Anyone claiming to know the exact weighting of AI citations is overselling. Providers publish crawler controls and broad guidance, not a complete formula for commercial recommendations.
A useful operating model is still possible. Evaluate every important claim through five lenses: retrieval, relevance, clarity, corroboration and freshness.
Retrieval: can the evidence be reached?
The page should be publicly accessible, return a successful response, render meaningful content, use a sensible canonical URL and be linked from the site. Include it in the sitemap when it belongs in search.
Crawler controls matter. OpenAI distinguishes OAI-SearchBot, GPTBot and user-triggered access. Blocking one is not necessarily the same as blocking every OpenAI product behaviour, so follow current first-party documentation.
Relevance: does the page answer this question?
A business can be credible yet irrelevant to a particular request. State the service, audience, geography, constraints and use cases explicitly. A general homepage rarely answers a detailed buying question as well as a focused page.
Relevance is not keyword repetition. It is factual coverage of the decision the user is making.
Clarity: can the facts be extracted safely?
Put definitions, prices, processes and limitations in plain language. Use descriptive headings, lists and tables where they help. Keep important facts in visible page content rather than hiding them only in images or scripts.
Structured data can label the organisation, service, article or breadcrumb, but it should agree with the visible page. Marking up a claim does not prove it.
Corroboration: who else supports the claim?
Independent sources reduce ambiguity. Relevant reviews, customer pages, professional memberships, directories, public records and editorial mentions can connect the entity to a real service and location.
The source should be credible for the claim. Companies House can verify a legal entity; it cannot prove service quality. Customer evidence can support delivery; it does not independently set an industry-wide benchmark.
Freshness and consistency
Outdated prices, closed addresses and conflicting descriptions create risk. Maintain high-value pages and profiles when facts change. Show a real modified date when the content has genuinely been reviewed.
Consistency should not become duplication. Use the same core facts across profiles, but write each page for its purpose and audience.
Selection is still variable
Passing all five tests does not guarantee inclusion. Systems may weigh sources differently, personalise context or omit providers to keep an answer concise.
Report coverage over time, not absolutes. A transparent AI-search programme distinguishes improvements to the evidence from outcomes controlled by another platform.
The takeaway
Audit citation readiness through retrieval, relevance, clarity, corroboration and freshness. These improve the evidence available to answer systems, but they do not create a guaranteed ranking position.
Questions people ask
No complete public ranking formula exists for business recommendations. Use published crawler guidance and observable source evidence without inventing certainty.
No. Structured data can clarify visible facts, but it neither proves those facts nor guarantees selection.
Relevant, genuine reviews can corroborate service and reputation. They work alongside clear owned content and other trusted sources.
No. Keep them as visible text. Images can support the page, but key facts should not depend on image recognition.
It can be mentioned more often because it has accumulated clearer public evidence. A newer business needs to build verifiable evidence, not merely publish more claims.