Key takeaways
- An AI visibility audit is a one-time baseline: how often AI engines mention, cite and describe a client correctly for real buyer questions.
- Build 30 to 50 prompts from the client's buyer questions, run them in six engines and score every answer on the same table.
- Add technical checks: AI crawler rules in robots.txt, indexability, structured data and ChatGPT Business Hub claim status.
- Deliver a findings deck with a prioritized fix list. Monthly tracking is a separate deliverable.
An AI visibility audit is the deliverable agencies sell when a client asks whether ChatGPT, Gemini or Google's AI features recommend them, and nobody has a straight answer. It is a one-time project with a fixed method, which makes it easy to resell as part of white label SEO: you get a baseline, a list of reasons and a fix plan. It is not the monthly report. Ongoing tracking belongs in GEO reporting for clients; this audit is what you run first so that the monthly numbers have something to compare against.
What does an AI visibility audit cover?
It covers three things: answers, sources and access. Answers means testing buyer questions in the main AI engines and scoring mentions, citations, accuracy and sentiment. Sources means finding which pages and third-party sites the engines rely on. Access means checking that AI crawlers can reach and read the client's pages.
| Audit block | What you check | Output |
|---|---|---|
| Prompt testing | 30 to 50 buyer questions across six engines | Scored answer table |
| Competitors | Which brands appear instead of, or next to, the client | Share of mentions by brand |
| Sources | Pages and third-party sites cited in answers | List of sources to win or fix |
| Accuracy | Prices, locations, services and claims stated in answers | Error list with the correct fact |
| Technical access | Robots.txt, indexability, structured data | Pass or fix per item |
| Profiles | ChatGPT Business Hub, Google and Bing profiles | Claim status and data gaps |
How do you design the prompt set?
Start from the questions buyers actually ask: sales call notes, support tickets, site search, Search Console queries and People also ask boxes. Write them in natural language, group them by funnel stage and service, and include comparison and local versions. Freeze the list before testing so results can be repeated later.
- Ask the client's sales team for the ten questions prospects ask before buying.
- Mix stages: problem questions, "best" and comparison questions, and brand questions such as "is [brand] good for...".
- Add location versions for local clients, written the way a person would ask.
- Include two or three questions where a wrong answer would hurt: prices, service areas, licensing.
- Record each prompt with an ID, service, stage and the answer the client would want.
Keep the prompts separate from the keyword map; the two overlap but answer different questions. Our keyword mapping template owns the page-level side.
Which AI engines should an AI visibility audit test?
Test the engines the client's buyers use: ChatGPT, Gemini, Perplexity, Google AI Overviews and AI Mode, Microsoft Copilot and Claude. Run each prompt in a clean session with no personal history, note the date, the location and whether web search was used, and save a screenshot or copy of every answer.
Answers change between runs, so run each prompt at least twice and record both. For Google, AI Overviews and AI Mode draw on pages that are indexed and eligible for a snippet; Google says there are no additional requirements or special optimizations needed. For Copilot, Bing Webmaster Tools now has an AI Performance report showing citations in Copilot and Bing AI answers, so pull it if the client has a verified Bing property.
How do you score AI answers consistently?
Use one row per prompt and engine, and five columns: mentioned, cited with a link, accurate, sentiment and competitors named. Score with yes, no or partial, not with invented points. Two people scoring a sample of the same answers is the easiest way to catch inconsistent judgments before the client sees them.
| Column | Values | Notes |
|---|---|---|
| Mention | Yes / no | Brand named anywhere in the answer |
| Citation | Yes / no + URL | Client page linked as a source |
| Accuracy | Correct / error / not stated | Copy any wrong fact word for word |
| Sentiment | Positive / neutral / negative | Based on how the brand is described |
| Competitors | Brand names | Every other business named in the answer |
Totals then come from counting rows, for example mentions in a given number of the 40 ChatGPT answers. Do not convert that into an "AI rank"; the reasons are in our GEO reporting guide.
What technical checks belong in an AI search audit deliverable?
