The indexes
ChatGPT's retrieval, in browse and search modes, is anchored on Bing's index, with model-side knowledge supplementing where retrieval is not invoked. Google AI surfaces (the Overview and Gemini's commercial behaviour) draw from Google's own index. The two indexes are not identical: subjects can be present in one and weak in the other for reasons that have nothing to do with their underlying authority.
The implication is that index-level work — indexability, schema accuracy, sitemap discipline — must be conducted in respect of both Bing and Google, not Google alone. This is a small operational point with material consequences for ChatGPT visibility.
The authority logic
Both systems weight authority. They differ in which authority they weight.
- ChatGPT weights editorial sources, professional bodies, and Wikipedia heavily; it is more willing to recommend a specific provider on the basis of editorial coverage.
- Google AI weights Google's broader trust framework, the Knowledge Graph, and Google Business Profile signals heavily; it is more conservative on direct recommendation but more confident on entity-level facts.
Subjects who hold strong editorial coverage but weak Google Business Profile signal score better on ChatGPT than on Google AI. Subjects who hold strong Knowledge Graph and GBP signal but weak editorial score the inverse. The single highest-leverage move for parity across the two systems is editorial coverage that updates Google's broader trust framework.
The content shape preference
Both systems reward citable content. They differ on what citable means in practice. ChatGPT will cite slightly longer passages, with looser attribution; Google AI prefers shorter, more attributed, more schema-anchored passages. A page that produces clean ChatGPT citations may need light editorial restructuring to produce Google AI citations.
Two scenarios, side by side
Scenario one — a metropolitan commercial law firm
On the prompt “best commercial litigation firm in [city]”, ChatGPT typically cites two to four sources: editorial coverage in the legal trade press, the firm's own credentialed-author pages, and a recognised legal directory. Google AI, on the same prompt, cites the firm's Google Business Profile, a recognised legal directory, and one editorial source — with markedly more conservative recommendation language.
Scenario two — a regional accounting practice
On the prompt “accountant for small business in [region]”, ChatGPT tends to cite trade press, professional body membership records, and the firm's own pages where credentialed authorship is visible. Google AI tends to cite the firm's Google Business Profile, the professional body member listing, and a regional editorial source — with strong preference for the local-intent sources.
A closing observation
ChatGPT reads the way a journalist reads. Google AI reads the way a directory editor reads. Subjects who succeed across both have made themselves the kind of business that survives both readings.
The institution's engagements are designed to lift parity across the two surfaces. Where parity cannot be reached in a single quarter, the work is sequenced: ChatGPT-favouring moves first (editorial, credentialed authorship), Google AI-favouring moves immediately after (Knowledge Graph, GBP, structured data depth).