Answers · Updated August 17, 2026
What is an AI visibility audit?
An AI visibility audit is a repeatable review of whether generative search systems can access, understand, cite, and accurately describe a business. It records prompt-level appearances and source URLs, checks entity facts and content eligibility, then connects referral or assisted visits to qualified leads and revenue without treating a citation as a ranking or guarantee.
What is an AI visibility audit?
An AI visibility audit is a repeatable review of whether generative search systems can access, understand, cite, and accurately describe a business. It records prompt-level appearances and source URLs, checks entity facts and content eligibility, then connects referral or assisted visits to qualified leads and revenue without treating a citation as a ranking or guarantee.
The audit is useful because four different outcomes are often collapsed into one number: being technically eligible, appearing in a generated answer, receiving a visit, and winning a customer. They are related, but none proves the next. A sound baseline keeps each layer visible and attaches the evidence needed to verify it.
Download the free CSV templateWhat should the audit measure?
Start with the sequence below. It prevents a common reporting error: presenting a citation count as though it were organic traffic or revenue. Each row has a different evidence source and a different acceptance rule.
| Layer | Question | Evidence |
|---|---|---|
| Eligibility | Can the system fetch, index, render, and select the page? | Crawler tests, directives, sitemap, URL inspection, server logs |
| Appearance | Was the business mentioned, cited, summarized, or recommended? | Versioned prompt sample with platform, market, date, answer, and cited URLs |
| Accuracy | Were the business facts and limitations stated correctly? | Approved fact sheet compared with the generated response and cited sources |
| Visit | Did an answer send a person to an owned page? | GA4 acquisition and landing-page reports plus server evidence where available |
| Qualified action | Did the visit produce an accepted call, form, booking, or sales-qualified lead? | Deduplicated key event plus destination or CRM receipt |
| Revenue | Did the qualified lead produce pipeline or collected revenue? | CRM opportunity and payment or invoice evidence with attribution limits |
| Change | What release, source, policy, or platform change may explain movement? | Release SHA, change log, prompt-set version, and before-after audit |
How do you run an AI visibility audit?
Use the same method for the baseline and each follow-up. Document every change to the prompt set or collection method. The aim is a decision record that another person can inspect, not a screenshot collection that cannot be reproduced.
1. Define the business facts
Write the approved name, category, offers, locations, service area, prices that are public, limitations, and canonical pages. This is the reference used to score factual accuracy; do not infer missing facts from an AI response.
2. Assign one page to each buyer intent
Map commercial, comparison, problem, research, and local questions to a single canonical owner. A new page is justified only when the result set shows a different intent, not because a synonym exists.
3. Verify technical eligibility
Check status codes, canonicals, robots rules, sitemap membership, rendered text, snippet eligibility, structured data, and the intended policy for named search crawlers. Record evidence instead of treating a successful fetch as proof of inclusion.
4. Freeze a prompt sample
Choose the buyer questions, platform, market, language, account state, date, and collection method. Keep the set versioned. Generated answers vary, so one screenshot is an observation, not a stable rank.
5. Record citations and correctness
For every response, capture the brands mentioned, owned and third-party URLs cited, the claims supported by each source, omissions, and factual errors. Separate an unlinked mention from a citation.
6. Review content and source gaps
Compare cited competitors and publishers with the current owner page. Improve direct answers, supporting detail, original evidence, entity clarity, internal links, and visible facts only where the evidence shows a real gap.
7. Connect visibility to business outcomes
Classify answer-platform referrals in analytics, preserve landing page and source, deduplicate conversions, and reconcile important leads to CRM or destination evidence. Report citations, visits, qualified actions, and revenue as different measures.
8. Repeat after controlled changes
Publish a dated change, submit changed URLs, verify the public release, and repeat the same sample. State when a platform change, low sample size, or missing account report prevents a causal claim.
What should an AI visibility audit not claim?
- Do not call citations rankings. Bing states that citations in its AI Performance report are references used in answers, not rank or authority scores.
- Do not promise placement. Search and answer systems select their own sources and can vary responses between users and runs.
- Do not invent universal uplift figures. Academic and vendor studies have bounded methods, datasets, and platforms; their findings do not guarantee the same result for a client site.
- Do not add markup for invisible claims. Structured data should describe content a visitor can verify on the page.
- Do not confuse referral traffic with a qualified lead. A key event needs a tested trigger and, for high-value actions, destination or CRM evidence.
Which official reports can support the audit?
Google says the same SEO fundamentals support eligibility for its AI search features and that pages still need to be indexed and eligible for snippets. Its newer guidance also emphasizes useful, unique content rather than a separate technical shortcut. Google announced dedicated generative AI performance reporting in June 2026 for a subset of Search Console users, so availability must be checked in the actual property rather than assumed. Read the official guidance: AI features and your website, AI optimization guide, and generative AI performance reports.
Bing Webmaster Tools added an AI Performance public preview with citations, cited pages, grounding queries, and trends. Bing cautions that citations are not rankings or authority scores. OpenAI advises publishers that want search discovery not to block OAI-SearchBot and recommends tracking ChatGPT referral traffic in analytics. See the Bing announcement and OpenAI publisher FAQ.
The term generative engine optimization was formalized in the GEO research paper. Treat its experiments as scoped evidence, not a universal formula. For implementation help, review Cognautic's generative engine optimization services, AEO explainer, and citation eligibility guide.
How should results be reported?
Report the baseline date, platforms, market, prompt-set version, sample count, owned and third-party citations, factual errors, referral sessions, qualified conversions, and revenue evidence. Add limitations beside the numbers. A clean report can say that crawler access and markup were verified today while indexing, citations, traffic, and sales remain future observations. That distinction makes the next action clear and keeps the audit honest.
People also ask
What does an AI visibility audit measure?
It measures technical eligibility, prompt coverage, brand mentions, cited pages and sources, factual accuracy, competitor inclusion, extractable answers, entity consistency, AI referral traffic, assisted conversions, qualified leads, and revenue evidence. The audit keeps those layers separate so a citation is not mistaken for a visit or a sale.
How do you check AI visibility?
Define a stable set of buyer questions, record the date, market, platform, response, brands mentioned, citations, and factual errors, then repeat the same sample on a schedule. Pair that record with Search Console, Bing Webmaster Tools, GA4, CRM, and server evidence where available. Results are samples because generated answers can vary between runs.
Can an AI visibility audit guarantee ChatGPT citations?
No. ChatGPT, Google, Bing, and other systems choose and vary their own sources. An audit can identify access problems, weak evidence, inconsistent facts, and missing answer coverage, then verify changes in a repeatable sample. It cannot force a platform to cite or recommend a specific business.
How often should AI visibility be audited?
Use a full baseline before major work, repeat a compact prompt and referral check monthly, and rerun the full audit after important site, brand, product, or platform changes. Keep the prompt set and collection method versioned so movement is interpretable rather than caused by an undocumented test change.
Is AI visibility the same as SEO ranking?
No. Organic ranking is a position in a search result set. AI visibility may be a citation, mention, summarized fact, recommended option, or referral from a generated answer. They share crawl, content, authority, and entity foundations, but need separate measurements and should both connect to qualified conversions.
How much does an AI visibility audit cost?
Cost depends on the number of products, locations, markets, prompts, platforms, and analytics systems in scope. Cognautic starts with a free consultation and provides a fixed written quote before paid work. The downloadable template on this page is free and does not require an email address.
Rather not DIY?
Need an AI visibility baseline tied to qualified demand?
If you’d rather have someone build this for you, that’s what we do. Start with a free consult — we map your workflows and name the smartest first move. No pitch, no pressure.