AI visibility audit: a 5-layer method and scorecard for B2B SaaS

An AI visibility audit measures whether assistants like ChatGPT, Perplexity, Gemini and Google's AI Overviews name your brand for buyer questions, and explains why. It checks five layers: crawler access, product facts, answer-ready pages, third-party evidence and measurement. Below you'll get the process, a 24-point scorecard and the mistakes to avoid.

What an AI visibility audit covers

An AI visibility audit answers two questions. First, what do AI assistants say about your category, and are you in it? Second, why do you get the result you get?

Most guides stop at the first question. They give you a prompt list and a spreadsheet, which is useful, and we cover that method in our free AI visibility check. But a mention rate on its own doesn't tell you what to fix. A full audit also looks at the inputs assistants rely on: whether their crawlers can reach you, whether the facts about your product are clear and consistent, whether your pages answer buyer questions directly, and what third-party pages say about you.

For a B2B SaaS company, that second half is where the work is. The answer to "best SOC 2 automation tool for a 50-person startup" is usually assembled from review sites, comparison articles, community threads and a few vendor pages. Your audit has to cover all of them.

AI visibility audit vs a classic SEO audit

People also search for this as a "GEO audit" or an "AI SEO audit". The names differ, but the scope is the same. Here is how it compares with the SEO audit your team probably already runs.

The two overlap more than the table suggests. Google says there are no additional requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible to show with a snippet. So a broken SEO foundation also breaks your Google AI visibility. Run both, and share findings between them.

The five layers of an AI visibility audit

We split the audit into five layers. The first four are inputs you can change. The fifth is the outcome you measure. Working in this order stops you from rewriting content when the real problem is a firewall rule blocking a crawler.

1. Crawler access

Assistants can only cite pages their systems can fetch. Check your robots.txt, CDN and firewall rules for each AI crawler, not only Googlebot. OpenAI's documentation is explicit: sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers. It also separates OAI-SearchBot (search) from GPTBot (model training), so you can allow one and block the other. Our guide to AI crawlers lists the main user agents and a robots.txt template for SaaS sites.

Also check rendering. If your pricing table or feature list only appears after JavaScript runs, fetch the raw HTML and confirm the content is there.

2. Product facts

Assistants describe you using whatever they can find. If your homepage says "platform for revenue teams", your G2 profile says "sales engagement tool" and an old press release says "email automation", expect a muddled answer. Audit the core facts: category, who it is for, main use cases, pricing model, key integrations and notable limits. They should match across your site, review profiles, LinkedIn and directories.

3. Answer-ready pages

Buyers ask assistants specific questions: how much it costs, whether it integrates with Salesforce, how it compares with a named competitor. Check whether you have a page that answers each of those directly, near the top, in plain text. For SaaS, the pages that matter most are usually pricing, integrations, comparison or alternatives pages, security and compliance, and use-case pages. Schema helps machines parse some of this, but don't expect it to do the job alone (see schema markup for AI search).

4. Third-party evidence

This layer often decides the result. In Ahrefs' study of 75,000 brands, branded web mentions had the strongest correlation with AI Overview brand visibility (0.664), well ahead of referring domains (0.295) and backlinks (0.218). Correlation isn't causation, as Ahrefs notes, but it tells you where to look. List every third-party page that assistants cite for your prompts, then group them: review sites, editorial listicles, community threads, partner and integration directories, analyst and press coverage. Note which ones mention you, which mention only competitors, and which are out of date. More on this in AI brand mentions.

5. Measurement

Measure the outcome with a fixed set of buyer prompts. Two rules matter. First, run each prompt several times, because answers vary a lot. SparkToro's research found that ChatGPT and Google's AI had less than a 1 in 100 chance of returning the same list of brands across two responses to the same prompt. Second, report a mention rate (how often you appear), not a "rank in ChatGPT". SparkToro found ordering even less stable than the lists themselves.

For Google, add Search Console. Google has started rolling out generative AI reports that show impressions in AI Overviews and AI Mode by page, to a subset of sites so far. They show impressions, not clicks, so treat them as a complement to your prompt set, not a replacement.

How to run an AI visibility audit, step by step

This is the process we'd use for a B2B SaaS company with one core product. Budget one to two working days for a first pass.

1. Write 25–40 buyer prompts. Cover five groups: category ("best X for Y"), problem ("how do I fix Z"), comparison ("A vs B"), alternatives ("alternatives to A") and brand ("is [you] good for Y"). Write them as a buyer would, with company size, role or stack where relevant.
2. Pick your assistants. ChatGPT, Perplexity, Gemini and Google (AI Overviews and AI Mode) cover most B2B research. Add Claude or Copilot if your buyers use them.
3. Run each prompt at least three times in clean, logged-out or temporary sessions, across at least two days. Log every brand named and every URL cited.
4. Score the outcome. Mention rate for you and your top three competitors, plus accuracy: is what the assistant says about you correct?
5. Audit layers 1–4 with the scorecard below. Score each check 0 (missing), 1 (partial) or 2 (done).
6. Map gaps to causes. If you're missing from comparison prompts and you have no comparison pages and no listicle mentions, that is your answer. Most findings fall into one or two layers.
7. Rank fixes by effect and effort, pick the top five, and set a re-audit date 60–90 days out using the same prompts.

An AI visibility audit scorecard you can copy

Here are the checks we use for layers 1–4. Twelve checks, two points each, 24 points in total. The score isn't an industry benchmark. It is a way to see which layer is weakest and to compare your own audits over time.

A low Access score is the most urgent, because nothing above it can work. A low Evidence score is the most common for younger SaaS brands, and the slowest to fix, so start early.

How many people search for an AI visibility audit

Demand for this is growing into its own search category. We pulled US search volumes from Ahrefs Keywords Explorer on October 22, 2026.

"AI visibility audit" has 700 monthly US searches with a keyword difficulty of 4, and Google shows an AI Overview for it. The top results are mostly free report tools and short checklists. Few cover the technical, factual and third-party layers together, and none we read had B2B SaaS page types or a scoring method. That gap is why we wrote this guide.

Common audit mistakes

• Running each prompt once. A single run is a sample of one. You'll mistake noise for a finding.
• Reporting position. "We rank third in ChatGPT" is not a stable measure. Report mention rate and accuracy instead.
• Only auditing your own site. If assistants cite three listicles and a Reddit thread for your category, those pages are part of your audit.
• Blocking the wrong bot. Blocking GPTBot to opt out of training is a valid choice. Blocking OAI-SearchBot removes you from ChatGPT search answers. Check which one your rules hit.
• No baseline. Without a dated first audit, you can't show that any of the follow-up work paid off.

What to do after the audit

Turn the findings into a short plan: access fixes this week, fact fixes this month, page and evidence work over the next quarter. Then track progress with the same prompts each month. Our guides to AI share of voice and tracking AI traffic in GA4 cover how to report it, and LLM visibility covers what tends to move the number.

If you'd rather start with an outside view, our free audit runs this same process on your category and sends you the scorecard.

FAQ