What is generative engine optimization (GEO)? How it differs from SEO and AEO

Generative engine optimization (GEO) is the set of practices that help the AI systems that write answers retrieve your content, understand it and cite it as a source. Where SEO competes for a position in a list of results, GEO competes for a place inside the answer itself.

The term has an academic origin: it comes from the paper “GEO: Generative Engine Optimization” by Aggarwal and fellow researchers at Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi, posted on arXiv in November 2023 and presented at the KDD 2024 conference, held in Barcelona in August 2024.

So it is not a Google or OpenAI product, and no platform has published a standard by that name. If you want the service rather than the theory, it is set out on the page for our GEO service.

Generative engines do not answer from memory alone: when they need current information they search, select a handful of pages and write the answer from them. This process is called grounding, or retrieval-augmented generation (RAG). It is how generative engine optimization works in practice: your site can only be cited if it can first be found and read.

On Google, AI Overviews and AI Mode draw on the Search index and on query fan-out: the system breaks the question into several related searches, runs them at once and combines what it finds. A site that covers a subject in depth has more ways in than one that publishes a single thin page.

Beyond Google, each assistant runs its own crawlers; some feed search and others feed training:

• OpenAI: OAI-SearchBot surfaces sites in ChatGPT search and GPTBot collects training data; both respect robots.txt.
• Anthropic: ClaudeBot (training), Claude-SearchBot (search) and Claude-User (user-requested visits) honour robots.txt.
• Perplexity: PerplexityBot indexes for its search engine, is not used to train models and respects robots.txt.

Google's position is that there is no special optimisation for its AI answers. Its guide to AI search, published in May 2026 and updated in July, says SEO best practices remain foundational, because AI Overviews and AI Mode are rooted in its core ranking and quality systems.

There are no additional requirements: to be eligible, a page must be indexed and able to show with a snippet, with no special files or markup. The guide acknowledges that the terms AEO and GEO are common, and warns that many of the hacks in circulation are not effective for its search engine.

There are two ways to read this. If your technical SEO and content fall short, no amount of GEO will make up for it on Google. And the guide speaks for Google only: ChatGPT, Claude and Perplexity use their own crawlers for search, so GEO applies the same discipline to more surfaces; it is not a secret formula.

SEO, GEO and AEO share a foundation and differ in their goal. SEO aims to rank pages in search results; AEO (answer engine optimization) aims to make your content the direct answer to a question; GEO aims to have generative models cite and recommend you when they write theirs.

A word of caution on AEO: since 7 May 2026, Google no longer shows FAQ rich results. Well-answered questions are still useful; what has gone is the visual reward. The detail is on the page for our AEO service.

At SERMAN we split the work like this: GEO is the umbrella for visibility in AI answers; we handle Google's surfaces (AI Overviews, AI Mode and featured snippets) in our AEO service, and conversational assistants such as ChatGPT, Gemini, Claude, Perplexity and Copilot in our GEO service.

Compared by goal, surfaces, signals and metrics, they look like this:

• SEO. Goal: your pages appear and earn clicks. Surfaces: organic results and Maps. Signals: crawling, helpful content and earned authority. Metrics: impressions, positions, clicks and conversions.
• AEO. Goal: to be the direct answer to a specific question. Surfaces: featured snippets, voice assistants and answer blocks. Signals: real questions and clear answers up front. Metrics: presence in those blocks.
• GEO. Goal: AI cites or recommends you. Surfaces: AI Overviews, AI Mode, ChatGPT, Gemini, Claude, Perplexity and Copilot. Signals: non-commodity content, crawler access, a consistent brand and trusted mentions. Metrics: impressions in AI features and citations.

In Spain, Google's AI search is already live. AI Overviews launched on 25 March 2025 in Spanish and English, for signed-in adult users, and AI Mode has been available in Spanish since October 2025 as an optional tab: the classic results view remains the default.

