GEO vs SEO: what actually changes, and what doesn't

GEO (generative engine optimization) means getting your pages used and cited in AI-generated answers such as Google's AI Overviews and AI Mode, ChatGPT search and Copilot. Google and ChatGPT still draw those answers from search results, so the SEO groundwork carries over. What changes is how pages get picked, how you measure the results, and which switches you control.

Where "GEO" comes from

GEO stands for generative engine optimization. The term comes from a research paper, "GEO: Generative Engine Optimization", first posted to arXiv in November 2023 and accepted to KDD 2024. Its authors introduced GEO to help content creators improve their visibility in generative engine responses. The same letters also mean geo-targeting (delivering content or ads by location), and at least one page ranking for "geo vs seo" in October 2026 is about that.

Microsoft now uses the term itself: in February 2026 it called Bing's AI Performance report "an early step toward Generative Engine Optimization (GEO) tooling". Google's guide to its generative AI features, updated in July 2026, names GEO and AEO (answer engine optimization) and says "many suggested 'hacks' aren't effective or supported by how Google Search actually works."

What stays the same: AI answers still start with search

Google's SEO Starter Guide defines SEO as helping search engines understand your content, and helping users find your site and decide whether to visit it. A client's pricing page that ranks for "payroll software for small business" and gets the click is SEO doing its job.

Google's guide says its AI features work by "relying on our core Search ranking systems to retrieve relevant, up-to-date web pages from our Search index", a technique it calls retrieval-augmented generation, or grounding. To be eligible, a page must be indexed and eligible to show in Search with a snippet, the same bar as a regular result, so a page that misses it can't appear in Google's AI features at all. Asked whether SEO is still relevant for generative AI search, the guide answers "In short, yes!"

ChatGPT search also leans on search: when OpenAI launched it in October 2024, it said the product uses third-party search providers as well as content provided directly by its partners. Even the 2023 paper that coined the term tested only pages already in Google's top 5. That reliance on search is the argument for keeping one content plan for Google and AI search.

What changes

	SEO	GEO
What you compete for	A position in a list of links	A share of one generated answer
What gets picked	A page, judged partly by its passages	Pieces of pages that earn a place in the answer
How many searches	The query the person typed	Several related searches run for one question
What you can measure	Clicks, impressions, position	Google: AI-feature impressions (own report), clicks only within Web data. Bing: citations
What you can switch off	Crawling, through robots.txt	Search and training crawlers separately, plus Google's AI-features setting

Pieces of pages

Passages are not new to search: classic Google Search already uses passage ranking, an AI system that identifies individual sections of a page to judge how relevant the page is to a search. Microsoft's guidance on AI search answers, from October 2025, describes what changes: "ranking still happens, but it's less about ordering entire pages and more about which pieces of content earn a place in the final answer."

More searches per question

Google's guide describes query fan-out: the model generates several related searches at once to fetch more results for one question. Google's page on AI features, last updated December 2025, adds that AI features can show a wider and more diverse set of links than a classic web search.

Ahrefs, a tool vendor, and ScaletoTop have both measured how often pages cited in AI Overviews also rank on the first page for the same keyword:

Study	Sample	Also in the top 10
Ahrefs, July 2025	Top 3 citations per AI Overview	76%
Ahrefs, March 2026	863,000 keyword SERPs, organic links only	37.1%
ScaletoTop, October 23, 2026	38 GEO and AI-visibility keywords, US	23% of 305 citations

Ahrefs' 2026 study used its own Brand Radar tool and an improved parser. Counting the first 10 blocks of the results page instead of organic links, it found 37.9%; by organic position, 26.2% of cited URLs ranked 11 to 100 and 36.7% outside the top 100. Given those method changes, the two Ahrefs studies aren't strictly comparable in our view.

In ScaletoTop's check, 37 of the 38 keywords showed an AI Overview. Their 305 citations (a page cited for several keywords counts once per keyword) pointed to 196 distinct pages. For 12 of the 37 keywords, no cited page appeared in the first-page organic results, and for "geo vs seo" itself, none of the 5 cited pages was among the 9 organic results. The check covers one day, one country, one niche and the first results page only, so it is a second, smaller data point beside Ahrefs' organic-only figure.

