People have started asking an assistant instead of searching. They type a full question, get one answer back, and often never see a list of links at all. GEO is the work of trying to be in that answer.

It stands for generative engine optimisation. It is new enough that a lot of what is sold under the name is guesswork wearing a confident face, so it is worth being precise about what it is, what can be measured, and what nobody can promise.

How it differs from SEO

SEO competes for a position in a list. Someone searches, gets ten blue links, and you want to be near the top. Success is a rank you can look up.

GEO competes to be the source behind a sentence. Someone asks an assistant, and it writes a paragraph assembled from things it has read. You are either part of that answer or you are not. There is no position two.

That second half matters more than people expect. An assistant describing your business with the wrong prices, an old address or a service you stopped offering is worse than not being mentioned.

What actually seems to help

Honestly: this field is young and the platforms do not publish their rules. What follows is what holds up in practice, not a formula.

• Say the same thing about yourself everywhere. Your name, what you do, where you operate, what you charge — identical on your site, your listings, your profiles. Models trained on contradictory facts produce contradictory answers.
• Write content that answers a question directly. A page that states the answer in the first paragraph and then explains it is easier to quote than one that builds to a conclusion in paragraph nine.
• Use structured data. Schema markup tells a machine what your pages are about rather than making it infer.
• Be mentioned somewhere other than your own website. Directories, publications, profiles. A business that only exists on its own domain is a weak signal.
• Keep the basics working. Pages that load, render without JavaScript, and are not blocked from crawling. Most GEO problems are ordinary technical problems.

What can be measured

Less than you would like, but not nothing:

• Whether you appear when a set of real questions is asked across the main assistants, checked on a schedule
• Whether what they say about you is accurate
• Whether your structured data is valid and your facts are consistent across the web
• Referral traffic from AI tools, where the platform passes it through

That is a baseline you can track over months. It is not a ranking report, and anyone showing you a tidy “AI rank” number has invented it.

What nobody can guarantee

I want to be blunt, because this is where the money gets wasted.

No one can guarantee that an AI assistant cites you, how often, in what words, or for which questions. The models change without notice, the companies do not publish how sources are selected, and the same question can return different answers to different people on the same day.

If someone offers you guaranteed AI visibility, they are either misunderstanding the technology or counting on you to.

Is it worth doing yet?

It depends on who your customers are.

Probably yes if you sell something people research before buying, if your buyers are comfortable with these tools, or if your competitors are already turning up in those answers and you are not.

Probably not yet if your customers find you through word of mouth and local search, or if your ordinary SEO is still broken. Fixing the foundation first is not a delay — most of that work is what feeds the models anyway.

The useful starting point is cheap: find out what the assistants currently say about your business. Sometimes the answer is nothing. Sometimes it is something wrong that has been quietly costing you.