GEO, or generative engine optimisation, is the work of making your business likely to be named when someone asks ChatGPT, Gemini or Claude for a recommendation. It is the assistant-era cousin of SEO. Instead of only chasing Google’s blue links, you are trying to be the firm the model mentions when a customer asks “who should I use”.
More people now ask an assistant before they search, so this matters. The good news is that much of it is clear writing, structured data, and not blocking the crawlers. The less good news is that you cannot pay ChatGPT to recommend you, and you cannot force a mention on a deadline.
What is GEO, and how is it different from SEO?
SEO tries to rank a page in search results. GEO tries to get you quoted or recommended inside an answer. The customer sees a short list of names, not ten blue links. If you are on that list, you get a chance at the job. If you are not, you may never know you were considered.
The underlying ingredients overlap. Assistants draw on pages they can read, listings they can trust, and reviews that suggest you are real. A site that already states what you do, where you do it and how to get in touch is halfway there. A site full of fluff, blocked bots and no structured data is easy to skip.
They are not the same job. A page can rank on Google and still be a poor source for an assistant if the facts are buried, the markup is missing, or GPTBot is banned in robots.txt. A page can be easy to quote and still sit on page two of Google. Doing both is the point of our AI optimisation and GEO work, from £99 as a one-off.
GEO is early. That is a reason to do the basics now, not a reason to buy a “ChatGPT ranking” package from someone who cannot explain GPTBot. There are no guarantees. Anyone selling a guaranteed mention is inventing a product.
How would this work for an electrician in Hassocks?
Take a real kind of query: “Who is a good electrician in Hassocks?” or “Recommend an electrician near Hurstpierpoint”.
The assistant does not have a favourite spark. It looks for sources it can fetch or has already seen. A useful source says, in the first screen, that the firm is an electrician, that it covers Hassocks and the villages around it, that it is registered with a scheme such as NICEIC or NAPIT, and how to phone. It repeats those facts in FAQ pairs (“Do you cover Ditchling?”, “Are you registered?”). It marks the page up with LocalBusiness and FAQ schema so a machine does not have to guess. It publishes a short llms.txt file at the root with the same essentials. It does not block GPTBot.
BW Electrical is a real Hassocks electrician we have built for. I will not claim ChatGPT currently names them, or that anyone else on this list is “the” AI pick. That would be an invented result. What I will say is the shape of the page they need is the shape assistants can use: trade, town, qualifications, a number, no waffle.
If the same customer asked for “the best electrician in Sussex”, the model has a worse job. Sussex is a county. One page will not stand for a whole county in search, and it will not stand for a whole county in an answer either. The model will hedge, name directories, or pick city firms it has more text about. That is another reason to be specific. “Electrician in Hassocks” is a question you can deserve. “Electrician for all of Sussex” is a question almost nobody deserves.
The same shape applies to Oilfire in Uckfield, Ruby’s Bakery in Burgess Hill, Sindall in Haywards Heath, Phoenix Piling, The Vests or C4C Group. Change the trade and the town. Keep the facts on the page, in the schema, and in a file a crawler can read in one go.
If you want a technical read on whether assistants can see the site at all, run the AI visibility checker. It will not conjure a ranking. It will show you whether the facts are even there to quote.
Which AI crawlers need to be allowed in?
Assistants cannot recommend a page they are not allowed to read.
robots.txt is a text file that tells crawlers which parts of the site they may fetch. Plenty of sites block bots they have never heard of, often because a WordPress security plugin, an old “block bad bots” list, or a template added the rules years ago. GPTBot (OpenAI), Google-Extended (the opt-out for Gemini training), ClaudeBot, PerplexityBot and related user-agents are the ones that matter for this job.
They are not the same as the search crawlers. Blocking GPTBot does not block Googlebot. Allowing Googlebot does not allow GPTBot. You can be fully visible in Google and invisible to ChatGPT’s training and retrieval.
OpenAI also uses more than one bot. GPTBot is the main crawler. OAI-SearchBot helps with results inside ChatGPT search. ChatGPT-User fetches a page when someone clicks or shares a link in the chat. Blocking one does not block the others. Check all of them.
The practical test is: open your robots.txt and search for Disallow rules that name those bots, or a blanket block that catches them. Then fix the file. If you would rather not read robots.txt before a coffee, use the free AI crawler checker. It shows the exact rule if something is blocked.