This suggests that appearing in the first-page organic results for the exact keyword covered a minority of what got cited, which fits query fan-out and Google's statement about a wider set of links. It doesn't show that rankings stopped mattering: cited pages may rank lower for the same keyword (Ahrefs puts 26.2% at positions 11 to 100) or for fan-out queries the check didn't cover.

Google's guide rules out the obvious reaction, a page for every fan-out query: creating separate content for every variation of a search, mainly to manipulate rankings or AI responses, violates its scaled content abuse policy.

What you can switch off

Since August 31, 2026, every site can opt out of AI Overviews, AI Mode and Discover's AI features with a Search Console setting. Inclusion is the default, and Google says the setting isn't used as a ranking signal elsewhere in Search. Google's July 2026 guide makes inclusion a condition of eligibility, while its December 2025 page still says there are "no additional technical requirements"; given the default, we take it that nothing needs changing unless someone opted the site out.

For crawling, Google's December 2025 guidance says robots.txt rules for Googlebot remain its control for Search, AI features included, and blocking Google-Extended, the token Google points to for AI training, doesn't affect inclusion or ranking in Google Search. OpenAI runs separate search and training crawlers (OAI-SearchBot and GPTBot) and says the settings are independent, so blocking GPTBot alone doesn't keep a site out of ChatGPT search.

The engines don't agree

Google's guide (updated July 10, 2026) and Microsoft's advice (October 2025, plus Bing's February 2026 post) disagree on chunking, schema and writing style.

	Google	Microsoft
Chunking content	Not required	Copilot parses content into pieces; use clear headings, tables and modular layouts
Schema	Not required for AI features; keep it for rich results	Use schema
Writing style	No need to write in a specific way for AI	Be cautious with em dashes
Is SEO enough?	SEO best practices remain relevant; no AI text files or special markup needed	Fundamentals remain essential, but are "just the starting point"

On schema, the gap is narrower than it looks, since Google keeps recommending structured data for rich results. Two types that older GEO advice recommends no longer produce rich results at all, though: Google's documentation changelog shows HowTo rich results gone since September 2023 and FAQ rich results since May 7, 2026.

The AI engines also differ among themselves. A 2025 preprint by Chen et al., using ranking-style prompts such as "Top 10 ... brands" on API versions of the models, found that AI search engines favored earned, third-party sources over brand-owned and social content, more than Google did, and differed significantly from each other.

Our recommendation: where a client's search traffic comes mostly from Google, follow Google's guide and don't restructure content for AI, but use Microsoft's structure advice, such as clear headings and tables, where it also helps readers; we don't expect that to cost anything in Google. Where Copilot or Bing traffic matters, follow Microsoft's structure and schema advice as well.

Measuring it: what the engines report in 2026

Google and Bing now report AI visibility to site owners, in different units.

Report	What it counts	Since
Google: Performance report, "Web"	Clicks, impressions and position, AI features included	Documented by Dec 2025
Google: Generative AI performance report	Impressions in AI features only	All sites Aug 31, 2026
Bing: AI Performance (public preview)	Citations, plus a sample of grounding queries	Feb 10, 2026

Google's Generative AI report reached some sites on June 3, 2026 before the full rollout. Bing's grounding queries are the phrases the AI used to retrieve the content it cited.

In the Performance report, per Search Console Help, an AI Overview takes a single position shared by all its links, a click on an external link in it counts as a click, and a follow-up question in AI Mode counts as a new query. The Generative AI performance report counts how many times links to the site were shown in an AI feature, and two links from the same site in one AI result count as a single impression in the chart total. Sites with too few impressions may not see it, and Google said at launch that it may add metrics over time.

Bing's AI Performance report covers Copilot, AI-generated summaries in Bing and some partner integrations. Because Google counts impressions and Bing counts citations, don't add the two together or compare one against the other.

We couldn't find a comparable report for site owners from OpenAI, Perplexity or Anthropic as of October 2026. As far as we can tell, seeing how often those engines cite a site means running prompts and record the answers, by hand or with a third-party tool (see what to measure before you buy a visibility tool).

