AI visibility glossary
Every number and label you’ll meet when tracking whether AI assistants recommend your brand, defined in plain English. No prior SEO knowledge assumed.
Mention rate
Out of every AI answer we collected, the share that named your brand. 40% means you showed up in 4 of every 10 answers.
The headline number, and the one worth watching over time. If we collected 100 AI answers to your tracked questions this month and your brand was named in 34 of them, your mention rate is 34%. Answers where AI recommended other brands but not yours still count in the total — those are real misses, and hiding them would flatter the number into uselessness. Mention rate is a share, not a rank, so it does not compete with your rivals' numbers: several brands can each be named in most answers, and in a healthy category they often are.
Strongest engine
The AI assistant that names your brand most often — where you're already winning.
The AI assistant that names your brand most often. Worth knowing because ChatGPT and Gemini disagree more than people expect — they search differently and read different pages, so the same question can produce different brand lists. If one engine consistently names you and the other never does, that is a signal about which pages each one trusts, not about your product. Treat the weaker engine as the opportunity: something it reads in your category is not discussing you.
Questions tracked
How many shopper questions we ask the AI assistants on your behalf, every day.
How many shopper questions we put to the AI assistants on your behalf, every day. Each one is asked to both engines, so ten tracked questions means twenty answers collected daily. This is also the number your plan caps, because it is what drives the cost of running your account. Fewer, better-chosen questions beat more vague ones: a question a real customer would type produces an answer a real customer would see.
AI engine
An AI assistant shoppers ask for recommendations — today ChatGPT and Gemini. We put your questions to each one daily.
An AI assistant that shoppers ask for recommendations — today ChatGPT and Gemini. We ask each one your questions separately and record its answers separately, because they are genuinely different systems that reach different conclusions. Both are queried with live web search switched on, so what you see reflects what a shopper asking today would be told, rather than what the model absorbed during training.
By engine
The same mention rates split by assistant, so you can see when ChatGPT and Gemini disagree about your category.
The same mention rates, split per assistant instead of pooled. Useful precisely when the two disagree: a pooled 30% could be 55% on ChatGPT and 5% on Gemini, which is a completely different situation from a steady 30% on both. The first says one engine has a blind spot about you and is worth investigating; the second says you are moderately present everywhere.
Question by question
Each question, and the brands AI named in its answers. The chip shows a brand's average spot in the answer and how often it appeared.
Each tracked question, and which brands AI named when answering it. This is where a single mention rate becomes actionable, because the average always hides a split: you might be named in nearly every answer to "is [your brand] legit" and almost never in answers to "best serum for oily skin". Those are different problems. The first is reputation, the second is discovery — and the second is where the customers you have never met are.
Tracking prompt
A shopper question we ask the AI assistants every day on your behalf, like "best vitamin C serum".
A shopper question we ask the AI assistants every day on your behalf, like "best vitamin C serum". Prompts are the whole input to the system, so choosing them well matters more than any setting. Write them the way a customer would type them, in their words rather than yours, and favour the questions asked by people who do not yet know your brand exists — those are the answers you are absent from and the ones worth winning.
Sentiment
How the AI talked about you in this answer — positive, neutral or negative. Being named isn't always being praised.
How the AI talked about your brand in a given answer — positive, neutral or negative. Being named is not the same as being recommended: an assistant can list you as the budget option, mention you as the one people complain about, or name you as the top pick, and all three count as a mention. Sentiment is what separates them. Most mentions land neutral, which is normal; a run of negative ones is worth reading in full rather than treating as a number.
Ranked #N
Where your brand came in this answer's list. #1 means AI named you before every other brand.
Where your brand came in one answer's list of recommendations. "#1" means AI named you before any other brand. This matters because position within an answer behaves like position on a search results page — the first recommendation gets disproportionate attention, and the fourth is often never read. A brand named consistently but always last has a different problem from one named rarely but always first.
Clippings
The answers AI actually gave, word for word. We keep them so you can read what shoppers are being told.
The AI answers themselves, saved word for word. Numbers tell you whether something changed; the clipping tells you what was actually said about you, in the sentence a shopper actually read. Keeping the full text also means a claim can be checked rather than trusted — if a mention rate moves, the answers behind the move are there to read.
Source
A web page an AI read before writing its answer. The pages AI trusts are the pages that decide who gets recommended.
A web page an AI read before writing its answer. Assistants do not answer shopping questions from memory: they search, open a handful of pages, and summarise them. Those pages are the sources, and they are the real decision-makers — the brands named in an answer are overwhelmingly the brands named on the pages behind it. Which is why the source list is more actionable than the answer itself.
Citation gap
A page AI cites when it recommends a competitor, but has never cited for you. Each one is a place you could earn a mention.
A page AI cites when it recommends a competitor, but has never cited for you. Each one is a specific, addressable place to go earn a mention — a magazine roundup you could pitch, a community thread you could join, a comparison page you could answer with a better one of your own. One honest limit: a gap means AI has never used that page in your favour. It does not mean your brand is absent from the page, because we record which pages engines cite and who for, and never read the page's own text.
Sources citing you
How many different websites AI has used at least once while recommending your brand.
How many different websites AI has used at least once while recommending your brand — your supporter list. Growth here is the leading indicator for mention rate: more independent pages discussing you means more chances that whatever an assistant happens to read this time is a page that names you. A single new frequently-cited source can move the headline number weeks later.
Site type
Where a page lives: your own site, a competitor's site, or independent press you could pitch.
Where a cited page lives: your own site, a competitor's site, or independent press. The distinction is about what you can do next. Independent press and community pages can be pitched or joined. A competitor's own page cannot be — but it tells you why AI recommends them, which is usually that they built a page answering exactly the question you are losing. Your own pages appearing here is a good sign: it means assistants read your site directly.
Gaps
Answers where AI recommended brands in your category but never named yours — the conversations you're missing.
Answers where AI recommended brands in your category but never named yours — the conversations happening without you. These are the most useful failures in the product, because they are specific: a real question, a real answer, and a list of the brands that got named instead. Unlike a low mention rate, which tells you something is wrong, an answer gap shows you exactly what was said and which pages said it.
For the ideas behind these terms rather than the definitions, start with how AI decides which brands to recommend.