How to Audit a Client's AI Visibility: A Consultant Workflow

More clients are asking a version of the same question: "Does ChatGPT recommend us?" Most agencies do not have a repeatable answer yet. This is the workflow we would use to turn that question into a scoped audit, a fix plan the client understands, and a monthly number you can report against.

1. Scope it around the questions customers ask

Before you scan anything, agree with the client on three to five search phrases, written the way a customer would type them into ChatGPT. For a local business that usually means service plus place: "emergency plumber in Katy", "kitchen remodel Cypress TX". For a B2B client it is the problem they solve: "payroll software for restaurants".

Get these from the client, not from a keyword tool. Ask what their best customers say on the first call. These phrases become the fixed set you test every month, so changing them later breaks your trend line.

2. Pick the pages that make money

For most small and mid-sized clients, five to fifteen pages are enough:

• The homepage
• Each main service or product page
• The main location or service-area pages
• The about or team page, where trust and authorship signals live

Skip the blog archive unless the client sells on expertise. Auditing forty posts rarely changes the plan, and it buries the findings that do.

3. Run the baseline

Scan the chosen pages with the client's main search phrase attached. In DidItIndex you can paste them as URL; search phrase, import them from the sitemap, or upload a CSV with the phrase in the second column. Each scan checks crawler access, page content, structured data and technical trust, and asks ChatGPT and Claude, with live web search, the client's question.

Save the baseline report. You will want it in three months when someone asks what changed.

4. Find out who AI recommends instead

The live test lists every site each assistant cited. That list is usually more useful than the score, for two reasons.

First, it shows which competitors are being named. Pick the one that appears most and run a side-by-side comparison. Both sites are tested with the same question, and the report summarizes where the competitor is ahead.

Second, it shows which sources the assistants trust for this market. In a local services test we ran, many cited pages were directories, review sites, a Reddit thread and a utility company's contractor list, not the businesses' own websites. Those sources become part of your plan, because getting the client properly listed and mentioned there is often faster than any on-site change.

5. Sort the fixes by who can make them

A list of forty failed checks is not a plan. Sort every finding into three groups:

• Blockers, fix first. Anything that stops AI crawlers reaching the site: robots.txt rules, firewall or CDN blocks, a noindex tag left over from a redesign. Nothing else matters until these are gone.
• Client can do it. Writing plain facts on service pages (warranty, license, hours, years in business), adding a page per location, completing directory listings, asking for reviews.
• Needs a developer. Structured data, server rendering for JavaScript-heavy sites, redirect chains, page speed.

Then order each group by severity. Critical items first, warnings next, informational items last or not at all. Clients respond better to five changes they can finish this month than to forty they cannot.

6. Explain it in their language

Your client does not need to know what a canonical tag is. Every DidItIndex report opens with a short plain-English "what to fix first" paragraph, written for a business owner. Use it as the opening of your summary, then add your own recommendations and pricing for the work.

Set expectations in writing. AI answers vary between runs and change as the web changes, so do not promise a placement. Promise the work, and promise to measure it the same way every month.

7. Re-check monthly and after every big change

Re-scan the same pages with the same phrases each month, or turn on monitoring so it happens on a schedule and you get an email only when something changes. Always re-check after a redesign, a plugin update or a hosting move. Those are the moments crawler access and structured data quietly break.

Report three things at each review: what was fixed, what the live tests show now compared with the baseline, and the next three fixes. That is a retainer line item the client can see the value of.

Tools for the job

The DidItIndex agency page covers bulk scanning, comparisons, monitoring and exports. For the manual side of the audit, our AI search audit checklist lists every item to check, and which competitors does AI recommend instead of you? goes deeper on the competitor step.