GEO vs SEO: Which Work Carries Over and Which Does Not
SEO gets your pages ranked in a search index. GEO gets your company named or cited inside AI answers. Google calls optimizing for its AI features "still SEO", so technical work carries over. Measurement does not: in our S&P 100 study, 32 of 48 named companies were named by all four assistants, a count lowered by missing ChatGPT answers.
Published · Last updated
By Stiven Hernandez, founder, UPAI · Updated 2026-10-05
Key takeaways
The short version of this page, in the order the sections below develop it. Each point is stated here in full so it can be read without the rest of the page, and the section that argues it gives the measurement and the source behind it.
- Google's own guide says optimizing for generative AI search is "still SEO": indexing, crawlability, page experience and original content all carry over.
- Google says site owners can ignore llms.txt files, content chunking and rewriting for AI when the target is Google Search. That guidance covers Google only.
- New work: allowing OAI-SearchBot, PerplexityBot and Claude-SearchBot, each of which has its own robots.txt rule.
- No single ranking exists. In our S&P 100 study, 32 of 48 named companies were named by all four assistants. ChatGPT returned fewer usable answers, which lowers that count by an amount we cannot state.
In our S&P 100 study, 48 companies were named by at least one of four assistants, and 32 of those 48 were named by all four.
Our own measurement: Four assistants (ChatGPT, Claude, Gemini, Perplexity) asked unbranded buying questions, run 2026-09-25; 342 usable answers in the eight sectors analysed. There ChatGPT returned text in 54 of 96 answers and the other three in 96 of 96, which lowers the all-four count by an unknown amount.. See the study
What is the difference between GEO and SEO?
SEO is the work of getting your pages indexed and ranked by a search engine; GEO is the work of getting your company named or cited inside an AI-generated answer. The term GEO comes from a paper by Aggarwal and five co-authors, first posted in November 2023 and accepted to KDD 2024, titled GEO: Generative Engine Optimization.
Google does not treat them as two disciplines. Its guide to generative AI features, last updated 2026-07-10, puts it this way:
"From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO."
That sentence covers Google Search only: AI Overviews and AI Mode. It says nothing about ChatGPT, Claude or Perplexity. Their makers document their own crawlers, and we have not found ranking guidance of the same kind from them.
Which SEO work carries over to GEO?
The technical base carries over unchanged, original content carries over and Google places it above its other advice, and per-keyword page production does not carry over. Google's AI features documentation states that there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary."
| Task | Carries over? | What the primary source says |
|---|---|---|
| Indexing and snippet eligibility | Yes, unchanged | Google: to be a supporting link in AI Overviews or AI Mode, "a page must be indexed and eligible to be shown in Google Search with a snippet". |
| robots.txt and crawl access | Yes, with more crawlers | Google: "robots.txt directives for Googlebot is the control" for how sites are crawled for Search, AI features included. OpenAI, Perplexity and Anthropic each document a separate search crawler. |
| Internal links, page experience, content in text form | Yes, unchanged | Listed in Google's AI features documentation among the SEO fundamentals that "continue to be worthwhile". |
| Original, first-hand content | Yes, and Google puts it first | Google says this will likely influence presence in generative AI search "more than any of the other suggestions in this guide". |
| Structured data | Keep it; it is not required | Google: "Structured data isn't required for generative AI search". Google adds that it still helps with eligibility for rich results. |
| A page per keyword variation | No | Google: you need not have "captured every variation" of a query, and creating separate content for every variation "primarily to manipulate rankings or generative AI responses" violates its scaled content abuse spam policy. |
| Keyword stuffing | No | The GEO paper found it gave "little to no improvement" in its test engine. |
| Link building | Unknown | Neither Google page cited here mentions it. We have not found a primary source from any assistant that says how links are weighed. |
| Rank tracking on one results page | No | Each assistant returns its own set of names. |
For Google, "carries over" is Google's statement about its own product. For the other assistants, the only thing we could read at a primary source is crawler access, not how they choose which company to name.
What does GEO add that SEO never asked for?
GEO adds three tasks that a traditional SEO programme never had to do: letting other companies' search crawlers in, measuring answers across several assistants instead of reading one rank report, and checking what sources you do not control already say about you. None of the three is a ranking tactic, and the third is the one most teams have no process for at all. Some tactics sold as GEO appear on Google's own list of things to ignore.
- Allow the search crawlers, separately from the training crawlers. OpenAI's crawler documentation says sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers, though can still appear as navigational links." GPTBot, which crawls content that may be used for training, is a separate setting: OpenAI says each one "is independent of the others". Perplexity recommends allowing PerplexityBot, and Anthropic says disabling Claude-SearchBot "may reduce your site's visibility and accuracy in user search results."
- Measure by asking, not by reading a rank report. Google offers a Generative AI performance report in Search Console. The crawler documentation we read from OpenAI, Perplexity and Anthropic describes access controls, not a performance report. The remaining method is to ask buyer questions and record the names, which we describe in how to know if ChatGPT mentions your brand.
- Read what third-party sources say about you. Google's guide notes that its AI features "can show what's being said about products and services across the web", and warns that seeking inauthentic mentions "isn't as helpful as it might seem."
Some tactics promoted as GEO are on Google's list of "things you can ignore for Google Search":
- llms.txt files. Creating one "will neither harm nor help your site's visibility or rankings in Google Search, as Google Search ignores them."
- Chunking. "There's no requirement to break your content into tiny pieces for AI to better understand it."
- Rewriting for AI. "You don't need to write in a specific way just for generative AI search."
That list is about Google Search only. The same guide says it is "completely fine" to keep such files for other services that use them. Google's steps for its own AI features are in our piece on appearing in Google AI Overviews.
