How to Run a GEO Audit (Step-by-Step Process)

A GEO audit answers one question: when someone asks an AI model for a recommendation in your category, does your brand show up — and if not, why not? Traditional SEO audits check rankings, backlinks, and crawl errors. A GEO audit checks whether ChatGPT, Perplexity, Gemini, and Google AI Overviews know your brand exists, trust it enough to recommend it, and can find structured information to cite.

The gap is bigger than most owners assume. A 2026 study from Outrigger (the Outrigger Visibility Index — 1,004 businesses, 5 AI models, 95,392 data points) found that 65.9% of businesses are effectively invisible in AI search. They rank fine on Google. They just never surface when a real buyer asks an AI assistant "who's the best option for X?"

This guide walks through a DIY GEO audit using the same five-pillar framework Outrigger uses on client onboarding. You can run every step manually. At the end, you will have a pillar-by-pillar score and a ranked list of fixes.

AI visibility is not a single number you can move with one tactic. It is the sum of five distinct signals, each of which the models weigh differently. A complete GEO audit scores all five.

The pillars are ordered by how directly they map to results, not by how you should fix them. AI Presence is the scoreboard. The other four are the levers. A brand that scores zero on AI Presence almost always has a specific, diagnosable weakness in one or two of the other pillars — usually entities or citations — and directory and entity consistency was the top raw predictor of AI visibility in the same study.

Run each pillar in order. Score each 0-100. Then use the composite to decide where to spend the next 90 days. For the full scoring methodology behind each pillar, the five-pillar audit breakdown covers the formulas Outrigger runs automatically.

Start with the scoreboard. You are testing whether the models actually recommend you, so you have to ask them the questions a buyer would ask.

1. Build 10 buying-intent prompts. Use your real category, not your brand name. Examples: "What are the best [category] providers in [city]?", "Can you recommend a [category] for [use case]?", "[Competitor] alternatives for [audience]", "Who should I hire for [service]?" The mistake to avoid is searching your own name — of course the model knows your name if you paste it in. You want to see if you surface unprompted.
2. Run each prompt against four models. ChatGPT, Perplexity, Gemini, and Claude. Perplexity is especially useful because it shows the source URLs it cited, which feeds directly into your citations pillar.
3. Record three things per response: Were you mentioned at all? Were you recommended (not just listed)? Which competitors appeared instead?
4. Score it. If you appear in 0 of 40 tests, your AI Presence score is 0. Appearing in 5 of 40 is roughly 15. Being mentioned in 20 and recommended in 10 is around 55. A category leader recommended in 25+ scores 85 or higher.

While you are here, note the competitor share of model — how often each rival appears. This is your benchmark. If DistroKid-equivalent shows up in 70% of your category prompts and you show up in 0%, that gap is the number you are trying to close. The share-of-model metric explains how to track this over time, and you can compare your category against public AI visibility benchmarks to see what "good" looks like in your vertical.

The AI presence test is uncomfortable, and that's the point. You are watching, in real time, the recommendation your prospect gets instead of you. Write down the exact competitors that appear — those are the brands winning the citations you're missing.

If AI Presence is weak, entities and citations are almost always the cause. These two pillars are where most invisible brands leak.

Entity audit (your brand's identity across the web): 1. Search your exact brand name on Google and collect the top 20 results. Note which platforms you appear on: Google Business Profile, LinkedIn company page, Crunchbase, Wikipedia/Wikidata, industry directories, review sites. 2. Read the description on each. Do they agree? A brand described as "music licensing" on LinkedIn and "music distribution" on Google Business is sending the models conflicting signals, and conflicting signals suppress recommendations. 3. Check your website for Organization and LocalBusiness schema markup. Missing schema means the models are guessing at your basic facts. 4. Note whether a Google Knowledge Panel exists for your brand. If it does not, that is a major entity gap. Entity consistency is among the strongest raw predictors of AI visibility in the Outrigger study.

Citation audit (the sources AI retrieves): 1. Take the source URLs Perplexity showed you in Step 1. These are the exact pages the models pull from for your category. 2. For each one — usually Reddit threads, Quora questions, and "best of" listicles — check whether your brand is mentioned. In most first audits, the answer is no, and a competitor is mentioned instead. 3. Count the ratio: threads where you appear vs threads where a competitor appears vs threads with neither. A brand named in 3 of 142 relevant threads has a citation score in the teens — and an enormous opportunity.

The reason to run entities and citations together is that they reinforce each other. When brand mentions correlate ~3x more strongly than backlinks with AI visibility, the citation pillar is doing most of the heavy lifting — but only if your entity is consistent enough for the model to connect the mention to the right brand.

The last two pillars are the most fixable, which is why they often produce the fastest wins.

Review audit: 1. Find your brand on Google Reviews, Trustpilot, G2, Capterra, Yelp, and any industry-specific review site. 2. For each, record total reviews, average rating, and the date of the most recent review. Recency matters — a profile with 40 reviews where the newest is 18 months old signals a stalled business. 3. Run the same check on your top three competitors. Your review footprint is only meaningful relative to theirs. Reviews are heavily weighted for local and product recommendations, so a large gap here directly suppresses AI Presence. The reviews-for-AI-visibility playbook covers how to close it.

On-page audit: 1. Open your key service or product pages and ask: could an AI model extract a clean, quotable answer from this page? Pages built as marketing brochures — vague headlines, no structure, no specifics — are hard to cite. 2. Check for citable structure: clear H2 headings phrased as questions, direct answers in the first two sentences of each section, tables, and specific numbers. Content with structured sections and expert attribution is cited ~65% more often by AI models. 3. Confirm you have FAQ content and expert attribution (named author, credentials). These are the E-E-A-T signals models look for when deciding whose content to trust. 4. Verify AI crawlers can reach your site — check your robots.txt for AI crawler rules that might be blocking GPTBot, PerplexityBot, or Google-Extended.

Score each pillar 0-100, then compute your composite. A weighted average that leans on AI presence and entities gives you a defensible single number to track month over month.

An audit that ends in five scores is only half done. The value is in the ranked action plan.

Sort every gap you found by impact-to-effort. High-impact, low-effort fixes go first. For most brands the sequence looks like this:

1. Fix entity inconsistencies (high impact, low effort). Align your descriptions across LinkedIn, Google Business, and your site. Add missing schema. This is often a one-week fix that lifts multiple pillars.
2. Seed citations where competitors appear but you don't (high impact, medium effort). Target the exact threads Perplexity surfaced in Step 1. This is the fastest way to start appearing in AI answers.
3. Close the review gap (medium impact, low effort). Systematize review requests to lift volume and recency.
4. Restructure your top pages for citability (medium impact, medium effort). Rewrite service pages with question-format headings, direct answers, and specifics.
5. Build entity authority (high impact, high effort). Pursue directory listings, a Knowledge Panel, and third-party press over the following quarter.

The 90-day playbook sequences these into a concrete campaign, and Outrigger's features automate the measurement so you can re-run the audit monthly and prove movement.

If you would rather not run all five pillars by hand, a free AI visibility audit runs every check in this guide automatically — the AI presence prompts across four models, the entity and citation scans, the review and on-page analysis — and returns your composite score with a ranked action plan. It is the same audit Outrigger runs on new clients, and it takes about two minutes to start.