Is SEO dead? What the click data says

No. Google's guide says SEO best practices remain relevant for generative AI search. Traffic is the harder question, and the evidence points two ways.

Pew Research Center studied the browsing data of 900 US adults in March 2025. In that sample, 18% of Google searches produced an AI summary. Users who saw one clicked a traditional result in 8% of visits, against 15% for users who didn't, and clicked a link inside the summary in 1% of visits. Long searches got summaries far more often: 8% of one- or two-word searches did, against 53% of searches with 10 words or more. The data covers one month of US browsing.

Google tells a different story. In an August 2025 post, it said total organic click volume from Search to websites had been "relatively stable" year over year, while user trends shifted traffic between sites, so some sites lost traffic and others gained. It called third-party reports of dramatic declines inaccurate. The post doesn't include the data behind that statement.

We think both can be true at once: fewer clicks on searches that show an AI summary, and steady totals overall. ScaletoTop's own content plan tracks 70 keywords, and 69 of them showed an AI Overview on October 22, 2026, which suggests that for topics like these, AI answers and classic results compete on the same results page. Google's May 2025 advice is to judge that traffic by conversions, such as sales and signups, as well as clicks.

GEO tactics with evidence behind them

Before repeating any percentage from the 2023 GEO paper, check what it tested. The engine was simulated: for each query it read only the top 5 Google results, and GPT-3.5 wrote the answer. The main metric, behind the percentages in the table, was a source's share of the answer's words, adjusted for position, so the results describe word share in a lab setup, not citations or traffic.

Method (paper's name)	Change vs. no optimization	Note
Quotation Addition	+41%	Best method in the main test; also best on Perplexity, at +22%
Statistics Addition	About +31%	Our calculation from the paper's Table 1
Cite Sources	About +27%	Our calculation from Table 1
Keyword Stuffing	Little to no improvement	10% worse than no optimization on Perplexity

The authors sum up their top three methods as 30-40% improvements, though their Table 1 puts Cite Sources a little lower, at about +27% by our calculation. Results varied by topic. The authors also report that lower-ranked sources gained more: with Cite Sources, visibility rose 115.1% for sites ranked fifth, while the top-ranked site's visibility fell 30.3% on average. And they write that they "didn't evaluate how GEO methods affect search rankings."

Read together with the engines' own documents, that evidence supports six practical rules:

• Keep pages indexed and eligible for a snippet in Google, and crawlable for Bing, which Microsoft calls essential.
• Check robots.txt for these search crawlers: Googlebot, OAI-SearchBot, PerplexityBot and Claude-SearchBot. In Search Console, check that nobody has excluded the site from Google's AI features.
• Add statistics, sources and quotations where they help readers. They were the paper's strongest methods, and Google's guide asks you not to "just recycle" what others have already said.
• Earn real coverage on other sites. Chen et al. found AI engines favor earned, third-party sources, and Google says seeking inauthentic mentions "isn't as helpful as it might seem."
• Don't publish a page for every fan-out query; Google treats that as scaled content abuse when the aim is to manipulate rankings or AI responses.
• For Google, skip llms.txt, chunking and special AI markup. Its guide says none of them is needed, though it still recommends structured data for rich results.

The short version

GEO vs SEO in four points, as of October 2026:

• Getting into AI answers starts with SEO work: Google retrieves the pages behind AI Overviews from its index using its core ranking systems, and ChatGPT search uses third-party search providers.
• GEO adds a few checks: whether AI search crawlers can reach the site, whether it is still included in Google's AI features, and whether other sites cover the brand on their own.
• Part of it can now be measured with the engines' own data: Google reports impressions in AI features and Bing reports citations, and the two aren't comparable.
• Be wary of anyone promising a fixed uplift: the 30-40% figures come from a 2023 test on a simulated engine, and one practitioner on Hacker News reported in June 2025 that in their runs the AI result set "often changed overnight", while a page-one ranking can last weeks.

For the crawler check, ScaletoTop's site audit scores a site on technical, content and GEO readiness, including whether AI search crawlers can reach it, and it can re-run on a schedule.