GEO vs SEO: what actually changed for UK local businesses
Most SEO advice still applies. What changed is the shape of the result, the number of winners, and the fact that you can no longer see your own performance.
Generative engine optimisation gets sold as a replacement for SEO. It isn't. Most of what makes a page rank well still makes it citable: clear content, a site that loads, a business that other credible sites talk about. If someone tells you to throw out your SEO and start again, they are selling you a rebuild you do not need.
Three things did genuinely change, and they change what you should measure.
The result has one winner, not ten
A search results page shows ten links, and positions two through six still receive clicks. An AI answer names three or four businesses in a sentence, and the rest do not exist as far as that customer is concerned. There is no page two to be on.
This compresses the value of being "nearly there". In classic search, moving from 15th to 8th was worth money. In an AI answer you are either in the shortlist or you are not, and the gap between fourth and fifth is the entire result.
Your own website is no longer the main source about you
An answer engine assembles its recommendation from sources it trusts, and it is unusual for a business's own site to be the most trusted source about that business. Directories, review platforms, local news, trade bodies and forum threads carry weight precisely because they are not you.
This inverts a lot of standard advice. Rewriting your homepage is largely work on the least influential document in the set. The higher-leverage question is which sources the engine is already citing for your service in your area — and whether any of them mention you.
You cannot see your own performance
This is the one that catches people out. With Google you have Search Console: impressions, positions, queries, all of it free. With AI answers you have nothing. No engine publishes what it said about you, to whom, or how often.
That absence is the whole reason a measurement tool exists. It is also why you should be suspicious of confident numbers. Nobody can tell you your "AI market share" — the engines do not publish the denominator. What can honestly be measured is what a given engine answers to a given question at a given time, recorded and kept. Anything beyond that is modelling, and should be labelled as such.
What carries over unchanged
- Being readable without JavaScript. Several retrieval paths do not execute it.
- Saying plainly what you do and where. Ambiguity is punished harder here, because there is no list of ten results for a customer to disambiguate between.
- Being mentioned by other credible sites. More important now, not less.
- Answering the actual question. Pages that answer a question directly, near the top, in plain language, get quoted. Pages that warm up for four paragraphs get skipped.
What is genuinely new work
- Knowing which questions your customers ask an engine, which are wordier and more conversational than keywords — "who's the best emergency plumber near me that's open now" rather than "emergency plumber leeds".
- Knowing which sources each engine trusts for those questions. This differs per engine and per trade, and it is measurable rather than guessable.
- Getting into those sources. Ordinary business development: listings, coverage, reviews, partnerships. Unglamorous, and it is most of the work.
- Re-measuring. Answers move on their own. One reading is a snapshot, and a snapshot cannot tell you whether anything you did worked.
A note on "AI-ready" scores
You will see scores claiming to tell you how AI-ready your site is, assembled from on-page checks. Those checks are worth doing. But a score computed entirely from your own site cannot tell you what an engine actually says about you, because it never asked one. Those are different measurements, and only the second one answers the question you care about.
The honest version of this is boring: ask the engines, record what they say, keep the evidence, and look again next month.
Our free check asks four engines three commercial questions about your trade in your area, and keeps the answers and the citations so you can read them yourself.