Check that the crawlers each engine uses for search are allowed in robots.txt, that key pages are indexable and render without scripts, and that structured data describes the business correctly. Separate search crawlers from training crawlers, because blocking one does not block the other, and the client may want different rules for each.
| Crawler | Used for | Documented by |
|---|---|---|
| OAI-SearchBot | ChatGPT search answers | OpenAI |
| GPTBot | Training OpenAI models | OpenAI |
| PerplexityBot | Surfacing and linking sites in Perplexity | Perplexity |
| Claude-SearchBot | Search quality in Claude | Anthropic |
| ClaudeBot | Training data | Anthropic |
OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, while GPTBot is a separate setting. User-triggered fetchers such as ChatGPT-User and Perplexity-User may not follow robots.txt, per the same docs. For the rest of the crawl checks, use our robots.txt audit and the technical SEO audit checklist rather than repeating them. Microsoft's guidance on inclusion in AI search answers also lists crawlability, metadata and schema as the baseline.
Should the audit check ChatGPT Business Hub?
Yes, as a status line. Record whether the business is claimed in ChatGPT, who holds admin, whether the agency has access and whether the details match the website. Claiming does not guarantee more mentions, but an unclaimed or wrong profile is an easy fix to list. Note that the product is a beta.
Search Engine Journal's walkthrough of the claim flow explains verification and agency invites. Ask the client who holds admin before the audit starts, because the fix usually needs their ID check.
How do you package and present the audit white label?
Deliver it under your brand as a short deck plus the raw scoring sheet: a one-page summary, the headline counts per engine, the competitor view, the accuracy errors, the technical findings and a prioritized fix list with owners. Present the method first, so the client understands the numbers are a dated sample, not a ranking.
Findings deck
Ten to fifteen slides in your template, method on slide two, fixes last.
Scoring sheet
Every prompt, engine, answer copy and score, so the client can check it.
Fix plan
Content, entity, technical and profile fixes ranked by effort and impact.
Our wholesale AEO and GEO visibility audit starts from $400, with monthly monitoring as a separate line; details are on the white label GEO and AEO service page. The fixes usually lead into the work described in how to get cited in AI answers. Or book a 30-minute call.
Frequently asked questions
How long does an AI visibility audit take?
It depends on the number of prompts and engines; plan for about a week once access and the prompt list are agreed. The first days go to agreeing the prompts, competitors and engines, then running the tests and recording each answer. The final days go to the written findings and a prioritized list of fixes.
Can tools replace manual prompt testing?
Tools help with volume, but read a sample of answers yourself to check accuracy and context. Check whether the brand is mentioned, whether the facts are correct, which sources are cited and how competitors are described. Tools can miss errors in wording that only a person who knows the client will notice.
How often should the audit be repeated?
Once as a baseline, then again after major changes. Monthly tracking runs on a smaller fixed panel. Keep the prompts, engines and settings the same each time so results can be compared month to month. Run a fuller audit again when the client launches new services, changes locations or sees a shift in leads.
Does blocking GPTBot remove a client from ChatGPT?
OpenAI documents GPTBot as the training crawler; ChatGPT search uses OAI-SearchBot, which is controlled separately. Review the robots.txt file with the client and decide which crawlers to allow, based on whether they want to appear in AI search answers. Blocking the wrong one can remove the site from places the client wants to be seen.
Is an AI visibility audit only for large brands?
No. Local businesses benefit too, especially where wrong hours, prices or service areas appear in answers. Small businesses can run a short audit with a handful of prompts about their services and city. Fixing inaccurate details at the source, such as the website and business listings, often has a visible effect.
Want these agency guides first? Choose White Label SEO LATAM as a preferred source on Google, and my articles are more likely to show up for you in Top Stories.
Follow on GoogleAdd as a preferred sourceSources
- Google Search Central, AI features and your website
- OpenAI developer docs, Overview of OpenAI crawlers
- Perplexity docs, Perplexity crawlers
- Claude Help Center, Does Anthropic crawl data from the web, and how can site owners block the crawler?
- Bing Blogs, New AI visibility insights in Bing Webmaster Tools (June 16, 2026)
- Microsoft Advertising, Optimizing your content for inclusion in AI search answers (October 2025)
- Search Engine Journal, How to submit your website to ChatGPT Business Hub (November 12, 2026)