How to Conduct an AI Visibility Audit: Complete Methodological Guide

TL;DR: An AI visibility audit measures how generative AI perceives, cites and recommends your brand. This guide presents our 7-step methodology: defining queries, multi-LLM testing, scoring, competitive analysis, technical diagnostics, action plan and monitoring. Applicable to any B2B or e-commerce site, with benchmarks from our 200+ audits in Belgium and France.

Why an AI Visibility Audit Is Essential in 2026

Most companies have no idea of their visibility with generative AI. They measure their Google rankings, organic traffic, conversion rate. But nobody knows whether ChatGPT mentions them when a prospect asks a question about their sector.

This is a strategic blind spot. According to a Gartner study (cited by the Financial Times, January 2026), 25% of informational searches in Europe now go through an LLM rather than Google. And this figure is growing at 40% per year. If your brand isn't cited in these responses, you're losing a quarter of your potential audience without even knowing it.

The AI visibility audit fills this blind spot. It answers 4 fundamental questions:

1. Do LLMs know you? Is your brand recognised as an entity in AI knowledge graphs?
2. Do LLMs cite you? On your key queries, are you mentioned in the responses?
3. Do LLMs recommend you? Is the tone of the citation positive, neutral or negative?
4. How do you compare to the competition? Who is cited in your place?

Our article on how to measure your AI Visibility Score covers the metric in detail. This guide focuses on the complete audit methodology.

Step 1: Define Your 20 Key Queries

The audit begins by identifying the queries your prospects ask AI. These are NOT classic SEO keywords. They are conversational questions, formulated as you would speak to an assistant.

Examples of the difference:

Classic SEO keyword	Conversational AI query
SEO agency Brussels	What is the best SEO agency in Brussels for an SME?
CRM software SME	Which CRM would you recommend for a 30-person SME in Belgium?
digital marketing training	What are the best digital marketing courses in France?
ERP comparison	Compare the 5 best ERPs for a 100-employee company

To build your list of 20 queries, we use 3 sources:

Source 1: Your sales team. What questions do prospects ask before signing? What objections recur? Each recurring question is a potential query.
Source 2: Google's "People Also Ask". Enter your main keywords in Google and note the PAA questions. Reformulate them in conversational style.
Source 3: Reddit and Quora. Search your sector on these platforms and identify the most upvoted questions. LLMs consume Reddit and Quora heavily in their training.

Distribute your 20 queries across 3 categories: 7 informational queries ("how", "why"), 7 comparative queries ("best", "difference between"), 6 transactional queries ("recommend", "which tool for").

Step 2: Systematic Multi-LLM Testing

Ask each of your 20 queries to the 4 main LLMs: ChatGPT (GPT-4), Perplexity, Gemini and Claude. Use the default versions (no "search" mode activated, except for Perplexity which searches natively).

For each response, note in a spreadsheet:

Direct citation: Is your brand named in the response? (Yes/No)
Link to your site: Does a link point to one of your pages? (Yes/No)
Citation position: If cited, in what position? (1st mention, 2nd, 3rd, etc.)
Tone: Is the citation positive, neutral or negative?
Competitor cited: Who is cited in your place?

This test takes approximately 2 hours for 20 queries x 4 LLMs. It is manual work, but there is not yet a reliable tool that fully automates this process. AI monitoring tools like Otterly or Peec can automate tracking over time, but the initial audit must be done manually to guarantee data quality. For tools, see our AI monitoring tools comparison.

Step 3: Calculate Your AI Visibility Score

The AI Visibility Score is a composite metric we developed at AISOS. Here is the formula:

AI Visibility Score = (Positive citations * 3 + Neutral citations * 1 + Links * 2) / (Queries * LLMs tested * 6) * 100

The denominator (6) represents the maximum score per query/LLM: positive citation (3) + link (2) + 1st position bonus (1).

Score	Interpretation	Benchmark (French-speaking B2B)
0-10%	Invisible: LLMs don't know you	55% of audited companies
10-25%	Fragmentary: cited on a few queries only	30% of audited companies
25-50%	Correct: solid foundation, optimisation needed	12% of audited companies
50%+	Excellent: topical authority recognised by AI	3% of audited companies

These benchmarks come from our 200+ audits conducted between 2025 and 2026 on B2B companies in Belgium and France. The reality is stark: 85% of companies have a score below 25%.

Steps 4-5: Competitive Analysis and Technical Diagnostics

Step 4: Competitive analysis. Repeat steps 2-3 for your 3 to 5 main competitors. Compare scores and identify who dominates which queries. You will often discover surprises: the SEO leader is not always the AI visibility leader. Smaller but better-structured players can dominate AI responses.

Step 5: Technical diagnostics. Use our 50-point SEO + AI checklist to audit the technical fundamentals of your site. The points most correlated with a good AI Visibility Score in our data are: the presence of structured data (schema markup), the existence of an LLMs.txt file, the quality of the HTML (text-to-code ratio), and the presence of structured FAQs.

For technical diagnostics, tools like Screaming Frog (European version, Brighton, UK) and Sitebulb (Manchester, UK) are indispensable. They identify HTML structure, schema markup and performance issues that LLMs penalise.

Steps 6-7: Action Plan and Monitoring

Step 6: Prioritised action plan. From the scoring and technical diagnostics, create a 3-phase action plan:

Phase 1 (weeks 1-4): Technical quick wins. Fix the robots.txt, add the LLMs.txt, implement missing schemas (Organization, Article, FAQPage). Expected impact: +5 to 10 points on AI Visibility Score.

Phase 2 (weeks 5-12): Strategic content. Create or restructure pillar pages, add FAQs, enrich existing content with factual data and tables. Expected impact: +10 to 20 points.

Phase 3 (month 4+): External authority. Obtain mentions in specialist media, contribute to sources that LLMs consume (Wikipedia, specialist forums). Expected impact: +5 to 15 points.

Step 7: Continuous monitoring. Repeat the multi-LLM test each month on your 20 key queries. Track the evolution of your AI Visibility Score over time. Use an AI monitoring tool (Otterly, Peec, ZipTie) to automate tracking between manual audits.

As Gianluca Fiorelli, SEO consultant based in Rome, notes: "The AI visibility audit is not a one-shot exercise. It's a continuous process, like SEO. LLMs evolve constantly, and your positioning in their responses can change within weeks."

FAQ: AI Visibility Audit

Conclusion: Measure Before Optimising

The AI visibility audit is the foundation of any serious AEO strategy. Without measurement, you're optimising blindly. With our 7-step methodology, you get a clear picture of your current position, your strengths, your gaps and the priority actions.

The first step is free: request your mini-audit and receive your AI Visibility Score within 48 hours. It's the most efficient starting point to understand your situation and prioritise your investments.

For the tools needed for the audit, see our SEO + AI tools comparison. And for the logical next step after the audit, read our AI content strategy guide.