The AI Visibility Audit: A 12-Point Framework for Brands

AI visibility audit framework for brands

The direct answer

An AI visibility audit measures whether a brand appears in answers to important buyer questions. It then checks technical access, content, brand information, product data and third-party evidence to identify possible causes. A useful audit sets out which improvements to make first. It does not stop at screenshots or a proprietary score.

What the audit must discover

The audit has four jobs: Establish the baseline, identify material errors, explain competitor advantage and rank fixes by commercial impact.

That requires evidence from both the answer layer and the web environment behind it. Testing ChatGPT alone is not an audit. Crawling the website alone is not an AI visibility audit either.

A free checker can reveal a symptom. A proper audit explains the technical, content and authority conditions creating it.

The 12-point framework

1. Commercial prompt map

Define the questions that matter to discovery and revenue. Include brand, category, fit, comparison, reputation and action prompts. Tag them by market, persona and value.

2. Cross-platform baseline

Run the prompt set across selected products such as ChatGPT search, Google AI features, Perplexity and Gemini. Record inclusion, position, description and citation. Include Claude or other surfaces where the audience uses them.

3. Competitor recommendation gap

Identify who appears when the brand does not. Look for patterns by segment and prompt type. The aim is to explain why the competitor has a stronger evidence position rather than merely count losses.

4. Message accuracy

Check the facts that could change a decision: Category, location, availability, audience, price model, features, ingredients, security, credentials and policies. Rank inaccuracies by commercial or regulatory risk.

5. Citation-source map

List the domains that support answers and classify them as owned, editorial, review, marketplace, community, institutional or other. Find repeated sources and topics where no credible source represents the brand.

6. Crawl and index eligibility

Review robots directives, rendering, canonicals, status codes, internal links, duplication and snippet eligibility. For Google generative Search a page still needs to be indexed and eligible within Search.

7. Entity consistency

Compare names, descriptions, addresses, people, products and relationships across the site, structured data, profiles and material third-party pages. Resolve conflicts at their authoritative source.

8. Content usefulness

Assess whether high-value pages give a direct answer, original value, evidence, expert context, update dates and useful next steps. Identify pages that exist only to target a phrase.

9. Product or service evidence

Check specifications, limitations, comparisons, methods, use cases, proof and commercial terms. Determine whether an external evaluator could understand fit without contacting sales.

10. Independent authority

Review relevant media, reviews, partner ecosystems, public experts and community discussion. Measure relevance and substance rather than raw mention volume.

11. Measurement readiness

Confirm analytics can identify known AI referrers, CRM fields capture self-reported discovery and the organisation can retain prompt observations over time. Define metric formulas before the first report.

12. Governance and change control

Assign owners for product facts, expert review, structured data, PR claims and monitoring. Set a response process for harmful inaccuracies. AI visibility is a continuing operating responsibility rather than a one-off content project.

When choosing how to deliver the audit, compare AI visibility platforms and agencies.

Turn the findings into action with our framework for improving AI search visibility.

Prioritise with the IMPACT model

Score each finding against:

Intent value: Does it affect a buying or merely informational prompt?
Materiality: Could it change trust, fit or conversion?
Prevalence: How often does it appear across tests?
Ability to fix: Can the organisation control the relevant source or system?
Confidence: How strong is the evidence linking the finding to the result?
Time to value: How quickly can the change be implemented and observed?

Do not multiply arbitrary numbers into a pseudo-scientific priority. Use the criteria to make the trade-off visible then apply judgement.

What the final audit should contain

A useful deliverable includes the prompt methodology, raw baseline, competitor view, accuracy risks, citation-source map, technical findings, content and authority gaps, measurement plan and a 90-day roadmap with owners.

Separate three categories of recommendation:

Platform-supported fundamentals: Crawlability, indexing, helpful content and accurate structured data.
Established marketing practice: Expert content, product truth, PR and customer evidence.
GEO experiments: Changes whose cross-platform effect is plausible but not guaranteed.

That separation makes the audit more honest and easier to approve.

What an audit should not promise

No audit can reveal a universal LLM ranking algorithm. No single run represents every user. No provider can guarantee a recommendation. Avoid claims that llms.txt, hidden prompts or special “AI schema” will unlock visibility. Google explicitly says its Search features do not require those devices.

Frequently asked questions

How long does an AI visibility audit take?

Scope determines timing. A focused single-market audit can be completed faster than a multi-brand international portfolio. Allow enough time for repeated testing, source analysis and technical investigation rather than a single automated scan.

What access does the auditor need?

Public testing can establish a baseline. Search Console, analytics, product feeds, CMS access, CRM data and existing research improve diagnosis and implementation planning.

Is a free AI visibility checker enough?

It can reveal a symptom. It rarely explains source selection, technical barriers, entity conflicts or commercial priority. Use it as triage rather than a strategy.

When should the audit be repeated?

Monitor the stable prompt set continuously and run a deeper audit after major launches, migrations, positioning changes or material platform shifts.

Sources

Google Search Central, generative AI optimisation guide: Google’s AI optimisation guidance
Google Search Central, helpful people-first content: Google’s guidance on helpful content
OpenAI, ChatGPT search: Introducing ChatGPT Search
IgniteStack, audit and gap-analysis model: IgniteStack