GEO vs SEO: what changes, what carries over

What carries over from SEO

Almost all of the foundations. This is the part vendors selling "the death of SEO" leave out. An assistant does not have a private index of the web sitting beside the one search engines use. When it needs current information it runs web searches, reads what comes back, and writes from that. Everything that determines whether your page comes back in those searches is classic SEO.

Specifically, these still decide whether you are even in the running:

• Crawlability and indexing. If a crawler cannot fetch it, no model can cite it.
• Server-rendered content. Content assembled by JavaScript in the browser is frequently not seen at all. This is a harder constraint for AI retrieval than it is for Googlebot.
• Topical relevance and depth. Pages that genuinely answer a question rank, get retrieved, and get quoted. The same asset serves both jobs.
• Links and authority. Retrieval leans on the same ranking signals, so the pages that rank are the pages that get read.
• Speed and clean HTML. Fetching has timeouts. Slow pages get dropped quietly.
• Local fundamentals. An accurate Google Business Profile and consistent name, address and phone details feed both the map pack and the assistant.

So the honest framing is: GEO is not an alternative to SEO. It is a second layer of work that only pays once the first layer exists. If your site is technically broken, fix that before anyone sells you an AI strategy.

What is genuinely new

Four things, and they are the reason a separate discipline exists rather than a new checklist item.

You are optimising for searches you never see

A user prompt does not become one query. Assistants typically expand it into several differently phrased searches, often with the current date attached, and retrieve from those. You are not competing for the sentence the user typed. You are competing for the machine's reformulations of it. That changes which phrasings you need to exist in, and it is why keyword lists translate badly.

Most of the deciding surface is not your website

When an assistant assembles a picture of your business, your own site is one source among many, and it is the least independent one. Review platforms, directories, local guides, "best of" listicles, Reddit threads and trade press are all read too. In practice, a substantial share of GEO work is not on your website at all. It is making the rest of the web describe you accurately and consistently.

Format is a ranking factor in a way it never quite was

Models lift text. Short declarative sentences, a direct answer under a question-shaped heading, tables with one row per dimension, plain facts with dates. Persuasive brochure prose is nearly unusable as source material, because there is nothing in it a model can safely state as fact.

Results are probabilistic, so measurement changes shape

The same question asked twice can produce two different shortlists. There is no rank to track. You freeze a prompt set, sample it repeatedly across assistants, and report appearance rates and trends. Anyone showing you a single clean "ChatGPT ranking" number is hiding the variance. There is more on this in what generative engine optimization is.

Is GEO replacing SEO?

No. It sits on top. The clearest way to see it: an assistant cannot cite a page it cannot retrieve, and retrieval runs through search infrastructure. So SEO remains the entry ticket. What is genuinely shifting is where the value lands after the ticket is punched.

Informational queries that used to send traffic increasingly get answered in place. If your content strategy was built on winning those clicks, that model is under pressure, and pretending otherwise does not help. Meanwhile the value of being the named recommendation goes up, because an answer that names four businesses concentrates demand far more sharply than a page of ten links.

The practical conclusion is unglamorous. Keep doing SEO. Stop measuring it only by traffic from informational queries. Add measurement of whether you are named in the answers, and add the work that is not on your own site.

GEO agency vs SEO agency: which do I need?

Diagnose before you buy. Three cases cover most businesses.

• You are invisible in ordinary search too. Hire SEO first. GEO has nothing to stand on, and a GEO engagement would spend its first three months doing SEO anyway.
• You rank well but assistants never name you. This is the GEO case. Your site is fine and the problem is elsewhere: how you are described across the web, whether third parties corroborate you, whether your content is extractable, whether AI crawlers are allowed in.
• You already have an SEO agency you like. Keep them. Add GEO as direction rather than duplication, and make the split explicit so nobody bills twice for the same fix. Plenty of our engagements are advisory for this reason.

Two things to ask any agency pitching AI visibility. First, what exactly do you measure, and does your reporting acknowledge that answers vary between runs? Second, what proportion of the work is off our own website? If the answer to the second is "none", they are selling you SEO with a new label.

If you want to see where you actually stand before deciding, the DIY method is written out in full on the GEO explainer, the mechanism is on how AI assistants choose businesses, and the rest of the common objections are on the FAQ.