Glossary
The AI search visibility glossary
The words we use to measure how AI assistants describe your brand - and what each one actually tells you.
AI search visibility is the measure of how often AI assistants name your brand when someone asks a question your business should be the answer to. Unlike a search ranking, an AI answer is assembled at the moment it is asked: a model retrieves passages from pages it trusts, weighs them against the phrasing of the question, and writes a short answer that names a handful of options. Ranking first in classic search does not guarantee a mention, and a mention does not guarantee you are described well - which is what our visibility audit measures.
That is why the vocabulary of this field matters. Terms like citation share, share of voice, prompt set and sentiment describe different parts of the same system, and each one is measured differently. A visibility score tells you how often you appear. Share of voice tells you how you compare. Citation share explains which pages are doing the persuading. Sentiment tells you whether the description helps you or hurts you. Track them together and a picture forms; track one alone and you are guessing.
The terms below are the ones we use in the aimentionz product and in our guides. They are written for marketing, content and SEO teams who need to brief work, read a report, and explain the results to someone else. Where a word is used differently elsewhere in the industry, we say so, because a shared definition is what makes a number worth arguing about.
New to the field? Start with AI visibility, prompt set and citation, then work through the guides in Resources. If a term you need is missing, tell us and we will add it.
Measurement
- AI visibility
- The share of tracked prompts in which an AI assistant names your brand in its answer. It is measured per model and per prompt set, then combined into a single 0–100 score so movement is easy to see.
- Why it matters: It is the closest thing the AI era has to a ranking: if you are not named, you are not in the consideration set.
- Prompt set
- The fixed list of questions run against each model on a schedule. A working set covers discovery, comparison and validation, and usually runs from ten to thirty prompts.
- Why it matters: Every other metric is a fraction of this list, so the questions you choose decide what the whole report can tell you.
- Share of voice
- Your brand mentions divided by all brand mentions across a set of prompts. It compares you with the competitors tracked alongside you rather than with your own past.
- Why it matters: It answers whether you are winning the category or merely present in it.
- Citation share
- The proportion of cited sources in AI answers that point at your own pages. It is reported next to the pages and domains that earn citations instead of you.
- Why it matters: It shows whether the models are quoting you or quoting other people about you.
- Non-branded visibility
- Visibility measured only on prompts that do not name your brand, such as 'best AI visibility tool for marketing teams'. Branded prompts are excluded from the calculation.
- Why it matters: It is the honest measure of demand: branded prompts flatter the score without telling you anything new.
- Sentiment
- How positively, neutrally or negatively a model describes your brand in its answer. Each reading is traced back to the exact phrases that produced it.
- Why it matters: A mention can be a warning. Sentiment catches reputational drift before it reaches a customer.
Models
- Model
- A specific AI assistant or underlying system, such as ChatGPT, Gemini, Claude or Perplexity. Each draws on different sources and returns different shortlists.
- Why it matters: Tracking one model gives you one opinion, not a market view - visibility has to be measured across several.
- Retrieval
- The step where an assistant gathers candidate passages from pages it trusts before writing an answer. Matching happens on the words in the question against the words on the page.
- Why it matters: You cannot be named in an answer if you were never retrieved, which is why plain, specific language outperforms clever headlines.
- Grounding
- Attaching an answer to specific sources so the model can point at where a claim came from. Grounded answers carry citations; ungrounded ones rely on training data alone.
- Why it matters: Grounded answers are the ones you can influence, because they are built from pages that can be written and improved.
Content
- Citation
- A source an AI assistant links to or leans on when answering a prompt. Citations are typically review sites, comparison pages, documentation and journalism rather than your homepage.
- Why it matters: Citations explain why an answer reads the way it does, which turns a visibility gap into a specific piece of work.
- Answer snapshot
- The stored copy of a model's full response to a prompt at the moment it was scanned, including the sources used. Snapshots are kept so changes can be compared later.
- Why it matters: It is the evidence behind every score - you can read exactly what was said, not just how the number moved.
Strategy
- Category prompt
- A question a buyer asks before they know which brands exist, such as 'what is the best way to track brand mentions in AI search'. It describes the job, not the product.
- Why it matters: Category prompts are where new demand is decided, so they deserve their own prompt set and their own share-of-voice reading.
- Visibility gap
- A prompt, topic or audience where competitors are named and your brand is not. Gaps are ranked by how often they occur so the biggest opportunity is worked first.
- Why it matters: It converts an abstract score into a ranked list of work with an obvious starting point.
Be part of the answer.
Put the vocabulary to work - start tracking how AI describes your brand across every major model.