Plain-language answer
Is this search or SEO stat true, and is it still current after AI Overviews?
The Fossil Checker answers two independent questions about a stat: a truth verdict from one versioned claims dataset, and a freshness verdict that dates the stat's primary source against the May 14, 2024 US launch of AI Overviews.
Short answer
Use evidence that can survive a second reader.
- The truth verdict is a human-reviewed record; the freshness verdict is computed by comparing the primary-source date against the AI Overviews launch date, never against Google's algorithm.
- A stat sourced before the launch that describes the results page itself is a fossil, because it measured a page that structurally no longer exists; a ranking-factor or corporate fact is not dated out by the launch.
- A stat whose magnitude is disputed or superseded reads as contested, direction intact and specific number not.
- A stat with no matching record returns a Self-Date checklist rather than a fabricated verdict, and every verdict and freshness call is a versioned record reviewed at least quarterly.
Worked example
See the decision in a fictional case.
- Id
- seo-is-dead
- Claim Text
- SEO is dead. AI Overviews killed organic traffic.
- Verdict
- CONTESTED
- Real Finding
- The click side is real. Zero-click is now 68% and AI summaries roughly halve the traditional result click rate (Pew: 8% with a summary versus 15% without). But dead is false. Alphabet Search revenue grew 17% to 63.3B in Q2 2026 and Google holds about 91% of search. Organic click-through is shrinking. The channel is not dead.
- Do Instead
- Separate the two claims. Report the click decline with its dated source, and report Google's revenue and share growth alongside it. Plan for fewer informational clicks, not a dead channel.
- Category
- Market Position
Use SEO Stat Fossil CheckerInspect the full methodTake it into the matching Petrichor work
What this answer cannot establish
- The freshness rule dates the source, not the algorithm; it does not predict rankings, traffic, or performance.
- The dataset is a reviewed sample, so a no-match is a research lead, not proof a stat is false.