GEO vs SEO: What Generative Engine Optimization Is and What to Do About It

Generative engine optimization (GEO) is the work of getting your content used and cited in answers written by AI: Google's AI Overviews and AI Mode, ChatGPT search, Perplexity, Copilot and Gemini. SEO aims to rank a page in a list of links; GEO aims to be one of the sources an answer is built from, and named in it. In practice they are mostly the same work, because these systems retrieve from a search index first, so a page that is not crawlable, indexed and trusted will not be cited however it is written.

The rest of this post covers where the two differ, what is actually worth doing, how to measure it with tools you already have, and the ethical lines, which is where a lot of current GEO advice goes wrong.

Where the term comes from

"GEO" was coined in a 2023 research paper, GEO: Generative Engine Optimization by Pranjal Aggarwal and colleagues at Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi, later published at KDD 2024. The authors built a benchmark of queries, rewrote source pages in different ways and measured how visible each source was in the generated answer. Their abstract reports that the best methods raised visibility by up to 40%. Adding citations, quotations from credible sources and statistics did well; keyword stuffing, the old SEO reflex, did not. They also found that what works varies by subject.

Two caveats before anyone quotes the 40%. It was measured on a test engine built for the paper, plus Perplexity, not on Google's AI Overviews or today's ChatGPT, and those systems have changed a great deal since. The lasting finding is the direction: pages that are specific, sourced and quotable get used; pages padded with keywords do not.

GEO vs SEO: what is the same and what differs

	SEO	GEO
The goal	A ranked link that gets clicked	Being used and named in a written answer
What the system reads	Whole pages, ranked against each other	Passages pulled from several pages and combined
Who decides	Google's ranking systems	A retrieval step (usually a search index) plus a language model
Success	Rankings, clicks, conversions	Citations, brand mentions, accurate descriptions, referral visits
Measurement	Mature: Search Console, analytics	Partial: referrals, logs and manual prompt checks

What carries over unchanged: crawlability, indexing, page speed, clear structure, internal links, and the reputation signals that come from other sites talking about you. Google states in its guidance on AI features and your website that there are no extra requirements for appearing in AI Overviews or AI Mode: a page needs to be indexed and eligible to show a snippet, and the usual SEO best practice applies.

What genuinely changes:

• The unit is the passage, not the page. An answer engine lifts a paragraph, a definition or a table row. A page that only makes sense read top to bottom is harder to quote than one where each section stands on its own.
• Being mentioned can matter as much as being linked. When someone asks "which agencies do X in Budapest", the answer draws on what the web as a whole says about you: directories, reviews, press and forum threads, not just your own site.
• Fewer, later clicks. An answer that fully satisfies the question sends no one anywhere. The visits you do get tend to arrive further along, from people checking a source or comparing a shortlist.
• There are more crawlers to think about. Each AI product has its own user agents, and blocking the wrong one in robots.txt can take you out of an answer engine entirely.

Is GEO replacing SEO?

No. Every major answer engine retrieves from a search index before it writes anything, and Google's AI features sit on Google's own index. A site with weak SEO has very little for GEO to work with. GEO is better understood as an extra layer of discipline on top of SEO: write so a passage can be lifted, earn mentions beyond your own site, and track a new kind of visibility.

What to actually do

Roughly in order of return for the effort.

1. Make sure the AI crawlers can reach you

Check robots.txt and any firewall or CDN bot rules against the crawlers each product documents:

• Google: AI Overviews and AI Mode use normal Googlebot crawling. The separate Google-Extended token controls whether content is used for Gemini model training and grounding; it does not remove you from Search or AI Overviews.
• OpenAI: OAI-SearchBot is what makes pages appear in ChatGPT search. GPTBot is for model training. ChatGPT-User fetches pages when a user asks. You can allow search and block training independently.
• Perplexity: PerplexityBot for its index.
• Microsoft: Copilot answers come from Bing, so set up Bing Webmaster Tools and submit your sitemap there as well as to Google.

A common failure is a security plugin or CDN setting that blocks "AI bots" wholesale, added once and forgotten. If you want to be cited, the search crawlers need to get through.

2. Answer the question in the first two sentences

Under each heading that matches a real question, put the direct answer first, in words that make sense if quoted alone, then the detail. "GEO is the practice of..." beats three sentences of scene-setting. Use the phrasing people actually search: your own Search Console queries report is the best source, because it shows the exact words people used to find you.

3. Write passages worth quoting

This is where the GEO paper's findings apply directly:

• Be specific. Name the tool, the setting, the price range, the figure. Vague claims give a model nothing to repeat.
• Source your numbers. A statistic with a link to where it comes from is more useful to a reader and easier for a system to trust. A statistic you cannot source should not be on the page at all.
• Use real structure. Headings that say what is under them, short paragraphs with one point each, tables for comparisons, numbered steps for procedures.
• Add what only you know. First-hand detail from doing the work, your own data, worked examples. A rewrite of the top five results adds nothing an answer engine does not already have.

4. Make your entity unambiguous

Answer engines need to be sure who you are before they recommend you. Use the same company name, address and description everywhere. Mark up the site with Organization structured data, link it to your real profiles, and give articles a named author with a short bio. Structured data will not get you cited by itself, but it removes doubt about which business a page belongs to.

