How to run an AI visibility audit — the complete step-by-step guide.

By Tom Howell, Co-Founder & CEO · Reviewed by the Visigent team July 2026 14 min read 3–4 hrs to execute

AI doesn’t rank — it recommends. An audit tells you whether you’re in the answer or invisible during discovery. You can’t fix what you haven’t measured.

+

Individual tests in a complete audit — 15 queries × 4 platforms. This thoroughness is what separates useful audits from surface-level checks.

Discovery is no longer about being seen. It's about being selected.

When a buyer asks an AI engine for a recommendation, there's no page two. There's an answer — and either you're in it or you're not. An AI visibility audit gives you a baseline measurement of where your brand appears (or doesn't) across the platforms reshaping how people discover products and services.

This guide walks you through the exact process: build your query list, run the tests, document the findings, and identify the gaps costing you recommendations.

What you'll need

✓ChatGPT accessFree or paid account

✓Google Gemini accessFree account works

✓Perplexity AI accessFree account works

✓Claude accessFree account works

✓A spreadsheet toolGoogle Sheets or Excel

✓3–4 hours of focused timePlus 15–20 customer queries

The audit, step by step

Build your query list

The quality of your audit depends entirely on the queries you test. Don't guess — use queries your actual customers ask. Aim for 5–7 in each of three categories:

Direct brand queries — include your brand name explicitly "What does [Your Brand] do?" "Is [Your Brand] good for [use case]?" "How much does [Your Brand] cost?"

Category queries — where your brand should appear as a recommendation "Best [product category] for [use case]" "Top [product category] tools in 2026" "[Product category] comparison"

Problem-solution queries — where AI might recommend you as the fix "How to solve [problem your product addresses]" "What's the best way to [task your product enables]" "Tools for [workflow your product improves]"

If you have Search Console data, pull the queries already driving traffic to your site — those are proven customer questions.

Pro tipInclude competitor brand names in a few queries ("Best alternatives to [Competitor]") to see whether AI engines recommend you in competitive contexts.

Create your audit spreadsheet

Set up a tracking sheet with these columns. This structure lets you analyze patterns later: which platforms favor you, which query types surface your brand, and where competitors dominate.

Run queries across all platforms

Now execute your queries systematically across each platform. This is methodical work — don't rush it. For each query:

• Open an incognito window. This prevents your history from influencing results.
• Enter the exact query from your list and wait for the complete response.
• Read the entire response. Don't just scan for your name — note the structure, the sources cited, and the recommendations made.
• Document everything: Did your brand appear, and where? Was your site cited? What exact language described you? Which competitors appeared? What sources were cited?
• Screenshot the full response with a clear filename like ChatGPT_BestCRMTools_2026-07-02.png.
• Log findings immediately. Don't trust your memory — capture details while they're fresh.

Repeat for every query on every platform. Yes — 15 queries × 4 platforms is 60 individual tests. That thoroughness is what separates useful audits from surface-level checks.

ImportantRun all queries within a 48-hour window. AI models update frequently — you want a consistent snapshot, not data spread across weeks.

Test with follow-up questions

AI conversations are dynamic. Sometimes your brand doesn't appear in the first response but surfaces when users ask follow-ups. For 3–5 of your most important category queries, ask natural follow-ups:

Example thread Initial: "Best project management tools for remote teams" Follow-up 1: "Which of these has the best mobile app?" Follow-up 2: "What about tools under $50 per month?" Follow-up 3: "Are there alternatives to [competitor mentioned]?"

Document whether your brand appears in follow-ups even if it didn't appear initially. This reveals whether you're being considered but filtered out by default parameters — price, features, use case.

Check citation sources

When AI platforms cite your brand, they're pulling from specific sources. Identify those sources — they're your current strengths. For every instance where you were mentioned and cited:

• In Perplexity: citations appear as numbered links. Click through to see the exact page.
• In ChatGPT, Gemini, or Claude: ask "What sources did you use to generate this answer?" or "Where did you find information about [Your Brand]?"

Record the source URLs and look for patterns: your own website, third-party reviews, Reddit threads, Quora answers, news articles, press releases. This tells you which content types AI platforms trust for information about your brand.

Analyze competitor visibility

Your audit isn't complete until you understand the competitive landscape. For every query where competitors appeared but you didn't, dig deeper. Build a competitor visibility matrix:

Example matrix. Yours should cover every category and problem-solution query you tested.

This reveals which competitors dominate AI recommendations, whether certain competitors own specific query types, and where no one is strongly positioned — your opportunities.

Watch the languagePay attention to how AI describes competitors. "Leading," "popular," "affordable," "enterprise-grade" — those descriptors are the positioning the model has learned.

"You can have 40% visibility on Perplexity and 10% on ChatGPT. That's not noise — that's actionable intelligence."

Calculate your AI visibility score

With all data collected, calculate your baseline metrics:

Mention Rate

(queries where you appeared ÷ total queries tested) × 100

Appeared in 12 of 60 queries = 20% mention rate

Citation Rate

(mentions with source citations ÷ total mentions) × 100

8 citations out of 12 mentions = 67% citation rate

Category Visibility — your most important metric

(mentions in category queries ÷ total category queries) × 100

Measures whether AI recommends you when users don't know your brand yet.

Platform Breakdown

mention rate, calculated separately per platform

40% on Perplexity but 10% on ChatGPT is actionable intelligence, not noise.

Identify visibility gaps

Now analyze your spreadsheet for patterns in where you're missing. Five gap types to look for:

Gap 01 · Platform gaps

Strong on one engine, invisible on others

Different AI models pull from different source ecosystems. Visigent research shows only 6.8% of cited domains appear on three or more platforms — optimization for one doesn't transfer automatically.

