How often, and how favourably, AI assistants mention your company when people ask them questions your product answers. Usually measured as the share of answers that name you across a fixed set of buyer questions.
Generative engine optimization (GEO)
The work of making a brand more likely to be named and cited in AI-generated answers. The term comes from a 2023 research paper of the same name. See What is GEO?
Answer engine optimization (AEO)
An older term for optimizing content to be the direct answer to a question, first in featured snippets and voice assistants, now used much like GEO.
LLM optimization (LLMO)
Another name for GEO, used when the focus is large language models specifically. The three terms describe overlapping work.
AI Overviews
Google’s AI-generated summaries at the top of some search results. Google says they follow the same controls as Search: Googlebot in robots.txt, plus noindex and nosnippet.
AI Mode
Google’s conversational search mode, which answers a question with a generated response and links. Counted with other Search traffic in Search Console.
Zero-click search
A search that ends without a click to any website because the answer was on the results page or in the AI answer itself. A recommendation can influence a purchase without ever producing a visit.
What gets measured
Mention
An AI answer naming your brand. Mentioned counts a mention only when the name lands on a word boundary, so “Done” isn’t found inside “abandoned”.
Citation
An AI answer linking to a page as a source. Only assistants that search the web while answering, such as Perplexity or ChatGPT with search, produce citations. A brand can be mentioned without being cited, and cited without being recommended.
Visibility (metric)
In Mentioned, the share of answers that named your brand: named in 7 of 10 answers is 70%.
Share of voice
Your brand’s share of all brand mentions in a set of answers. If answers named brands 40 times and 10 of those were you, your share of voice is 25%.
Average position
Where your brand appears when an answer gives a ranked list. Lower is better. Mentioned leaves it blank, not zero, when you were named only in prose.
Sentiment
Whether the sentence that names you is positive, neutral or negative. Mentioned reads cue words in that sentence and shows them, rather than scoring the whole answer.
Buyer question
A question a buyer would type before they know your name, such as “what’s the best CRM for a 20-person sales team?”. It leaves your brand out, so it measures a recommendation rather than recall. See buyer questions.
Prompt (tracked prompt)
One question an AI visibility tool asks on a schedule. Tools count prompts differently, so compare questions asked per month, not the label.
Locked questions
Questions written once and asked word for word on every run, so a change in the trend is a change in the answers, not in the questions. Mentioned locks each tracked keyword’s five questions when you create it.
Blind spot
A model or a question where you’re never named. In Mentioned’s Actions, a model blind spot is an assistant that never names you while another does; a prompt blind spot is a question that never names you while your other questions do.
AI referral traffic
Visits that arrive from an AI assistant, such as chatgpt.com or perplexity.ai. Some arrive with no referrer and show as Direct. See tracking it in GA4.
How assistants answer
Large language model (LLM)
The model behind an assistant, such as GPT, Claude or Gemini. It writes answers from patterns learned in training, and sometimes from pages it fetches while answering.
Training data
The text a model learned from. What it says about your category without searching comes from here, which is why it can lag your latest launch by months.
Knowledge cutoff
The date after which a model has no training data. Anything newer reaches it only if it searches the web.
Grounding
Giving a model fresh, specific sources at the moment it answers, such as search results or a fetched page. Google uses the term for Gemini answers built on Google Search.
Retrieval-augmented generation (RAG)
The general technique behind grounding: retrieve relevant documents, then have the model answer from them. Assistants that search the web are doing a form of it.
Hallucination
A confident statement that isn’t true, such as a wrong price or a feature you don’t have. Assistants hallucinate most about brands they know little about.
Entity
A thing a search engine or model treats as one distinct object, such as your company, with a name, a website and profiles elsewhere. Consistent naming across the web helps tie mentions to the right entity.
Crawlers and files
AI search crawler
A bot that indexes pages so an assistant can find them while answering: OAI-SearchBot for ChatGPT, Claude-SearchBot for Claude, PerplexityBot for Perplexity. Blocking one removes you from that assistant’s search results.
AI training crawler
A bot that collects pages that may be used to train future models, such as GPTBot, ClaudeBot and CCBot. Blocking one doesn’t remove you from live search answers. See which crawlers to allow.
User-triggered fetcher
A bot that opens a page because a person in a conversation asked, such as ChatGPT-User or Perplexity-User. Some operators say these may not follow robots.txt.
Google-Extended
A robots.txt token, not a crawler, that controls whether Gemini may train on or ground answers in pages Googlebot fetched. Google says it doesn’t affect Search.
robots.txt
A text file at the root of a site that tells crawlers which paths they may fetch, read according to RFC 9309. A request, not a lock. Check yours with the AI crawler checker.
llms.txt
A markdown file at the root of a site with a summary and links to its key pages, written for language models. Proposed at llmstxt.org in 2024. See what it does and doesn’t do.
llms-full.txt
A companion to llms.txt, not part of its spec, holding the full text of a site’s key pages in one file.
Structured data (schema markup)
Machine-readable facts in a page, usually JSON-LD using the schema.org vocabulary, such as your organization’s name and logo or a product’s price. Check a page with the structured data checker.
Server-side rendering
Sending a page’s content in the HTML itself rather than building it with JavaScript in the browser. Matters because the major AI crawlers fetch JavaScript files without running them.
See the terms in practice
Visibility, share of voice and blind spots are easier to understand on your own brand. The free check asks Claude and Gemini five buyer questions for your category and shows which answers named you, the exact sentence, and who was named instead.
