I'm Gen Furukawa, founder of SuperMarketers. I've run this exact audit across dozens of B2B SaaS companies, and the result is the same almost every time: the founder assumes they show up in AI search, the audit shows they don't, and a competitor with a worse product is being recommended in their place. The audit takes 30 minutes. The gap it reveals has usually been costing pipeline for months.
Most articles that rank for "AI visibility audit" tell you to buy a platform. Profound, AthenaHQ, and Scrunch are good products, and you'll want one eventually. But you do not need any of them to get your baseline. You need 10 queries, four browser tabs, and 30 minutes. Here is the exact procedure.
What an AI visibility audit checks
The audit answers one question: when your buyers ask AI about your category, does your company come up? It checks three things in each answer.
Are you named? Not your category, not a generic description of the problem. Your company, by name, in the generated answer.
Who is named instead? The competitors the engine recommends are your real AI search competitors, which is often a different list than the one in your pitch deck.
How consistent is it across engines? A company can be cited in Perplexity and invisible in ChatGPT. One engine is a data point. Four engines is an audit. That difference is why this is built around answer engine optimization across every surface, not a single tool.
What you need: the four engines
Open four tabs. These are the surfaces where your buyers actually run their research, and each one weights sources differently, so you need all four.
The default for category and "best tools" queries. The largest audience and the one your buyers reach for first. Use a logged-out or temporary chat so memory does not personalize the answer.
The most citation-forward engine. It names sources inline, so it is the clearest read on which pages and companies it trusts for a given query.
Increasingly used for research and comparison questions. Tends to reward well-structured, clearly attributed sources over high-volume content.
Feeds Google AI Overviews, the answer block a growing share of your buyers see before any blue link. What Gemini cites is what Google increasingly surfaces.
Use a clean session in each. Personalized history can name a company the engine would not surface for a stranger, and you want the answer a new buyer sees, not the one tuned to you.
Step 1: Build a 10-query test bank
This is the part that determines whether the audit is useful. The 10 queries have to be the questions a buyer actually types, not the headlines you wish you ranked for. Spread them across four types so you measure your whole funnel, not just your brand.
- Three category queries."Best [category] tools for [use case]." "Top [category] software for [your ICP]." "What is the best [category] platform in 2026." This is where buyers start, and where being absent costs the most.
- Three problem queries."How to solve [the problem you solve]." "How do I [job your product does]." "Why is [problem] hard." Buyers who haven't named a category yet ask these first.
- Two comparison queries."[Top competitor] alternatives." "[Competitor A] vs [Competitor B]." If you aren't named as an alternative to the leader, you aren't in the consideration set.
- Two brand queries."What is [your company]." "Is [your company] any good." This checks whether the engine knows who you are and describes you accurately.
Write them down and keep them. These same 10 queries are the bank you re-score every month, which is the only way your numbers stay comparable over time. If you want a deeper version of this step, our AEO methodology uses a 20-query bank, but 10 is enough for a first baseline.
Founders write queries the way they'd write a value prop. A buyer does not search "AI-native revenue intelligence for mid-market." They search "best sales forecasting tool." Phrase every query the way a stranger would type it at the start of their research.
Step 2: Run each query in all four engines
Run all 10 queries in ChatGPT, Perplexity, Claude, and Gemini. That is 40 answers. Work down one engine at a time so you stay in a consistent session, then move to the next.
For each answer, read it the way a buyer would. Don't skim for your own name and stop. Read the whole recommendation: which companies lead, how the engine describes each one, and whether it gives a reason. The description matters as much as the mention, because that is what your buyer reads before they ever reach your site.
This is the step most people skip and the reason most "audits" are wrong. Checking one engine and assuming the rest match is how founders convince themselves they're visible when they're cited in one place out of four.
Step 3: Record who gets cited (you, competitors, no one)
Build a simple grid: 10 queries down the side, four engines across the top, 40 cells. In each cell, record one of three outcomes.
- 01You are namedThe engine recommends your company by name. Note whether the description is accurate, because an inaccurate mention is its own gap to fix.
- 02A competitor is namedWrite down which one. The competitors that recur across the grid are your real AI search rivals. These cells are where you are actively losing the recommendation.
- 03No one is namedThe engine answers generically without naming any company. This is open territory, and it is often the easiest citation to win because no incumbent holds it yet.
When the grid is full, the audit is essentially done. The pattern across 40 cells tells you more than any single answer: where you're invisible, who is taking the slot, and which queries are still unclaimed.
Step 4: Score yourself on the nine dimensions
Start with the headline number. Your citation rate is the count of answers where you appear divided by 40, times 100. If you were named in 6 of 40 answers, your citation rate is 15 percent. A score below 20 percent on your high-intent queries is a gap that is actively costing you pipeline.
Then go deeper. Citation rate tells you the score, not the why. The nine-dimension AI Visibility Score is the rubric we use to explain it, scoring the structural reasons an engine cites you or doesn't.
| Dimension | What it measures |
|---|---|
| Technical readiness | Can AI crawlers reach and parse your pages (robots.txt, llms.txt, render) |
| Content architecture | Definitions, steps, and FAQs an engine can lift cleanly |
| Entity clarity | One consistent, specific description of your company everywhere |
| Authority signals | Credible third-party mentions, not just your own site |
| Topical depth | Focused coverage of one category instead of scattered posts |
| Freshness | Pages updated recently enough to stay cited |
| Schema coverage | DefinedTerm, FAQPage, Article, and HowTo JSON-LD in place |
| Cross-engine consistency | Cited in more than one engine, not just one |
| Competitive position | Where you sit against the names that recur in your grid |
In the audits we run, most companies score 2 or 3 out of 10 on their first run. The target is 7 or higher. The point of scoring the dimensions is that it converts a vague "we're not showing up" into a specific list of structural fixes, and the structural fixes are what actually get you cited by ChatGPT.
Step 5: Find your three biggest gaps
You now have 40 cells and a score. Do not try to fix everything. Sort the gaps by impact and pick three.
The three biggest gaps are the high-intent queries where a competitor is cited and you are not. High-intent means a buyer running that query is close to a decision: category and comparison queries usually outrank generic problem queries here. A competitor being recommended on "best [category] tools" is worth more than no one being named on a vague how-to.
Those three queries are your first three pages to rebuild for extraction. Fixing the three highest-intent gaps moves more pipeline than fixing ten low-intent ones, because that is where buyers are actually choosing. Everything else goes on the list for later months.
A founder with finite runway who fixes three high-intent pages well beats one who half-fixes thirty. Three is small enough to actually ship this month and important enough to move your citation rate on the next audit. Depth on the queries that matter beats breadth across the ones that don't.
What to do next
You have a baseline, a citation rate, and three ranked gaps. Two paths forward.
Do it yourself. Rebuild those three pages for extraction: a direct definition in the first 100 words, numbered steps, five FAQ questions with self-contained answers, and JSON-LD schema. Then re-run this same 30-minute audit in 30 days and watch the cells flip. The method compounds because the same 10 queries make every month comparable to the last.
Have it run for you. If you'd rather not spend the 30 minutes, the free Displacement Report below runs the scan for you and shows the gap. If you want the full version, the Blueprint audit is the front end of our AI visibility system: it scores all nine dimensions across a 20-query bank, maps your real AI search competitors, and hands you a 90-day roadmap of which pages to build in what order. It costs $1,000 and credits to your first month if you continue.
But the manual audit is free, and you can start it in the next 30 minutes. Most founders won't, and that is exactly why their competitors are cited and they aren't. The ones who run it are the ones who close the gap.