope
A verifier can only confirm what independent sources already record. During a developing story, sources conflict, correct themselves, and lag reality, so a claim that is true may still return unverifiable simply because nothing authoritative has published it yet. Verdicts reflect the evidence available at the time of the check. This flags claims for your own judgment; it is not a newsroom's editorial process, and it does not adjudicate contested political disputes for you.
Why 'a source exists' is not enough
Viral claims often point to a real article or study that does not actually say what the claim asserts. The link works; the support does not. AI summaries make the same mistake, attributing to an outlet a fact the outlet never reported.
Mephistopheles checks whether the source actually supports the claim, not just whether the source exists. See citations vs claims.
Frequently asked questions
How do I fact check an AI news summary?
Paste the summary into /chat. Each factual claim and attributed quote is extracted and grounded against independent sources, including current web results. You get a per-claim verdict, so you can spot altered quotes, wrong attributions, and outdated figures, which the EBU and BBC study found were the most common problems in AI news answers.
Can it verify a viral social media claim?
Yes. Extract the claim, ground it against independent sources, and read the verdict. If the claim is genuinely supported, it shows the source; if no authoritative source confirms it, it returns unverifiable rather than guessing. That honest 'not yet confirmed' answer is often the most useful result during a fast-moving story.
Why does a breaking-news claim come back unverifiable?
Because no authoritative independent source has confirmed it yet. During developing stories, reporting conflicts and lags, so even a true claim can be unverifiable at the moment you check. Verdicts reflect the evidence available at that time, so it is worth re-checking as coverage settles.
Does it decide who is right in a political dispute?
No. It checks specific factual claims against independent sources. For genuinely contested matters it will return disputed and show the conflicting sources rather than pick a side. It flags claims for your judgment and does not adjudicate opinion or contested political interpretation.
Verify what your AI just told you.
Paste any AI answer and Mephistopheles checks each claim against independent sources — no sign-up to try.