THE STAT THAT CONFUSED EVERYONE.
OtterlyAI published a finding in 2024 that has become inescapable in AI-search articles: roughly 80% of URLs cited by ChatGPT do not rank in Google's top 100 for the same query. The number has been corroborated by other observers in slightly different forms (HubSpot, Profound, several independent studies).
Most articles repeat the stat without explaining why. The mechanism is straightforward once laid out, and the strategic implication is significant: your AI visibility playbook is not your Google playbook.
BING INDEX DEPENDENCY.
ChatGPT retrieves from Bing's index, not Google's (article N. 11 covers the dependency in detail). Bing and Google have different indexes - similar but not identical. A page that ranks well in Bing may not rank in Google for the same query, and vice versa.
When ChatGPT pulls a citation from Bing, the underlying page is whatever Bing surfaced for that query at retrieval time. If that page is not in Google's top 100 for the same query, you get the OtterlyAI stat.
Implication: optimising for Google does not automatically optimise for ChatGPT. Bing has its own ranking signals (article N. 11), and ChatGPT inherits them.
CONTENT EXTRACTABILITY.
ChatGPT does not just rank pages; it extracts content from them and re-presents it. A page can be ranked highly in Bing but have content that is hard to extract (JS-rendered, paywalled, image-heavy). ChatGPT may skip it for citation in favour of a lower-ranked page with more extractable content.
Result: extractable but lower-authority pages can outrun authoritative-but-extraction-hostile pages in citations. This systematically biases citations toward pages that:
- Are server-rendered (not JS-only).
- Have clean semantic HTML with answer-first structure.
- Are not paywalled or behind heavy interstitials.
- Use schema markup for entity grounding.
RECENCY WEIGHTING.
ChatGPT (and especially Perplexity) weight recency heavily for many query types. A six-month-old article on a fast-moving topic may be cited above an older but more authoritative source. Google's ranking is more conservative on freshness; long-standing high-authority pages often hold their slots even when newer content exists.
This is why long-tail recent content survives in AI surfaces. The Wikipedia article on a topic may rank #1 in Google but a recent specific deep-dive can outrank it in ChatGPT's citation slot if the query is fresh-event-shaped.
DOES NOT OPTIMISE FOR CHATGPT.
THE STRATEGIC IMPLICATION.
Your Google strategy and your ChatGPT strategy overlap, but they are not the same. Specifically:
- Bing setup is mandatory for ChatGPT; Google ranking does not transfer.
- Server-side rendering is mandatory for ChatGPT; Google can render JS but ChatGPT cannot.
- Long-tail specific content can win in ChatGPT; the same content may rank poorly in Google because of authority/backlink gaps.
- Recency matters more in AI surfaces; freshness-decay is faster than in Google.
WHAT THIS DOES NOT MEAN.
It does not mean ignore Google. Most sites still get more traffic from Google than from AI surfaces; the AI surfaces are growing but not dominant. The right framing is two parallel optimisation tracks with significant shared work (technical SEO basics, schema, content quality) and a small set of divergent priorities (Bing, recency, extractability).
THE BOTTOM LINE.
The OtterlyAI stat is real and the mechanism is mundane. Bing is not Google. Extraction is not ranking. Freshness is weighted differently. Once you internalise the differences, the AI playbook is not mysterious - it is just a different optimisation surface that happens to overlap 60-70% with the Google playbook and diverge 30-40% in ways that compound over time.