Every definition below carries the frequency we actually measured, so the term is not just explained but priced. The full method and the sample sit in 100 GEO audits: the report.
glossary009 Agency
GEO glossary
the eight terms that decide whether an AI engine can read you, verify you and quote you — defined with the numbers from our own 100 audits, not from a dictionary.
Generative Engine Optimization (GEO)
What it is. Optimization for AI engines that generate answers rather than lists of links: ChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek.
Why it matters. SEO competes for a position on a results page. GEO competes for a mention inside a sentence the model writes when a buyer asks who to pick. Across the 100 audits in our published dataset the two showed no correlation: sites with strong organic traffic and zero AI presence were common, and so was the reverse.
Three layers, in order: readability (the crawler can fetch your content), verifiability (the model can confirm you are a specific real company), citability (there is something on the page worth quoting).
Answer Engine Optimization (AEO)
What it is. In practice a synonym for GEO, used more often when the emphasis is on direct answers and featured snippets rather than on generated prose.
Why it matters. The two terms are used interchangeably by most practitioners, including us. We target both because buyers search with both, and because no one has yet won the naming argument.
If a distinction is drawn at all, AEO leans toward answer boxes and voice results, GEO toward the text an LLM composes.
llms.txt
What it is. A plain-text file at the root of a domain that tells AI systems what the site is, what its key pages are, and how to describe the company.
Why it matters. It is cheap insurance rather than a ranking lever: an hour of work, no downside, and the only place where you get to state in your own words what your company is and which similarly-named things it is not. In our dataset 87% of audited sites had no llms.txt at all, and where one existed it was usually a link dump of the blog.
The part that has changed answers for us is the disambiguation paragraph — naming the collisions explicitly.
AI crawler
What it is. The user-agents AI companies send to fetch web pages: GPTBot and OAI-SearchBot (OpenAI), ClaudeBot (Anthropic), PerplexityBot, Google-Extended, CCBot and others.
Why it matters. They are not browsers. Several of them execute no JavaScript, so a site that renders content client-side can return a valid 200 response containing nothing a model can read. One site in our report scored 100/100 for SEO in Lighthouse while GPTBot received 1,133 bytes of empty shell from the same URL.
Check it in one command: request your page with the crawler's user-agent and read what comes back.
brand recall
What it is. Whether an engine knows your brand by name and describes it accurately when asked about it directly.
Why it matters. This is the easy half, and it lulls people. Across 100 audits, 93% of brands were known to AI and described accurately. Being known is not being recommended.
Measured with branded prompts: “what is [brand]”, “is [brand] legit”, “tell me about [brand]”.
category recall
What it is. Whether an engine names you when a buyer asks who to pick in your category, without naming you first.
Why it matters. This is the half that converts, and it is where the market fails: only 4.5% of category answers named the audited brand. The gap between 93% and 4.5% is the entire commercial argument for GEO.
Measured with category prompts: “best [category] for [use case]”, “who should I hire for [job]”.
entity anchor
What it is. A record outside your own website that lets a model confirm you are a specific, real organization — a company page, a directory record, a knowledge-base entry, consistent identifiers repeated across all of them.
Why it matters. Close to 100% of audited sites had none. Without an anchor you are a string of characters rather than an entity, and a model choosing whom to quote has no way to tell you apart from something else sharing your name.
Consistency matters more than quantity: three profiles that agree beat ten that contradict each other.
citability
What it is. Whether a page contains something a model can lift: a definition in one paragraph, a comparison, a price, a dated figure with a source, a named author.
Why it matters. Models quote what is easy to quote. Anonymous content with no dates and no sources was present in 89% of our audits, and it is the profile a model discounts when it has to pick a source.
Answer-shaped beats comprehensive. A 60-word definition gets quoted; a 2,000-word essay around the same idea often does not.
the terms are the map. the audit tells you which of them is costing you money right now.