GEO and SEO: what actually changes

Your SEO work is not wasted, and generative engine optimisation is not a replacement for it. The pages that rank are often the pages an assistant reads. What changes is the target: ranking puts you on a list somebody chooses from, and being cited puts you inside the answer they were given.

Four things that carry over

Being crawlable. If your content only appears after JavaScript runs, many crawlers see an empty page. This matters more for assistants than for search engines, because the crawlers that feed them are generally less patient.

Topical depth. A site with twenty real pages about one subject is read differently from a site with one page about twenty subjects. That was true before and it is still true.

Being linked to and mentioned elsewhere. An assistant that has seen your name on three independent sites treats it differently from a name it has only seen on your own.

Page speed and structure, up to a point. They do not directly make you more quotable, but a page a crawler gives up on is a page that cannot be quoted.

Four things that genuinely change

Keyword matching stops being the mechanism. An assistant is not looking for your page because it contains a phrase. It is looking for a page that lets it answer confidently. A page stuffed with a phrase and thin on substance is worse than useless here, because it reads as something to distrust.

Position ten is gone. Ranking tenth still puts you on the page. Being the tenth thing an assistant might have said means it said nothing about you.

Hedging costs you. "Solutions tailored to your needs" gives an assistant nothing to repeat. "Covers 25 High Courts and 8,200 courts and tribunals" gives it a sentence. The second one gets quoted.

The feedback loop is slower and noisier. Rank tracking gives a number every day. An assistant's answer varies between runs, so a single check tells you little and you need a range rather than a figure.

Where the two disagree, and which to follow

There are places where good SEO advice and good GEO advice genuinely conflict. These are worth knowing, because following the wrong one costs you months.

• Keeping the reader on the page. SEO rewards holding attention and burying the answer so somebody reads on. An assistant wants the answer immediately and will skip a page that withholds it. Put the answer first.
• Covering everything on one long page. A comprehensive hub can rank well. A summary is lifted from a passage, so a page that answers eight questions loosely often loses to eight pages that each answer one tightly.
• Avoiding mention of competitors. Conventional advice is to keep them off your site. Assistants answering a comparison question reach for the source that has already weighed the options, and a page that names nobody else has weighed nothing.
• Publishing prices. Sales teams resist it and it is one of the strongest single things you can do here, because a business with a public price can be placed in an answer and one without often cannot.
• Volume. Publishing regularly is reasonable SEO advice. Here, a page nobody asks a question about contributes nothing at all, however often you publish.

Where this went wrong in our own run

On 6 August 2026, on clawlaw.in, the pricing page rendered its prices only after scripts ran. A visitor saw the prices. A crawler received a page with no prices in it at all. The page was perfectly good SEO and it was unreadable to the thing we needed to read it.

The consequence was not that prices went unmentioned. It was worse: the prices the assistants quoted came from an app store listing instead, and that listing carried different plan names and different amounts from the website. ChatGPT noticed the contradiction and said so in its answer.

So a page that ranked, on a site that was otherwise working, handed the assistants a competing version of our own prices. Nothing about keywords or content quality would have surfaced that. Only fetching the page the way a crawler does would.

What to do first if you already do SEO

Take the ten questions your buyers actually ask, written the way they would type them. Ask them of two or three assistants without naming your company. Write down who gets named and which pages get cited.

That list of cited pages is the most useful thing you will get. It tells you which sources the assistants trust for your category, and almost always includes at least one you did not expect and are not on.

Then do the mechanical check that costs nothing: fetch your most important pages the way a crawler does, without scripts, and see whether your prices, specifications and claims are actually in what comes back. This is the most common silent failure and it is invisible when you look at the page in a browser.

What this does not change

It does not make your existing rankings less valuable. Traffic from search still converts, and in most businesses it still converts better than anything else.

It also does not mean your old content was a mistake. Most of it was written for a channel where being interesting earns attention. Assistants do not browse, they answer, and that is a difference in the channel rather than a fault in the writing.

Common questions

Should I stop doing SEO?

No. The overlap is large and search still sends real buyers. Treat this as a change in what winning looks like for work you are already doing, not as a new budget line that replaces it.

Will being summarised cost me clicks?

Sometimes, and it is worth planning for honestly. Somebody who gets their answer from a summary may never visit. That is why the questions worth targeting hardest are the ones where a person still has to decide, compare, price or contact, because those are the questions where being named leads somewhere.

Do the same pages work for both?

Often, with one change: move the answer to the top. A page that answers in its opening lines and then earns the rest with detail tends to do well in both, and a page that withholds the answer does well in neither any more.

Which should I measure?

Both, separately, and never averaged. Rankings move daily and predictably. Mention rates move slowly and noisily, and need several readings and a range before a change means anything.

My agency says they already do this. How do I check?

Ask them for the question set and confirm that not one question names your company. Then ask how many assistants each question was asked on, and whether each answer came from a live search or from memory. If they cannot answer those three, they are reporting something other than your visibility.

What about structured data and schema?

It helps machines parse a page and it does not substitute for having something worth parsing. Mark up a page with a real price, a real date and a real list and it helps. Mark up a page of adjectives and you have made adjectives machine readable.

How we know this

We read the assistants rather than guessing what they do. In our own audit of 1,260 saved answers from ChatGPT and Gemini across the United States, the United Kingdom and India, 82.0% of every citation landed on a site the business itself owned, and review directories took 0.7%. Counted as one domain per reading, 16,516 of 20,143.

That is the whole asymmetry: the pages you own are the pages an assistant quotes back to a buyer. See the dated readings and the method on our proof page.

Find out whether assistants name your business, on your own domain and your own buying questions.

By Aditya Goel, Founder, AI Knows Us. Reviewed 1 November 2026.