How to Run an AI Visibility Audit: A Step-by-Step Playbook

42% of enterprise buyers now consult ChatGPT or Perplexity before visiting a vendor's website. That number was 11% in early 2024. If you have not checked how AI platforms describe and recommend your brand, you are flying blind on a channel that influences nearly half your pipeline.

An AI visibility audit answers three questions. Does AI mention your brand when buyers ask about your category? Does it cite your content as a source? Does it recommend you, or does it recommend your competitors?

This is the process we use at Cite Solutions to audit brands across every major AI platform. You can run it yourself.

Why an AI Visibility Audit Is Different from an SEO Audit

SEO audits check whether your pages rank on Google. AI visibility audits check whether AI systems recommend your brand. These are different things.

Only 12% of URLs that ChatGPT cites rank in Google's top 10 results. A page can sit at position one on Google and be completely absent from AI answers. A page that does not rank on Google at all can be the primary citation source for ChatGPT or Perplexity.

The signals are different. Google evaluates backlinks, page authority, and keyword relevance. AI systems evaluate passage extractability, factual density, third-party validation, and content freshness. A strong SEO audit and a strong AI visibility audit will surface different problems and different priorities.

For background on how AI systems differ from traditional search, see GEO vs SEO: what is the difference.

Step 1: Select Your Audit Prompts

The foundation of an AI visibility audit is the prompt list. These are the questions your buyers ask AI when researching your category. We call them golden prompts.

Aim for 20 to 30 prompts. They should cover three types of queries.

Recommendation queries. "What is the best CRM for mid-size B2B companies?" "Which project management tool should a 50-person team use?" These are the queries where AI names specific brands.

Comparison queries. "HubSpot vs Salesforce for small sales teams." "Monday vs Asana for marketing agencies." These queries force AI to evaluate your brand against competitors.

Problem-solving queries. "How do I improve sales pipeline visibility?" "What tools help with employee onboarding?" These queries may or may not name brands, but they reveal whether AI associates your brand with the problems your product solves.

Build your prompt list from three sources. First, ask your sales team what questions prospects ask in discovery calls. Second, check your search console data for long-tail queries that suggest comparison or evaluation intent. Third, look at Reddit threads and forum posts where users ask for recommendations in your category.

For a detailed guide on prompt selection methodology, see how to select prompts for LLM tracking.

Step 2: Run Baseline Queries Across Five Platforms

Test every prompt on five surfaces.

ChatGPT (with web search enabled). The highest-volume AI search platform with 883 million monthly users and 2 billion daily queries. ChatGPT citations turn over fastest, with a half-life of 3.4 weeks.

Perplexity. Cites sources in 97% of responses. The most citation-transparent platform. If your content is citable, Perplexity will show you.

Google Gemini. Growing from 5.7% to 21.5% market share in one year. Favors editorial and institutional sources over community content.

Google AI Overviews. Now appear in 25% of Google searches. These are not a separate platform but a layer on top of Google Search that affects whether users ever click through to your site.

Claude. Smaller market share (4.1%) but used by technical decision-makers and researchers. Worth including if your ICP is enterprise or developer-focused.

For each prompt on each platform, record three things. Is your brand mentioned? Is your website cited as a source? Is your brand recommended as a solution? Also record which competitors appear and in what position.

The whole process takes 3 to 4 hours if you do it manually with 25 prompts across 5 platforms. Automate it if you plan to repeat monthly.

Step 3: Score Your Visibility

Four metrics tell you where you stand.

Share of Model. What percentage of your golden prompts produce a response that mentions your brand? If you test 25 prompts and your brand appears in 8 responses, your Share of Model is 32%. This is your headline visibility number.

Citation Rate. When AI mentions your brand, how often does it cite your website as a source? High mentions with low citations means AI knows about you but does not trust your content enough to link to it. That is a content quality problem.

Recommendation Rate. When a user asks "what is the best X," does AI recommend your product? Being mentioned in a list is different from being the recommended solution. Recommendation Rate measures whether AI positions you as the answer.

Sentiment Score. How does AI describe your brand? Positive, neutral, or negative? Look at the specific language AI uses. "Industry-leading" is different from "a budget option." Sentiment affects whether a mention helps or hurts.