Gap 02 · Query type gaps

Found by name, invisible during discovery

If you appear in brand queries but not category queries, users can find you when they already know you exist — but you're invisible when they don't. This is the most damaging gap because it blocks new customer acquisition.

Gap 03 · Citation gaps

Mentioned but never cited

AI knows you exist but doesn't trust your website as an authoritative source. This points to issues with domain authority, content quality, or technical accessibility.

Gap 04 · Position gaps

Technically visible, practically invisible

If you consistently land in the last third of responses, you're in the answer but out of the decision. Users often stop reading after the first 2–3 recommendations.

Gap 05 · Competitive gaps

Competitors own queries you should contest

Review the sources AI cited for those competitors — that's what you need to replicate or surpass.

Document source patterns

Look at every source AI platforms cited across all queries — for your brand and competitors — and categorize them:

• Owned properties: brand websites, blogs, documentation
• Review sites: G2, Capterra, TrustRadius, Software Advice
• Community platforms: Reddit, Quora, Stack Overflow, niche forums
• News and media: press releases, news articles, industry publications
• Social proof: YouTube reviews, LinkedIn posts, X threads
• Wikipedia: brand or category pages
• Academic / research: studies, whitepapers, research papers

Calculate what percentage of citations comes from each category. If 70% of competitor citations come from Reddit and you have zero Reddit presence, that's a gap — and a to-do.

Create your baseline report

Compile everything into a report you can reference in 90 days when you re-audit. Include:

• Executive summary: overall mention rate, platform-by-platform breakdown, top three visibility gaps, competitor positioning.
• Detailed findings: query-by-query results, screenshots of key mentions and notable absences, source analysis, and the competitive matrix.
• Recommended actions: 3–5 priorities driven by your gaps — e.g. "Increase Reddit presence — competitors get 60% of citations from community platforms," or "Fix citation rate on owned content — mentioned but not cited 40% of the time."

Save the report with a clear date stamp. In 90 days, repeat the entire audit and compare results to measure improvement.

Common mistakes

• Testing too few queries. Five queries won't reveal patterns. You need 15–20 minimum for reliable data.
• Using only brand queries. You'll miss the bigger problem — invisibility during customer discovery.
• Not documenting exact language. The sentences AI uses to describe you (or competitors) reveal what the model has learned about positioning.
• Skipping screenshots. AI responses can change within days. Screenshots are your proof and reference.
• Testing across multiple weeks. Spread the audit over three weeks and you're comparing different model versions. Compress it into 48 hours.
• Ignoring follow-up behavior. Initial responses are just the start. Your brand may surface conditionally in follow-ups — that's data.
• Not recording competitor data. If everyone in your category is invisible, that's different from you being invisible while competitors dominate.

Common issues, solved

ProblemAI platforms give different answers when I repeat the same query.

Fix

Normal — generative AI includes randomness. Run each query once and document that result. Don't re-run seeking a "better" answer. The variability itself is data.

ProblemAI refuses certain queries or says it can't make recommendations.

Fix

Rephrase to be informational. Instead of "What's the best [product]?" try "What are popular options for [use case]?" Platforms provide information more willingly than explicit recommendations.

ProblemI can't tell if my brand was cited because there's no link.

Fix

Ask directly: "What sources did you use for this answer?" or "Where did you find information about [Your Brand]?" Most platforms reveal their sources when asked.

ProblemThe audit shows zero visibility — we don't appear anywhere.

Fix

Common for newer brands or those without strong online presence. It's not a failure — it's a baseline. Now you know the starting point and can measure every improvement from here.

ProblemWe appear for competitor comparisons but not category queries.

Fix

You're known in the market but not associated with the category itself. Publish category-defining content, not just competitive positioning.

ProblemDifferent team members get different results for the same query.

Fix

Use incognito browsing to eliminate personalization, and confirm everyone is on the same model version — different versions produce different results.

Now change the outcome

Your audit tells you where you stand. The next step is improving visibility where you're weak — this is where measurement ends and execution begins.

• Low mention rate in category queries: publish comprehensive content that positions you as a category authority. AI platforms need to see you discussed in category contexts repeatedly, across multiple sources.
• Mentioned but not cited: improve technical accessibility and content quality. Add structured data, improve page speed, and make sure your most important pages are crawlable.
• Competitors dominate specific platforms: analyze where their citations come from and build presence in those same ecosystems. If they're getting cited from Reddit, you need Reddit visibility.
• Platform-specific gaps: different AI engines trust different sources. Research which sources each platform favors and prioritize those.

Run this audit every 90 days. AI visibility changes faster than traditional search rankings — quarterly measurement lets you track improvement and catch problems before they compound.

The audit is the readout. The work is what changes the answer. Tracking your visibility doesn't move it — rewriting content with passage-level structure, building citations, and earning mentions in the sources AI actually crawls does.

Every statistic in this piece follows Visigent's research standards: named sources, dated figures, corrections on request to help@visigent.com.

Key Numbers

The 10 Steps

Build your query list

Create the audit spreadsheet

Run queries on all platforms

Test follow-up questions

Check citation sources

Analyze competitor visibility

Calculate your visibility score

Identify visibility gaps

Document source patterns

Create your baseline report

Platform behaviors, model versions, and citation mechanics reflect publicly observable behavior at the time of writing and may change. The 6.8% cross-platform citation figure comes from Visigent's research corpus. Re-verify platform-specific behavior each time you audit.

Or skip the spreadsheet. We’ll run the audit for you.

Get your AI audit: a full read on how ChatGPT, Gemini, Perplexity, Claude, and Grok see your brand — plus the execution plan to change the answer, not just track it.

Related research

Buyer’s Guide

Academic Paper

The Princeton paper, decoded. The academic foundation of GEO.

Comparative Analysis

Six platforms promise AI citations. Most don’t finish the job.