What generative engine optimization actually is
GEO is optimising your pages so an AI engine cites them when it answers a buying question. Same goal as SEO in one sense, be the source, but the reader has changed. A search engine ranks ten links and lets the human choose. A generative engine reads the sources, writes the answer itself, and names one or two. You're no longer competing for a click. You're competing to be the sentence the AI lifts.
That reframes the whole job. The machine isn't scanning for a keyword and a meta title. It's asking, 'does this page plainly answer the thing I was asked?' A page that reads like a warehouse stock feed can rank fine in Google and still get passed over by ChatGPT, because ChatGPT can't turn a stock code into a recommendation it trusts.
How GEO differs from SEO
SEO earns you a position in a list. GEO earns you a mention inside an answer. In SEO you win the click; the buyer lands on your page and you sell to them there. In GEO the AI has already answered by the time the buyer sees anything, so the win is getting named as the supplier worth going to, and the traffic that follows is smaller but far warmer.
The signals overlap but they aren't the same. SEO cares a lot about links, rankings and technical health. GEO cares about whether your page is a citable passage: a clear question, a direct answer, in the buyer's words, with facts a machine can quote without hedging. The old work still matters, an AI can only cite a page it can crawl and understand, but a page that ranks tenth can still get cited if it answers the question more cleanly than the nine above it.
One more difference worth naming: there's no single results page to check. Your visibility now lives across ChatGPT, Perplexity, Gemini and AI Overviews, and each one runs its own searches and trusts its own sources. GEO means thinking about all of them, not just Google.
Why this matters now
The behaviour has already moved. Buyers ask ChatGPT 'who sells this and can they deliver' and act on the answer without opening a search engine. Perplexity answers with citations built in. Google puts an AI Overview above the results for a growing share of queries, and a lot of those queries now end without a single click, the buyer got what they needed from the answer box.
For a merchant that's the whole game. If the AI recommends three suppliers for your product and you're not one of them, you've lost the sale before your site loaded. And the shortlist is short, an AI answer names two or three names, not a page of ten. Being on page one of Google is no longer the same as being in the room.
It's early, which is the opportunity. Most of your rivals are still optimising for last year's search box and haven't checked whether an AI names them. The merchants who fix their pages for citation now get named while the field is thin.
A framework you can act on
Start by finding out what the AI actually says about your category. Ask ChatGPT and Perplexity the real buying questions, 'best place to buy X online in the UK', 'who delivers Y next day', and write down who gets named. That's your baseline: either you're cited, or you can see exactly who's taking your spot.
Next, match your pages to the searches the AI runs, not the question the buyer typed. AI engines rewrite a question into their own queries before they search. Line up four things on your best pages against those queries: the URL slug, the page title, the H1 and the first sentence. Those are the spots a machine weighs when it decides what a page is about. A page at /pipe/15mm-copper-pipe-3m titled in plain English beats the same product sat at /ProductDetails/?Code=CU15X3 every time.
Then make each page a clean, quotable answer. One intent per page. Lead with the direct answer, in the buyer's words, backed by facts a machine can lift, stock status, sizes, delivery, price. Add FAQ blocks that answer real questions in full sentences, and mark them up so the engines can read them. Last, check the plumbing: AI crawlers like GPTBot and PerplexityBot have to be allowed to fetch your pages, or none of the above counts.
Where merchant sites fall down
The usual failure is language. Product and category pages come out of an ERP, MACE, Spruce, K8, built around stock codes, not the words a buyer types. A title like 'CU PIPE 15MM X 3M TABLE-X' means something to your warehouse and nothing to a machine trying to match 'where to buy 15mm copper pipe 3m lengths'. Rewrite the visible title and lead copy to the buyer's words, and leave the stock code in the body for your team.
The second failure is access and structure: pages an AI can't crawl, one page trying to be three things at once, no clear answer near the top, no FAQ schema. None of these need an ERP rebuild. Most systems let you override the visible title and first line, which is where the leverage is. Fix those, and a page that was invisible to an AI becomes a source it can quote.