What GEO, AEO, cited sources, llms.txt and the rest mean — in plain words, with what each one means for a founder doing this without a marketing team.
Making a product easy for AI assistants such as ChatGPT, Claude, Perplexity, Gemini and Grok to find, understand and recommend when people ask them which product to use.
It covers your own site (can assistants read it, does it say plainly what you do, who it is for and what it costs) and the pages assistants lean on when they answer (lists, directories, reviews, forums). GEO and AEO are other names for the same work.
Another name for AI search optimisation: improving how a brand appears in answers written by generative AI, rather than in a list of search results.
A search result is a link someone may click; a generated answer names a few products and explains why. GEO is about being one of the names, described accurately.
Another name for AI search optimisation, from the idea that assistants are answer engines: they reply with an answer rather than a page of links.
The term is older than chatbots — it started with featured snippets and voice assistants — and now mostly means the same work as GEO.
How often, and how, AI assistants name your brand when your customers ask them about your category.
Measured from real answers to real questions, not estimated. Being named is half of it; the other half is whether the description is accurate and favourable. A brand nobody has asked about yet has no visibility number at all, rather than a flattering one.
An AI assistant people ask instead of searching — BrandReco measures ChatGPT, Claude, Perplexity, Gemini and Grok.
Each one answers the same question differently, cites different pages and changes over time, which is why a question has to be asked on each of them, repeatedly, to know where you stand.
A customer question asked to the AI engines on a repeating schedule — for example “best booking app for a small salon” — to see who they recommend.
Good tracked questions are the ones your customers really ask before choosing. Each answer is kept in full with the sources it cited, so a change can be traced to the day it happened.
An answer an AI assistant writes after searching the web, built from the pages it found and usually citing them.
Most “which product should I use” questions get grounded answers, which is why the pages an assistant finds — not only what it learnt in training — decide who it names.
A page an AI answer was built from and pointed to as its basis.
If the pages an assistant cites for your category never mention you, it cannot recommend you from them. Getting onto the cited pages you can get onto — a list, a directory, a thread — is one of the fastest ways to start being named.
Your brand named in an AI answer or on a web page, with or without a link.
A citation is a page the answer used; a mention is a name the answer said. Assistants learn a brand from what independent sites say about it, so mentions elsewhere on the web feed the mentions in answers.
An article ranking the products in a category — “9 best AI SEO tools for founders” — and one of the pages AI assistants cite most when asked which product to pick.
Many are published by a vendor in that category and rank its own product first. Those are its marketing, and asking to be added rarely works; independent lists and directories are the ones worth a note.
Where a question stands across the engines: open (nobody is consistently named), contested (several brands are), owned (one competitor is named almost every time) or yours.
An open question is the cheapest one to win: the first brand with a clear page answering it often becomes the answer.
A question you could realistically become the answer to within weeks, because no brand dominates it and the answers lean on pages you can get onto.
A hard question is one a competitor is named for on most engines, built from its own pages and major publications. Working on winnable questions first is how a small brand gets named at all.
A site AI assistants can read, and that says clearly what the business does and who it is for. BrandReco checks both every week and tells you in plain words what is missing.
It comes from your own site, so it is the part you fix on your own: each fix that goes live moves you closer, and the AI-ready badge unlocks once your site gets there.
A plain-text file at the root of a site (/llms.txt) that tells AI assistants, in a few lines, what the site is and which pages matter.
It is a proposed convention, not a standard every assistant reads, but it costs minutes to publish and gives any assistant that does read it the facts in your words.
The programs AI companies send to read web pages — GPTBot, ClaudeBot, PerplexityBot and others — and the robots.txt file that tells them which pages they may read.
A robots.txt that blocks them, often by accident or by a hosting default, keeps your pages out of the answers. Some bots read for live answers and others for training; you can allow one kind and not the other.
Facts about a page written in a fixed format (usually JSON-LD, using the schema.org vocabulary) so machines can read them without guessing: who the company is, what the product costs, the questions a page answers.
Common types are Organization (the company), Product or SoftwareApplication with offers (what it costs) and FAQPage (questions and answers). It only helps when it matches what the page visibly says.
The part of Organization structured data that lists a company's other official profiles — LinkedIn, Crunchbase, X, GitHub — so machines can tell they are all the same company.
It is how an assistant connects your site to your profiles and stops confusing you with another company of a similar name.
A thing an AI assistant can tell apart from everything else — your company, with one name, one site and one description that agree wherever they appear.
When your site, your profiles and the directories describe you differently, the assistant is less sure who you are and names you less.
One thing you state is true about your business — what it is in a line, who it is for, what it costs — that AI answers are checked against.
Wrong facts in AI answers (an old price, the wrong audience, a feature you don't have) cost customers, so each fact an answer gets wrong becomes something to fix.
One small, specific change — on your site or somewhere AI assistants read — that is checked after you mark it done, by re-reading the page or file it was about.
Done means the change is live and was seen, not that a box was ticked. Some actions, like a reply in a community, can't be checked from outside; those say so.
An open standard that lets an AI assistant connect to another service and use it — read its data and take actions — on your behalf.
BrandReco's MCP server lets Claude, ChatGPT, Claude Code or Cursor read your actions, do the ones they can, and ask BrandReco to verify them.
See which of these is holding your product back.
Run the free checkWe use cookies that are needed to sign you in, and — only if you agree — self-hosted analytics to count page views. No advertising or third-party trackers. Cookie Policy.