SEO vs GEO: what changes, what does not, and how to do both
Published · Viscovery, Pepper Square
The short answer. Search engine optimization (SEO) helps a website rank in a list of results and earn a click. Generative engine optimization (GEO) helps a brand be read, trusted and named inside the answers ChatGPT, Claude, Gemini and Perplexity compose. GEO builds on SEO: nothing that cannot be crawled is cited. It adds three things search did not ask for: permission for a new set of crawlers, content written as quotable answers, and evidence outside your site that backs up what it says.
What stays the same
A site that is slow, unindexable, thin or duplicated fails both. Clean URLs, one canonical host, a sitemap, valid structured data, real content depth and pages that load quickly on a phone are prerequisites for being cited as they are for ranking. The assistants build much of their picture of the web from the same crawls and the same search results SEO works on; a page Google cannot find is not going to be quoted by Perplexity. If the SEO basics are not in order, GEO work is premature.
What changes
| SEO | GEO | |
|---|---|---|
| Where you appear | A ranked list of links | A sentence in a composed answer, sometimes with a citation |
| Who reads your site | Googlebot, Bingbot | GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, Claude-User, PerplexityBot, Google-Extended, CCBot, Applebot-Extended; each needs permission in robots.txt |
| The unit that wins | A page matched to a query | A passage that answers a question on its own |
| What authority means | Backlinks and domain strength | Corroboration: third-party mentions, consistent facts, dated and attributed content |
| Structured data | Rich results; FAQ and HowTo rich results withdrawn for most commercial sites in 2023 | Entity clarity: Organization, WebSite, Service, FAQPage read by assistants |
| Site description | Title and meta description | Those, plus llms.txt and a plain entity sentence on the homepage |
| How you measure it | Rankings, impressions, clicks | Mentions, citations, share of answers, accuracy of the description. No click is recorded |
The crawlers are different and need permission
Each assistant reads the web through its own user agent, and robots.txt decides per agent. A rule can be present without anyone having decided it: it came in with a template or a plugin. Blocking Google-Extended is a deliberate choice for some publishers, to keep content out of Gemini’s training; it also limits what Gemini can use from the site. In SEO a robots.txt mistake costs a page; in GEO one line can remove the site from an assistant entirely. This is the first thing the free AI visibility check reads.
The unit of content is a passage
Search ranks pages against queries. An assistant reads a page, selects a passage, and quotes or paraphrases it. Content that works for this has the question as a heading, the direct answer as the first sentence under it, and a self-contained paragraph of roughly 120 to 160 words that survives being lifted out of context. Most marketing pages open with why the topic matters and answer the question in paragraph three. Answer engine optimization, AEO, is the name for this discipline.
Authority becomes corroboration
SEO’s authority is largely links. An assistant deciding whether to repeat a claim looks for different evidence: the same facts stated consistently elsewhere, content dated so the model can tell it is current, a named organization behind the site, other sources saying what you say. A brand can have a strong backlink profile and still read as unverifiable, and the assistant will prefer a smaller source it can place.
Measurement has no click in it
A customer who asks an assistant and gets a full answer may never visit any site. Rankings, impressions and clicks do not record whether the assistant mentioned you, described you accurately, or recommended someone else. GEO is measured by asking the assistants the questions that matter and recording the answers, repeatedly, because the answers vary between runs. That is what the audit does that an analytics tool cannot.
A 30-day plan that does both
- Week 1, access. Fix robots.txt for the ten AI crawlers. Publish llms.txt. Confirm one canonical host and a sitemap. Hours, not days.
- Week 2, identity. Put a one-sentence definition of the business under the H1 and in the meta description. Add Organization and WebSite structured data. Name the organization in the footer.
- Week 3, answers. Take the ten questions customers ask and give each a page or a section, question as heading, direct answer first. Add FAQPage schema that matches the text.
- Week 4, evidence. Date the content pages. State the same business facts identically on every profile and directory. Earn mentions from sources in your field; a listing on a page that already ranks for your category is worth more than a generic directory.
Then measure: run the free check to confirm the access and identity work landed, and ask the assistants your questions to see whether the answers changed. Repeat monthly.
Is GEO replacing SEO?
No. Search still sends most commercial traffic, and the assistants lean on the same crawlability and search results SEO has always worked on. GEO is a set of additional requirements on a sound SEO foundation, not a rebrand of it. What is changing is the share of buying decisions that begin with a question to an assistant rather than a query to a search box. We have not measured that share; the brands that appear in those answers will be the ones that did both kinds of work. The short version is on the FAQ.
Where to start
Run the free AI visibility check. It reads your robots.txt, llms.txt and homepage and reports whether the assistants can read you: week one of the plan above, measured. The paid report does the diagnosis for weeks two to four: who the assistants recommend for your keywords instead of you, and what to build first.