Do not open the whole site to every bot on earth if you have a reason not to. Do not block the assistants and then wonder why they never mention you. That is the usual Sussex-sized mistake: a decent local site, a locked door.
What are schema, FAQ pairs and llms.txt?
These are the three pieces people mean when they say “do GEO” and then wave their hands.
Schema is structured data, usually JSON-LD, on the page. LocalBusiness (or a more specific type such as Electrician) tells a machine your name, trade, area, phone and URL. FAQPage marks up question-and-answer pairs. None of this is a ranking spell. It is a clean label on facts you should already be stating in English. If the visible text says you are a plumber in Brighton and the schema says you are a restaurant in Manchester, you have made things worse.
FAQ pairs are questions written as a customer would ask them, with short factual answers. “Do you cover Hassocks and Hurstpierpoint?” “Are you NICEIC registered?” “Do you do consumer units and EV chargers?” Assistants lift this format easily. So does Google, which can show FAQ-style results when it chooses. Write answers you would stand behind on the phone. Do not write “We are the leading, trusted, number-one choice”. That sentence is how you get ignored.
llms.txt is a short file at the root of the site (yoursite.co.uk/llms.txt) that summarises the business for language-model crawlers: name, trade, area, contact, and a line about what you do. It is not official in the way robots.txt is, and not every crawler uses it yet. Keep it factual, short, and in line with the page.
Together: the human reads the page, the schema labels the facts, the FAQ pairs give quotable answers, llms.txt gives a one-page brief, and robots.txt lets the crawler in. That is the checklist.
What should I not bother doing?
Do not stuff “best electrician best electrician” through the page. Do not invent awards. Do not invent reviews. Do not buy a guest post titled “Top 10 Hassocks Electricians” that nobody reads and an assistant may treat as junk. Do not pay anyone who promises that ChatGPT will name you by Friday.
Do not treat GEO as a replacement for local SEO. If the Google Business Profile is empty and the name, address and phone do not match, you are still hard to trust. Assistants look at the public web, and that includes Maps and reviews. The profile work in the Google Business Profile guide still comes first for a local trade.
Do not expect a county-wide mention from a one-page site. One page will not rank for a whole county, and it will not become the answer for a whole county. Be the Hassocks electrician, the Uckfield heating engineer, the Burgess Hill bakery. Specific is citable. Vague is disposable.
Do not panic if a prompt fails today. Models refresh on their own timetable. You can do everything right and still wait. That is the same honesty as local SEO: weeks to months for the Maps pack, longer for harder terms, and Brighton slower than Hassocks. GEO is not faster just because the interface is a chat box.
GEO from £99 is the one-off version of this checklist on your site. SEO from £99 is the search side. A website, if you need one, is £99 plus £29 a month, or £299 a year. None of those prices buy a mention. They buy a site that is possible to mention.
Frequently asked questions
Is GEO different from SEO?
They overlap. SEO aims to rank in search results. GEO aims to be mentioned in AI-generated answers. Clear facts, consistent details, reviews, schema and crawler access help both. You can be strong at one and weak at the other if the bots are blocked or the copy is vague.
Can I pay to be recommended by ChatGPT?
No. There is no advert slot that puts your name into a normal recommendation. You influence the answer by being clearly described, consistent, well reviewed and readable by the crawlers. Anyone selling a guaranteed ChatGPT mention is selling something they do not control.
How do I know if AI assistants recommend my business?
Ask them the question a customer would ask, and run a checker on your own site. See whether you appear, how you are described, and whether GPTBot is allowed in. Repeat now and then. A single prompt is a snapshot, not a ranking report.
What is GPTBot?
GPTBot is OpenAI’s crawler. If robots.txt blocks it, ChatGPT’s systems are asked not to fetch your pages. You can still show up in Google. You are just harder for that assistant to learn from or quote. Check GPTBot, OAI-SearchBot and ChatGPT-User separately.
What is llms.txt?
A short text file at the root of your site that states the essentials for language-model crawlers: who you are, what you do, where, and how to contact you. It is not a magic ranking file. It is a clear summary that should match the website.
Will GEO get me recommended this month?
Sometimes you will see a better description quickly once the page is readable and factual. Mentions are not on a switch, and models update when they update. Treat it like local SEO: do the work, wait weeks to months, and do not buy a deadline.
How much does GEO cost?
Our GEO work starts at £99 as a one-off: FAQ pairs, schema, crawler access and an llms.txt summary. SEO is also from £99. Neither is a monthly retainer unless you ask for more later.