24 Sept 20265 min read

How to audit your AI search visibility

Find the answers that leave you out. Build a clearer, better-supported picture of your business where buyers are looking.

The Vyera Team

Product & Research

An AI visibility audit checks whether your business appears in AI-generated answers, whether your website is cited and whether the information is accurate. Start with a fixed set of buyer questions, record the answers and sources, then repair the gaps that affect real purchase decisions.

Start with the questions buyers actually ask

Choose one audience, one market and ten questions. Use sales conversations, support requests and search data where available. Include a mix of discovery, comparison and verification questions.

  • Discovery: "Which competitor monitoring tools suit a small B2B SaaS team?"
  • Problem: "How can a founder track competitor pricing changes?"
  • Comparison: "How do monitoring tools differ from manual competitor research?"
  • Fit: "Which tools combine competitor research with marketing workflows?"
  • Verification: "What does [your business] do, and who is it for?"

Keep branded questions separate from unbranded questions. Being described correctly when somebody already knows your name is useful, but it does not show that new buyers can discover you.

Record the answer and the evidence

Choose the AI services your buyers use. Run the same questions in fresh conversations and note the date, service, visible model or mode, language, location and whether web search was used. Save the complete answer and the source URLs. Repeat observations before declaring a trend.

Use this record for every response:

FieldWhat to capture
Question and contextExact wording, audience and market
Brand mentionPresent or absent in the answer
RecommendationRecommended, mentioned neutrally or excluded
Website citationExact page URL, if cited
AccuracyCorrect claim, error or uncertain claim
ActionSource or page to review, with an owner

A citation and a recommendation are separate observations. A page can be cited without the business being recommended. An answer can name a business without linking to its website.

Work out which problem you have

If the brand is absent, examine the cited sources and whether your site clearly answers the question. If the brand is present but misrepresented, find the current authoritative page and any conflicting older information. If a relevant page is unavailable to crawlers, fix access before rewriting it.

For Google's AI search features, normal search eligibility still matters: pages need to be indexable and eligible for snippets. Google does not require a special AI schema or an llms.txt file. See its AI optimisation guidance.

For ChatGPT search, check OAI-SearchBot access, including infrastructure rules that may block it. OpenAI documents search crawling separately from GPTBot's training controls; allowing training is not a prerequisite for enabling search crawling. See OpenAI's crawler documentation.

Repair one important page first

Start with the page most closely connected to the buyer's decision. A product page should state what the product does, who it serves, its main workflows and relevant limitations. A pricing page should explain billing units and plan conditions. A comparison page should support its claims with current evidence.

Make the core explanation readable as text. Use concrete examples, link supporting documentation and remove contradictions across your own pages. Correct profiles you control; request corrections to independent sources where appropriate. You cannot dictate what an independent publisher or AI system says.

Make your business easy to understand, and make the important claims easy to check.

1

10 buyer questions

2

3 visibility measures

3

1 priority page

These are starting recommendations for this workflow.

Measure progress without manufacturing certainty

Keep a stable question set and record mention rate, website citation rate and claim accuracy separately. Mention rate is the share of valid sampled answers that name your brand. Citation rate is the share that link to your domain. For accuracy, review identifiable claims against current evidence and keep uncertain claims separate.

Report sample sizes and failed requests separately. Compare like-for-like observations, and track relevant visits and enquiries alongside visibility.

A worked AI visibility baseline

Here is a small synthetic dataset to demonstrate the audit. It is not a test of Vyera or a measurement of any AI provider. Each row represents a different buyer question asked once under the same recorded test conditions.

Question intentBrand namedDomain citedClaim review
Find a suitable toolYesYesTwo accurate claims
Solve a monitoring problemNoNoNo brand claims
Compare two approachesYesNoOne inaccurate claim
Check a required integrationYesYesOne accurate claim and one uncertain claim
Verify product pricingYesNoOne accurate claim

This sample contains five valid answers. Four name the brand, so sampled mention rate is 4 ÷ 5 = 80%. Two cite its domain, so sampled domain citation rate is 2 ÷ 5 = 40%. These measures answer different questions and should remain separate.

There are six brand claims: four accurate, one inaccurate and one unresolved. Accuracy among verifiable claims is 4 ÷ 5 = 80%; verification coverage is 5 ÷ 6, approximately 83%. Reporting both prevents an audit from hiding uncertainty inside a favourable accuracy score. If there are no verifiable claims, record accuracy as not applicable.

The immediate task is to inspect the source behind the inaccurate claim. The unresolved integration claim needs product evidence before it can be scored. A missed citation alone does not establish a technical fault. With only one response per question, this is an illustration of the arithmetic, not a reliable performance estimate.

For your own baseline, record a fixed panel of ten questions across two chosen services, with three observations per question per service: 60 planned observations. Separate results by service and intent, report completion counts, and retain the same setup on the next run. Those sample sizes are an operating suggestion, not a statistical guarantee. Record product version, search mode, date, market and language alongside the responses.

How Vyera fits into the workflow

Vyera connects AI visibility tracking with marketing action. Use this audit to choose the buyer questions to monitor and the page improvements to prioritise.

Make your next fix count

Know which questions matter before you rewrite another page. Explore how Vyera connects market visibility with your next marketing action.

Frequently asked questions

No. A search ranking concerns a result's placement for a query. AI visibility can include a brand mention, recommendation or source citation, and can vary across services and repeated answers.

No. Clear content and technical access support discoverability, but they do not guarantee indexing, citations or recommendations.

Start with a weekly sample if you can review it consistently. Retest after material changes, while allowing for crawl delays and variation in generated answers.

Ready to
compete smarter?

Your market, watched while you build.

Free to start · No credit card required

teams already joined