Companion to “Is Your Website Ready for AI Search? 5 Checks That Matter.”
Revision 1. Documentation checked November 4, 2026.

Purpose and scope

Use this worksheet to identify discovery barriers, content gaps, and inaccurate answers for one important public service or topic. It works for a personal website, business, or public agency. It is an original suggested workflow, not a certification, ranking formula, or completed audit of any website.

Google Search is the main technical reference. The OpenAI crawler example illustrates why each additional service needs its own review. Settings and interfaces can change.

Start here

  1. Choose one public page and one task it should support. Add other pages after this first pass.
  2. Ask the content owner to verify the correct answer. Obtain Search Console access and technical help where needed.
  3. Record each check as Verified, Issue, Unknown, or Not applicable, with evidence and an owner. Unknown means unverified.

Use public information and fictional test scenarios. Make changes only within your authority. Keep authenticated or private material protected; robots rules are not access controls. [R1]

A. Technical discovery and settings

Check Where to look and what to do What verification looks like Typical owner
1. Public access Open the intended public URL signed out. Ask the technical team to investigate unexpected login walls, error pages, or bot challenges. The intended content loads; access problems have evidence and an owner. A browser success alone does not prove crawler access. Web/platform
2. Indexed page Run URL Inspection in Google Search Console. Review indexing status and the reason for any exclusion. Request indexing after fixing eligible issues. Recorded status and selected canonical URL, with unresolved issues explained. A live test alone does not prove indexing. Search Console owner
3. Page and server directives Inspect the HTML robots meta tag and HTTP X-Robots-Tag header. Review unintended noindex, nosnippet, and content-level data-nosnippet restrictions. [R2] Effective directives match the intended publication policy. The crawler can access the page to read those directives. Developer/platform
4. Robots and hosting rules Review /robots.txt and relevant CDN/firewall events for the specific affected pages and intended crawlers. Do not remove all protections or trust a user-agent name alone. [R1] A technical review explains intended permissions and any blocked legitimate access. Hosting/security
5. Sitemap Locate the CMS-generated XML sitemap or sitemap index. Check that intended URLs are present and use Search Console's Sitemaps report to investigate processing problems. [R3] A usable sitemap has consistent intended URLs. Sitemap inclusion helps discovery; it does not guarantee indexing. CMS/platform
6. Canonical URL Inspect rel="canonical"; compare with redirects, internal links, sitemap URLs, and Google's selected canonical. [R4] Signals consistently identify the preferred version. A canonical is a signal, not a command or universal requirement. Developer/CMS
7. Google AI inclusion Review Search Console → Settings → Search generative AI. Check the effective include/exclude setting and parent inheritance. [R5] The recorded effective setting matches the authorized decision. This is separate from ordinary search eligibility and model-training preferences. Property owner/policy owner
8. Other AI search access For ChatGPT search, review OAI-SearchBot robots rules and relevant hosting access against current published IP ranges. Review GPTBot training preferences separately. [R6] Search and training decisions are documented independently. Repeat provider-specific checks for other services you choose to support. Platform/security/policy
9. Rendered text and navigation Compare the visible page with Search Console's rendered inspection where available. Check whether essential text and links to it are present. [R7] Key instructions can be read and reached. Missing content or broken next-step links have a reproducible issue. Developer/content

B. Complete and maintain the answer

These are editorial recommendations for running the review, not a claim that each item is a separate ranking factor.

Verification: The service/content owner confirms the answer against authoritative information. This checklist cannot establish currency merely from a recent timestamp.

C. Test whether the source becomes a useful answer

This is a proposed repeatable exercise. It has not been run against a website as part of this package.

  1. Define expected answers. Start with five real audience questions. Have the content owner record required facts and the correct next-step URL.
  2. Run comparable searches. Record ordinary search and selected AI search products/modes, exact question, date, relevant locale, answer, and cited URLs. Use public information. Separate personalized results from a neutral baseline where possible.
  3. Review accuracy. Ask a knowledgeable human to check jurisdiction, eligibility, exceptions, dates, and whether the action link reaches the correct service.
  4. Repeat and classify. Use a second wording and repeat at a later time. Distinguish an owned-content error, access/retrieval issue, AI answer error, and task-completion problem.
  5. Fix and retest. Correct owned material through normal review. Report external errors through provider feedback when appropriate. Better source content cannot guarantee an external answer will change.

Copyable review record

Field Your entry
Website and public page URL
User task and exact question
Expected answer and authoritative source
Technical checks: verified / issue / unknown / not applicable
Search product, mode, date, and locale
Observed answer and cited URLs
Missing or incorrect conditions
Correct next-step URL and observed destination
Classification and consequence of issue
Responsible owner and action
Due date, retest evidence, and closure decision

Authored fictional example

Site: An imaginary city's permit page. Question: “Can I apply without a contractor?”
Expected answer: The fictional service owner says owner-applicants are allowed only in specified cases.
Illustrative issue: One supporting PDF contains that exception, while the main page omits it.
Suggested action: The owner verifies and incorporates the condition into the main service answer, checks the application link, and repeats the search review.

This is an authored scenario, not an observed AI response, actual jurisdiction rule, or measured result.

D. Keep measurement useful

Record search visibility, citation activity, referrals, answer accuracy, and completed tasks separately. Bing's AI Performance report supplies citation information for supported experiences; citations are not an accuracy score. [R9] Google's Generative AI performance report provides impressions across supported Google Search AI features. [R10]

Recommended management view: important task → verified source → current owner → open issue → corrective action → retest outcome. Avoid combining these into an unsupported “AI readiness percentage.”

Do not spend effort on a supposed Google requirement for special AI schema, llms.txt, or a mandatory writing pattern. Google does not require those for its AI search features. [R11]

Sources and maintenance

Documentation was reviewed on November 4, 2026. Recheck provider guidance before implementing changes.

Source inspiration: Searchable's six-step video, November 2, 2026. This companion is independently authored and adds operations, ownership, and answer verification. It does not reproduce the creator's paid tools or claim measured results.