Most free AI visibility checkers test a few prompts and turn the result into a dramatic score. That is useful for a sales conversation, not for running a search programme.
This checklist is designed to be repeated. Save the date, product, prompt, answer, citation and market conditions so the next audit can show an actual change.
Prompt and competitor baseline
List 30 to 50 prompts customers might genuinely use. Include service, problem, comparison, price, location and suitability questions. Avoid inserting your brand name into non-branded tests.
For each prompt record whether you appear, position or prominence in the answer, cited URL, named competitors, answer accuracy and any negative or outdated claims.
Technical access
Check robots.txt, response codes, canonical tags, rendered text, internal links and sitemap inclusion. Review the documented crawler names directly rather than copying an outdated bot list.
Test priority pages in Google Search Console. A page can be accessible to one crawler and still absent from a particular answer, but technical exclusion should be ruled out first.
Entity and business facts
Confirm the legal and trading names, current domain, address, phone, founders, service area and core services. Compare the site with Companies House, Google Business Profile, LinkedIn and relevant trusted directories.
List conflicts and obsolete profiles. Decide which source is authoritative for each fact and correct the public trail where you control it.
Content and proof
Check whether every valuable prompt has a page that answers it directly. Score the page for clear answer, scope, pricing, process, limitations, author, updated date, evidence, sources, FAQs and next action.
Separate claims from proof. A case study needs a named subject or a clear reason for anonymity, a real timeframe, method and verifiable outcome. Remove illustrative statistics that could be mistaken for client results.
Third-party corroboration
Record profiles, reviews, customer references, partnerships, memberships, awards, press and specialist directories. Judge whether each source is relevant and credible, not merely present.
Identify the sources that repeatedly support cited competitors. That becomes a focused outreach and reputation plan rather than a generic link-building target list.
Measurement and action plan
Confirm analytics can identify referral sources and that forms record lead source where appropriate. Add a simple ‘How did you hear about us?’ option because not every AI-assisted journey produces a clean referral click.
Prioritise fixes by commercial value, confidence and effort. Re-test monthly using the same core prompt set, then add new prompts separately so the baseline remains comparable.
The takeaway
The output should be an evidence matrix and a prioritised backlog: prompt gaps, technical blockers, identity conflicts, weak pages, missing corroboration and broken measurement. Reject opaque scores that cannot show their inputs.
Questions people ask
Thirty to fifty well-chosen prompts is a practical starting set for a small business. Quality and commercial relevance matter more than volume.
Monthly is usually enough for a small business. Keep the core prompts stable and record material site or market changes.
Yes. A spreadsheet with consistent fields is more useful than an opaque automated score, although repeated checks take time.
Track them separately. Branded accuracy matters, but it does not demonstrate discovery among people who do not already know the company.
Confirm the priority page is public, returns a successful response, is not blocked and contains meaningful rendered text.