GEO vs SEO: what actually changed, and what to do about it.

What actually changed

Two things, and neither is “Google stopped mattering”.

First, the answer often replaces the click. Someone who asks a question and receives a written answer with three sources may never visit a tenth result. Ranking first is still valuable, but it is no longer the same as being chosen.

Second, the unit of competition moved from the page to the entity. A search engine matches a query to a document. An answer engine decides which sources it trusts on a subject, then composes. That decision is made about you — your organization, its consistency, what others say about it — not about one page.

What did not change: an engine cannot cite a page it cannot crawl, cannot parse, or that contradicts itself. Which is why the foundation is still SEO.

Side by side

SEO and GEO compared

	SEO	GEO
Goal	Rank in a list of results	Be cited inside a written answer
Unit optimized	The page	The entity (company, brand, person)
Core question	Does this page match the query?	Can this source be trusted, and is it unambiguous?
Main signals	Backlinks, keywords, speed, relevance	Entity consistency, verifiable facts, third-party mentions
Winning format	Comprehensive page targeting a keyword	Direct answer with numbers, a table, named sources, an FAQ
Technical layer	sitemap.xml, robots.txt, Core Web Vitals	Same, plus linked JSON-LD with stable @id, llms.txt, AI crawlers allowed
Measurement	Rankings, impressions, clicks	Share of voice in answers, citations, referrals from AI domains
Time to effect	3–12 months	3–12 months, technical fixes visible in weeks
Fails when	Page is thin or slow	Claims are unverifiable or the entity is ambiguous

Where they overlap — which is most of it

The overlap is larger than the marketing around GEO suggests. Crawlability, page speed, clean HTML, internal linking, content that genuinely answers something, and a sitemap that reflects reality serve both. If your SEO fundamentals are weak, GEO work will not compensate; you are asking an engine to trust a source it cannot read properly.

The honest version: GEO is SEO plus three things — an entity that is stated unambiguously everywhere, facts that can be verified, and content written as an answer rather than as a page targeting a keyword.

Where they genuinely differ

1. Ambiguity is fatal in a way it never was for SEO

Two sites using the same brand name for different businesses, a founder sharing a name with someone more famous, a company that pivoted but still reads as its old self: Google copes, often by ranking both. An answer engine facing a contradiction tends to cite neither and name a competitor instead. Resolving ambiguity has no SEO equivalent, and it is usually the highest-return work available.

2. A false claim costs more

Keyword stuffing made a page rank worse. An unverifiable claim makes a source less trustworthy — and that judgement applies to everything else you publish. We have seen this on our own sites: a single inflated technical claim in a file written for machines is worse than no file at all.

3. Third-party mentions weigh more than links

SEO counts links. Answer engines weigh what independent sources say about you in prose: reviews, forum answers, press, directories. Three people describing your business accurately on Reddit can outweigh a page you wrote about yourself.

How to split the budget

Where to put effort, by starting point

Your situation	Do first	Rough split
No SEO foundation, slow or uncrawlable site	Technical SEO. Nothing else pays until this is fixed.	100% SEO
Decent SEO, no AI visibility work	Entity consistency, JSON-LD with stable ids, llms.txt, crawler access — a few days of work.	75% SEO / 25% GEO
Both in place, competing on content	Citable content: one question per page, real numbers, named sources.	50 / 50
Strong in both, plateaued	Third-party mentions and reviews. The slow lever, the decisive one.	40% SEO / 60% GEO

How to measure something with no rankings

This is the part most teams skip, and then cannot tell whether any of it worked.

• Share of voice. Write down 15–25 questions your customers actually ask. Ask the engines, on a fixed cadence — monthly is enough — and record for each: were you named, in what position, and which sources were cited instead. The trend over months is the metric; a single run is noise.
• Referrals from AI domains. In analytics, segment traffic from chatgpt.com, perplexity.ai, claude.ai and copilot. The volume is small and the conversion rate is usually high: the person arrived with the problem already articulated.
• Branded query volume. If AI engines start naming you, people start searching your name. Search Console shows that before revenue does.
• What not to measure: a “GEO score” from a tool that cannot see inside the engines. If a vendor sells you a number without showing you the questions it asked, the number means nothing.

We run exactly this on our own brands: a fixed question list per company, asked monthly, recording who gets recommended. It is the same diagnostic we run for clients, and the first month is usually uncomfortable.

What we would do first, in order

1. Fix crawlability and speed if they are broken. Everything else depends on it.
2. Make the entity unambiguous: same name, same description everywhere; linked JSON-LD; disambiguation from similar names.
3. Allow the answering crawlers and add an llms.txt. Half a day.
4. Publish three to five pages that answer a real question with real numbers and named sources.
5. Start the measurement before the content, so you have a baseline.
6. Then, and only then, work on third-party mentions — the slowest and most durable lever.

Frequently asked questions