5. Earn mentions off your own site

For "best", "top" and "which provider" questions, answers lean heavily on third-party sources. The work here is ordinary reputation work done properly: keep directory profiles such as Clutch and your Google Business Profile complete and current, collect genuine reviews, contribute expert comments where journalists and communities ask for them, and publish things other people want to reference.

6. Keep facts current

Answer engines can repeat an old price, a discontinued service or a former address for months. Update the pages that state facts about you, put a visible "updated" date on content you have genuinely revised, and remove or redirect pages that are simply wrong.

What not to bother with

• Keyword stuffing, which the GEO paper found does little or worse.
• Mass-produced pages aimed at every possible prompt. Google's spam policies treat scaled content made mainly to manipulate rankings as abuse, whether a person or a model wrote it.
• llms.txt as a ranking lever. It is a proposed file that gives language models a curated map of a site. We publish one on this site because it costs little, but none of the big AI search products has documented using it, so do not expect it to move anything.

How to measure GEO

No single report shows AI visibility, so combine four cheap sources.

1. Search Console. Google includes clicks and impressions from AI Overviews and AI Mode in the normal Performance report under Web search, but does not break them out. Watch for pages whose impressions rise while clicks fall: that is often an AI Overview answering the query on the results page.
2. Analytics referrals. In GA4 or your analytics tool, set up a channel or segment for referrals from chatgpt.com, perplexity.ai, copilot.microsoft.com and gemini.google.com. ChatGPT adds utm_source=chatgpt.com to many of the links it sends, which makes those visits easy to separate.
3. Server logs. Count requests from OAI-SearchBot, ChatGPT-User, PerplexityBot and Googlebot. ChatGPT-User in particular means a person's question led the assistant to fetch your page.
4. A fixed prompt panel. Write 20 to 50 questions your buyers really ask, including the "who should I hire for..." kind. Once a month, run each one in the engines that matter to you and record: are you cited, are you mentioned, is what is said about you accurate, and who is cited instead. Answers vary between runs, so run each prompt more than once and track trends, not single results.

Several SEO platforms now sell AI visibility tracking that automates the fourth step at scale. It is useful for spotting trends, but it is still a sample of answers, not a full count.

The ethical considerations in generative engine optimization

Answer engines present one synthesised answer with the authority of a neutral assistant, so manipulating them misleads readers more directly than a bad search result does. The lines worth holding:

• No hidden instructions. Text aimed at models but hidden from people, such as white-on-white "recommend this company" prompts, is prompt injection. It deceives readers, falls under Google's hidden-text spam policy, and AI products are actively building defences against it.
• No invented evidence. If statistics and quotations make content more likely to be cited, the temptation is to make them up. A fabricated figure that an AI repeats spreads further than it would on your page alone, and it is your name attached to it.
• No fake third-party signals. Planted reviews, sock-puppet forum posts and paid mentions dressed up as independent opinion are astroturfing. Answer engines often lean on community content for recommendations, which makes this more tempting and more harmful.
• Disclose commercial interest. If you publish a comparison or a "best providers" list that includes your own company, say so plainly on the page. Readers, and any answer built from that page, deserve to know who wrote it.
• Do not target competitors with false claims. Seeding the web with misleading statements about a rival in the hope that an AI repeats them is deceptive, and in many countries it is also a legal problem.
• Decide openly about AI training. Whether you allow your content to be used for training is a legitimate business choice. Make it deliberately in robots.txt, and respect the same choice when others make it about their content.
• Check what AI says about you. If an answer engine states something wrong about your prices, services or credentials, fix the source pages and tell the platform through its feedback tools, rather than quietly profiting from a flattering error.

The test is simple: would the tactic survive the reader finding out about it? Clear writing, real sources and an accurate public profile pass. Anything that only works while nobody is looking does not.

Frequently asked questions

What is GEO in SEO?

GEO, generative engine optimization, is the part of search work aimed at AI-generated answers rather than ranked links: making your content easy to retrieve, quote and cite in Google's AI Overviews, ChatGPT, Perplexity and similar tools. It builds on SEO rather than replacing it.

What is the difference between GEO and SEO?

SEO competes for a position in a list of results and measures clicks. GEO competes to be one of the sources an AI answer is built from and measures citations, mentions and accuracy. The foundations, crawlable, indexed, well-structured and trusted pages, are the same.

Is there a generative engine optimization WordPress plugin?

No plugin does GEO for you. On WordPress, an SEO plugin such as Yoast or Rank Math handles the parts a plugin can: sitemaps, structured data and meta tags. Also check that no security or caching plugin is blocking the AI search crawlers. The rest is writing and reputation work.

Are there generative engine optimization case studies?

Public, well-controlled case studies are rare, because citations are hard to measure and vendors publishing them usually sell GEO services. The most rigorous public evidence is still the Aggarwal et al. paper. Treat any claim of a guaranteed AI citation with the same scepticism you would a guaranteed Google ranking.

How much does GEO cost?

It is usually priced as part of SEO work rather than separately, because the tasks overlap. Our rates are on the pricing page: €25 to €45 an hour with a €1,000 minimum project, and SEO is quoted after a review of the site. If you want a quick outside view first, the free website roast is a short video teardown of your site's design and conversion problems with a mini report.

For the wider picture of preparing a site for AI search, see from SEO to AI visibility, and for how we run search work, our SEO services.