Score each metric per platform. Your ChatGPT visibility and your Perplexity visibility are independent measurements. Citation patterns drift differently across platforms, so aggregate scores hide important variation.

Step 4: Map Your Competitors

For every prompt where your brand does not appear, record who does. This competitive map reveals three things.

Who owns your category in AI. Some competitors may appear in 80% of responses while you appear in 10%. That gap is your visibility deficit.

Which competitors are investing in GEO. If a competitor appears frequently with cited sources and structured content, they are actively optimizing for AI. If they appear with uncited mentions, AI is relying on training data rather than current content.

Where the gaps are. Some prompts may return no strong brand recommendation. These are open opportunities where the first brand to create citation-ready content can claim the position.

Step 5: Identify Gaps and Prioritize Fixes

Your audit will surface a mix of issues. Prioritize by impact and effort.

Content gaps. Prompts where neither you nor competitors have strong content. These are the fastest wins because you are not displacing an entrenched competitor, you are filling a vacuum.

Citation gaps. Prompts where your brand is mentioned but not cited. This usually means your content exists but is not structured for extraction. Fix the passage structure on your key pages.

Recommendation gaps. Prompts where competitors get recommended and you do not. This is the hardest gap to close because it typically requires building authority through third-party validation, not just improving your own content.

Freshness gaps. Pages that were cited two months ago but have fallen out. Content older than 30 days loses citation probability by 40%. Set up a refresh cycle for your highest-value pages.

Technical gaps. Is PerplexityBot blocked in your robots.txt? Does your content render in static HTML? Do you have an llms.txt file? These are quick fixes with outsized impact.

Common Audit Findings

After running hundreds of AI visibility audits, the same patterns show up repeatedly.

Most brands are invisible on at least one platform. A brand visible on ChatGPT may be completely absent from Perplexity, or vice versa. Platform-specific gaps are the norm, not the exception.

Competitor visibility surprises. Smaller competitors with better-structured content frequently outperform larger brands. AI systems do not care about brand size. They care about content quality and citation-readiness.

Stale content is the most common fixable issue. The majority of citation gaps trace back to content that has not been updated in 60+ days. Freshness is the single highest-ROI fix in most audits.

Schema markup is almost always missing. FAQ schema produces a 350% citation increase, yet most sites do not have it. Adding structured data to your top 10 pages is usually a quick win.

Third-party validation is the hardest gap. Brands with no G2 reviews, no Reddit discussions, and no editorial mentions struggle to break through regardless of how well their own content is optimized. AI systems need independent confirmation before recommending a brand.

How Often to Audit

Run a full audit quarterly. Run a lighter monitoring pass monthly, focused on your top 10 golden prompts across all platforms. Citation domains churn 40 to 60% monthly, which means your visibility picture can change significantly in 30 days.

If you are actively making GEO improvements, increase monitoring to weekly for the first 8 weeks to track the impact of changes. After that, monthly monitoring plus quarterly full audits is sufficient for most brands.

FAQ

What is an AI visibility audit? An AI visibility audit measures how AI search platforms like ChatGPT, Perplexity, Gemini, and Google AI Overviews describe, cite, and recommend your brand. It tests a set of buyer-relevant prompts across multiple platforms and scores your visibility on metrics like Share of Model, Citation Rate, and Recommendation Rate.

How long does an AI visibility audit take? A manual audit with 25 prompts across 5 platforms takes 3 to 4 hours. A professional audit with competitive analysis, scoring, and a prioritized action plan typically takes 5 to 7 business days.

Is AI visibility the same as SEO? No. Only 12% of URLs cited by ChatGPT rank in Google's top 10. AI visibility and SEO measure different things, use different signals, and require different optimization strategies. Most brands need both.

How often should I run an AI visibility audit? Full audits quarterly, lighter monitoring monthly. Citation domains churn 40 to 60% per month, so your visibility picture changes fast. If you are actively making GEO changes, monitor weekly for the first 8 weeks.

What is the most common finding in AI visibility audits? Stale content. Most citation gaps trace back to pages that have not been updated in 60+ days. Refreshing your highest-value pages on a 30-day cycle is typically the highest-ROI fix.