Your buyer asked ChatGPT for the three best tools in your category this morning. You weren't one of them. You'll never see that in Google Analytics, because there was no click to record.
That gap is what AI visibility measures. It's the share of relevant AI answers that mention, recommend or cite your brand across ChatGPT, Gemini, Perplexity, Claude, Copilot and Google's AI Overviews. If classic SEO asks "where do we rank?", AI visibility asks "are we in the answer at all, and what does it say about us?" This guide gives you a working definition, the five numbers behind it and a two-week plan to set your baseline. If you already track LLM visibility, use it as a checklist to see what you're missing.

A working definition of AI visibility
AI visibility is how often, how prominently and how accurately AI assistants present your brand when people ask questions you should own.
Three parts matter in that sentence.
How often
How prominently
How accurately
Frequency is the obvious one. Prominence is first-mention versus a footnote in a list of eight. Accuracy is whether the model quotes your current pricing or a 2023 review that says you lack SSO.
A brand can score high on one and fail the other two. Being mentioned in every answer as "the expensive option" is visibility, but not the kind you want.
Why "AI visibility" replaced half a dozen other terms
Over the past two years the industry tried GEO, AEO, LLM SEO, AI search optimization and answer engine optimization. Those names describe the work. AI visibility describes the outcome, which is why it stuck with marketing leaders and CFOs.
You don't report "GEO" to a board. You report whether the brand shows up when buyers ask. That's also why most SEO suites now bolt on an "AI visibility" tab. The term became the budget line.
The five numbers behind an AI visibility score
A single score is useful for a dashboard. It's useless for deciding what to fix. Break it into these five metrics instead.
1. Inclusion rate
The percentage of your tracked prompts where your brand appears at all. This is the headline number. Track it per engine, because ChatGPT and Perplexity often disagree by 20 points or more on the same prompt set.
2. Position and prominence
Where in the answer you appear. First-named brands get chosen far more often than the fifth name in a bulleted list. Log whether you're the primary recommendation, one of several or a passing mention.
3. Share of voice
Your mentions divided by total brand mentions across the same prompts. This is the competitive view. You might sit at 40% inclusion and still lose, if your main rival sits at 75%.
4. Citation share
How often the engine links to your own domain as a source, versus a review site, Reddit thread or competitor's comparison page. Mentions build awareness. Citations send the occasional click and, more importantly, show the model trusts your page as the source of truth.
5. Sentiment and accuracy
What the answer actually says. Is your pricing current? Are your features right? Are you framed as "best for enterprise" when you sell to startups? Wrong facts cost deals quietly, because the buyer never tells you why they didn't book a demo.
What actually drives AI visibility
Models build answers from two sources. One is what they learned in training. The other is what they retrieve live from the web when a question needs fresh facts. You influence both, with different levers.
For retrieval, the basics come first.
Let AI crawlers in
Put facts in HTML
One fact, one page
Check robots.txt and your CDN for blocks on GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot and Google-Extended. Then move pricing, plan limits and integrations out of images, PDFs and JavaScript tabs into plain text.
For training data and third-party consensus, the work is slower. Models lean heavily on what others say about you. G2 and Capterra profiles, "best X for Y" roundups, Reddit threads, analyst lists and Wikipedia all feed the picture. If ten sources call you "a budget tool for freelancers", one homepage headline won't change the answer.
Entity clarity ties it together. Consistent naming, Organization and Product schema, and a clear "what we do, for whom" statement on your About page help models connect mentions back to the right brand. Teams with a generic or shared brand name feel this most.
How to measure your AI visibility in two weeks
You don't need a six-month project. You need a fixed prompt set and a habit.
Week one is about building the baseline.
Write 30 to 50 prompts your buyers actually ask. Mix category questions ("best CRM for a 20-person agency"), comparison questions ("HubSpot vs Pipedrive for agencies") and fact questions ("does X integrate with Slack"). Pull them from sales call notes and support tickets, not from your keyword tool.
Run each prompt across at least four engines. Record inclusion, position, competitors named, sources cited and any wrong facts. Repeat each prompt two or three times, because answers vary run to run.
Week two is about finding the gaps.
Sort prompts by where competitors appear and you don't. Those are your priority pages. Then list every wrong fact and trace it to a source. Most errors come from your own hidden content, an outdated third-party page or a competitor's comparison post.
Doing this by hand works once. It doesn't scale to weekly tracking across five engines. A dedicated LLM brand visibility tool like LLM Search Console runs the same prompts on a schedule, tracks share of voice against named competitors and flags when a model starts repeating something wrong about you.
Common mistakes when teams start tracking
A few patterns show up again and again.
Teams track only branded prompts. Of course ChatGPT mentions you when someone types your name. The prompts that matter are the unbranded ones where buyers haven't decided yet.
Teams treat one run as truth. AI answers are probabilistic. A single screenshot proves nothing. Trends across repeated runs do.
Teams chase traffic. AI referral traffic is real and growing, but most AI-influenced research ends without a click. Inclusion and share of voice tell you more than sessions.
Teams fix the homepage and stop. Models pull from pricing pages, docs, comparison pages and third-party listings far more than from a hero headline.
Your next step
AI visibility isn't a new channel to add to the list. It's a measure of whether the channel your buyers already use to research ever puts you in front of them.
Start small. Write ten prompts a buyer in your category would ask today. Run them in ChatGPT, Perplexity and Gemini. Count how many times you appear and who appears instead. That one exercise will tell you more about your AI visibility than any traffic report.
Want a practical breakdown like this every week? Subscribe to the LLM Search Console newsletter on Substack for hands-on playbooks on AI visibility, competitor tracking and getting cited by the engines your buyers trust.


The no-click part is what makes this hard to sell internally. A buyer gets three names from ChatGPT, picks one, and nothing in the analytics ever shows the shortlist you weren't on.
The accuracy part is the one founders underestimate most. Pricing is the usual miss in what I see: an old plan, a discontinued feature, a stale founder bio. You can't edit the answer, so the fix is upstream. Find the old page or directory listing it's pulling from, publish one clear page with current pricing and facts, make every profile agree, then re-run the prompt weekly, logged out, and score it for accuracy, not just presence.