Does AI know your business exists?
When a customer asks ChatGPT for the best in your category, do you even come up? See what 9 AI models actually know about your business — and who they recommend instead.
One business. Nine verdicts.
Harbor & Vine
Illustrative sampleFarm-to-table restaurant · Asheville, NC
3 of 9 models can't identify Harbor & Vine at all — and when asked for the best option without naming it, it's recommended just 34% of the time. Competitors get named more often.
Is your site set up for AI?
harborandvine.com
No JSON-LD structured data found on the homepage.
Fix: Add schema.org JSON-LD — at minimum a LocalBusiness block (name, address, URL, hours). Add FAQPage where relevant so AI can extract answers directly.
No /llms.txt found.
Fix: Add an llms.txt at your site root — a short Markdown overview AI systems read for context.
No /sitemap.xml found and none referenced in robots.txt.
Fix: Publish a sitemap.xml and reference it in robots.txt so crawlers find every page.
Only ~340 characters of text in the served HTML.
Fix: Much of your content is JavaScript-rendered — server-render the key content so AI crawlers see it.
No H1 on the homepage.
Fix: Use exactly one H1 and organize with H2/H3 headings phrased like the questions people ask.
GPTBot, ClaudeBot, PerplexityBot, Google-Extended and Bingbot are all allowed.
Both a clear title and meta description are present.
og:title and og:description present.
Make AI learn about you
The audit is the diagnosis. This is the treatment — paste-ready assets for your site and a prioritized checklist of where to establish your business so AI systems can find, understand, and cite you.
Harbor & Vine is a farm-to-table restaurant in Asheville, North Carolina, serving seasonal Appalachian cuisine sourced from local farms. It focuses on a rotating menu built around regional produce and a curated wine list.
The primary source models and AI Overviews use for local businesses — hours, category, reviews, location.
Do: Verify the listing and fill every field: category, hours, services, photos, description.
LLMs build entities from Wikidata — a structured record is one of the strongest signals that you exist.
Do: Add an item with your name, description, category, website and location.
Review sites are cited heavily and establish you as a real, active business in your category.
Do: Claim your Yelp profile and the top directories for your industry; keep name/address consistent.
Checking how much traffic AI sends you tells you almost nothing about whether AI knows you. Nearly every AI answer is zero-click — the model describes or recommends a business and the user never leaves the chat. So low AI traffic could mean the models know you well but nobody clicks, or that they've never heard of you. The only way to tell them apart is to measure what the models actually recall from training.
“How much traffic does AI send me?”
Measures clicks — the tiny, lossy tail of AI usage. A model can recommend you to thousands of people and show up as a flat zero in your analytics.
“Does the model actually know me?”
Measures training recall — whether you exist to the AI at all, whether it recommends you unprompted, and whether what it believes about you is even true.
Four things, across nine models
Not a vibe check. A structured audit that turns fuzzy AI anxiety into a scoreboard you can act on.
Recognition
We ask each model what it knows about your exact business — from training only, no web search. It answers clearly, vaguely, or not at all, and tells us if it's confusing you with someone else.
Unprompted recommendation
The real test. We ask each model for the best option in your category and city — without ever naming you — across many phrasings a real customer would use, and measure how often you come up on your own.
Accuracy
Of the facts a model claims about you, how many are true? We crawl your own site for ground truth and flag what the models get wrong — the wrong city, a defunct product, a founder who left.
Share of voice
When the models recommend businesses in your category, who do they name — and how often is it you versus your competitors? Your slice of the AI conversation, quantified.
“AI” isn't one thing
Each model trained on different data at a different time. One knows you cold; the next has never heard of you. Checking only ChatGPT is a blind spot the size of every other model your customers use.
ChatGPT
Claude
Gemini
Grok
DeepSeek
Llama
Mistral
Perplexity
CohereFind out where you're invisible.
Run your business through all 9 models and get the full scoreboard — recognition, unprompted recommendation rate, accuracy, and who the AIs name instead of you.
FAQ
How do I know if AI knows about my business?
The only reliable way is to ask the models directly, from their training knowledge, with search turned off. If a model can describe your business without looking it up, it learned about you during training. AI Colosseum's visibility audit does this across 9 models at once — ChatGPT, Claude, Gemini, Grok, Perplexity, DeepSeek, Llama, Mistral and Cohere — and reports which ones recognize you, how accurately, and whether they recommend you unprompted.
Why doesn't AI referral traffic tell me if AI knows my business?
Referral traffic only counts the rare cases where someone clicks a link inside an AI answer. Most AI answers are zero-click — the model describes or recommends you and the user never leaves the chat. So low AI referral traffic can mean the models know you well but nobody clicks, or that they don't know you at all. Measuring what the models actually recall from training separates the two.
What is an AI visibility audit?
An AI visibility audit measures what large language models genuinely know about your business from their training data, rather than what they can retrieve live. A good audit checks four things across multiple models: recognition (do they know you exist), unprompted recommendation (do they name you when asked for the best option in your category without being told your name), accuracy (is what they believe about you correct), and share of voice (who they recommend instead).
Does ChatGPT recommend my business to customers?
It depends whether ChatGPT learned about you during training and files you under the right category and location. The test is simple: ask 'what's the best [your category] in [your city]?' without naming your business, several times. If ChatGPT names you some of the time, it recommends you unprompted. The audit runs this across 9 models with varied phrasings so you get a recommendation rate, not a single lucky answer.
Why check 9 AI models instead of just ChatGPT?
Because 'AI' isn't one thing. Each model was trained on different data at different times, so one may know your business clearly while another has never heard of it. Checking only ChatGPT tells you nothing about the customer using Gemini, Grok or Perplexity. Seeing all 9 side by side shows exactly where you're invisible and where you already win.
How much does the AI visibility audit cost?
It runs on AI Colosseum credits. A full 9-model audit uses roughly 26–128 credits depending on depth (how many recommendation runs per model), and the exact cost is shown before you run it. New accounts start with 100 free credits and no credit card, which is enough to audit your business.