AI search source readiness checklist
1. Official business facts
A customer should be able to identify the organization, category, offers, people, locations, contact paths, policies and eligibility limits without piecing together contradictory pages.
2. Technical access
Important information should be available in indexable, readable HTML through a stable canonical URL. Search crawlers need permitted access and crawlable links. JavaScript, robots controls, login walls and rendering failures can all change what a system can retrieve.
Google states that its established search best practices remain relevant to AI features and that no special AI markup is required. OpenAI separately documents OAI-SearchBot controls for ChatGPT search. Neither source promises that an accessible page will be selected.
3. Clear page ownership
A system cannot confidently retrieve an official answer when several pages make different claims about the same service, location or policy. Declare one owner for each important fact and decision, then consolidate contradictions.
The title, H1, opening answer and body should agree about the page's job. A service page can carry depth and nuance without forcing the customer to infer what is being offered.
4. Extractable answers with necessary limits
Place the direct answer near the question it resolves. Define the entity and action explicitly, then add the qualifications, dates, examples and sources required to interpret it correctly.
5. Proof that matches the claim
An owned page can establish official facts. It cannot independently prove that the company is the best option, caused a customer's growth or is trusted across a market.
Map each performance or superiority claim to approved evidence. State the baseline, result, window and attribution limit when those facts are known. Remove or qualify the claim when they are not.
6. Corroborating public sources
Profiles, directories, partners, publishers, associations, reviews and forums can corroborate identity or opinion. They should agree with the official record where the same fact appears.
Do not manufacture the corroboration. Fake accounts, purchased reviews, coordinated votes, disguised promotion and invented third-party praise destroy the very trust the layer is meant to create.
7. Measurement that separates visibility from value
Record the exact question, platform, market, retrieval state, date, answer, cited URL and factual errors. Keep mentions, citations and recommendations separate. Then measure any observable visit, enquiry or sale without pretending every answer creates a click.
Retest on a fixed schedule or after a material source change. A single favourable answer is an observation, not durable market ownership.