Why is there no single ranking to report in GEO?
There is no single ranking to report in GEO because the assistants do not return the same companies, so a reading from one is not a reading of the market. We measured this in our S&P 100 study, run on 2026-09-25. Four assistants were asked unbranded buying questions in eleven industries; the eight with enough usable answers to analyse account for 342 usable answers. Forty-eight companies were named by at least one assistant. Of those 48, 32 were named by all four, a measure we call Model Agreement.
One limit sits on that number. In those eight sectors ChatGPT returned text in 54 of 96 answers, while Claude, Gemini and Perplexity returned 96 of 96. Fewer answers means fewer chances to name a company, which pushes "named by all four" down. So 32 of 48 likely understates agreement, and we cannot say by how much.
| Assistant | Distinct S&P 100 companies named | Answers with text, eight sectors |
|---|---|---|
| Gemini | 46 | 96 of 96 |
| Claude | 39 | 96 of 96 |
| ChatGPT | 36 | 54 of 96 |
| Perplexity | 34 | 96 of 96 |
What does not rest on ChatGPT: the three assistants with complete answers still named different numbers of companies.
These are S&P 100 companies, so this is not a figure for mid-size firms. Financials, Health Care and Information Technology were excluded for too few usable answers. Only Perplexity searched the live web; the other three answered from what they had learned. And a mention is not a recommendation, a click or a sale.
The practical consequence: a GEO report needs one line per assistant. Google's guide adds a point about tooling: "No third-party tool has access to our internal ranking or AI systems." Our comparison of tools that track brand visibility in AI search covers what those products can and cannot observe.
Does the research behind GEO prove that it works?
The research behind GEO proves less than it is quoted for: in a 2023 test bench, some edits changed how much of a generated answer a source received. The paper's headline sentence is that GEO "can boost visibility by up to 40% in generative engine responses." Four details belong beside it.
- The figure is a relative gain on the paper's own metrics. The authors write that "the best methods improve upon baseline by 41% and 28% on Position-Adjusted Word Count and Subjective Impression respectively." The first is a word count weighted by position in the answer. Each part of the second "is evaluated using GPT-3.5", so a language model does the scoring.
- The engine was the authors' own pipeline. The answer was "generated by the gpt3.5-turbo model". A separate check ran on Perplexity.ai.
- The sources were already ranking. "Only the top 5 sources are fetched from the Google search engine for every query." The paper tested which of five retrieved pages got more of the answer, not how a page gets retrieved.
- "Up to" is a ceiling for the best method. In the Perplexity.ai check, keyword stuffing "performs 10% worse than the baseline." That check used "a subset of 200 samples", with the source text supplied "as file uploads" rather than found by Perplexity's own search.
Both metrics score a source's share of one generated answer. Neither is whether a buyer's question returns your company name. We do not know of a peer-reviewed study that measures that on current assistants.
When does the GEO vs SEO split not matter?
The GEO vs SEO split does not matter when your buyers reach you through Google alone, because for Google the two are the same work by Google's own account. It starts to matter under three conditions:
- Your buyers ask assistants other than Google. Then crawler access and per-assistant measurement are new work.
- The assistant answers without searching. In our study three of four assistants answered from what they had learned. Changing a page today does not change that answer today.
- You are indexed but unnamed. Ranking and being named are separate observations. Our piece on whether GEO is worth it for a mid-size company covers that decision.
Questions people ask next
These are the follow-ups we are asked most often once the answer above has landed. The same rule applies to each: where a figure appears, its source and year sit in the same sentence, and where we do not know something we say so rather than estimate it.
- Is GEO replacing SEO?
- No. Google's guide for site owners states that its generative AI features are rooted in its core Search ranking and quality systems, and that optimizing for them is still SEO. A page that is not indexed cannot be shown as a supporting link in AI Overviews or AI Mode.
- Do I need an llms.txt file for GEO?
- Not for Google. Google's guide says Google Search ignores llms.txt files and that creating one will neither harm nor help visibility or rankings in Google Search. We have not read a primary source from another assistant stating that it uses them for search answers.
- Does keyword research still matter for GEO?
- Knowing what buyers ask still matters. Building a page for every wording does not. Google's guide says you need not worry about lacking long-tail keywords or capturing every query variation, and that creating separate content for every variation primarily to manipulate rankings or AI responses violates its scaled content abuse spam policy.
- Can a company rank first on Google and still be missing from AI answers?
- It can, outside Google; we have not measured how often. A ranking is a position in one index, and being named by an assistant is a separate observation. Our S&P 100 study did not compare Google rankings with assistant answers. In it, 32 of the 48 companies named by at least one of four assistants were named by all four. ChatGPT returned fewer usable answers than the other three, which pushes that agreement down by an amount we cannot state.
Related
Other questions in this library that cover the ground next to this one. Each is answered the same way — the direct answer first, then the measurement behind it — and each carries its own sources, so you can check any of them on its own page without taking this one on trust.
Sources
Every figure on this page comes from one of the sources below, each linked so you can open it and check the number in its original context rather than against our summary. Where a figure is our own measurement, its method and sample size are stated with it.
- Google Search Central — Optimizing your website for generative AI features on Google Search (last updated 2026-07-10)
- Google Search Central — AI features and your website
- Aggarwal et al. — GEO: Generative Engine Optimization, KDD 2024 (arXiv 2311.09735v3)
- OpenAI — Overview of OpenAI Crawlers
- Perplexity — Perplexity Crawlers
- Anthropic — Does Anthropic crawl data from the web, and how can site owners block the crawler?