When a written answer sits at the top, there is less reason to click. According to Pew Research Center (July 2025, using March 2025 browsing data from a panel of 900 US adults), users clicked a traditional result on 8% of visits where an AI summary appeared, against 15% without one. Google disputed the methodology, and the data is from the US.

According to Bain & Company (April 2026), 22% of Spanish consumers say they have replaced search engines with generative AI most of the time. Even so, Google remains by far the dominant search engine in Spain: GEO is added to SEO; it does not replace it.

For a brand, the question is no longer only where it ranks, but whether it is part of the answer. In sectors where customers compare at length before deciding, such as hospitality and tourism or professional services, that presence counts before the first visit.

Four levers account for the useful work in GEO: content that is not generic, AI crawler access, a consistent brand entity and genuine mentions in trusted sources. None of them is a trick; all of them take sustained work.

Structure supports all of them: blocks that open with the answer, headings that ask the real question and facts with a date and a source. The four levers, one by one:

• Non-commodity content. Google's guide names it as the most influential long-term factor in AI search: an original perspective, first-hand experience and expert judgement. This is the territory of content marketing.
• Crawler access. Check robots.txt, your CDN and your firewall: since 1 July 2025, Cloudflare has blocked known AI crawlers by default and asks new domains to choose whether to allow them. This check belongs to technical SEO.
• A consistent brand entity. Name, activity, location and services should match across your site, your business profile and the sites that talk about you. According to Google, Business Profile and Merchant Center feeds help you appear in its AI answers too.
• Genuine mentions. Media, trade publications and real reviews are the kind of source AI answers draw on. They are earned through digital PR and citable content, not by chasing inauthentic mentions.

Much of what is sold as GEO has nothing behind it. Google has published its own list of things you do not need to do for its search engine, and some tactics can breach its spam policies, which since May 2026 also apply to generative AI responses in Search. These are the most repeated myths:

• “llms.txt gets you ranked in Google's AI.” No. Google Search does not use it, and in June 2026 the company clarified that it affects visibility neither positively nor negatively. You can publish one in case other systems read it.
• “You have to chunk content for AI.” Chunking, and rewriting copy purely for models, are on Google's list of unnecessary practices.
• “One page for every variation of the question.” Creating pages for each query variation mainly to manipulate rankings or AI responses breaches the scaled content abuse policy.
• “Blocking Google-Extended takes you out of AI Overviews.” No. Google-Extended controls whether your content is used to train Gemini models, not your presence in Search. To limit AI Overviews and AI Mode there are snippet controls, noindex and a Search Console setting that excludes your content from the AI features; according to Google, using it is not a ranking signal for the rest of Search.
• “We guarantee you will appear in ChatGPT.” No platform offers guaranteed placement. What can be done is to improve eligibility and visibility, and to measure both.

AI visibility can now be measured with official tools, though only in part. Search Console and Bing Webmaster Tools report when your pages appear or are cited in their generative answers; for the other assistants you have to combine further signals. Three sources cover the essentials:

• Search Console. The generative AI performance reports, available to every site since 31 August 2026, show impressions, pages, countries and devices for AI Overviews and AI Mode. They do not include clicks or CTR.
• Bing Webmaster Tools. The AI Performance report, in public preview since February 2026, shows citations of your pages in Microsoft Copilot and in Bing's AI summaries.
• Key query reviews. A fixed set of questions your customer would ask, checked regularly in each assistant and each language, to see whether you appear and who is cited in your place.

AI answers are ceasing to be purely organic and purely informational: ads are arriving, and so are agents that act on the user's behalf. It pays to separate what is live in Spain from what is not.

Agents first: Google's guide already describes how they may read a site's screenshots, DOM and accessibility tree. Purchases made by an agent are still at an early stage in Spain, so the sensible move is to prepare your data: complete product pages, up-to-date feeds and an accessible site. We develop this in our agent-ready web and commerce service and in the article on agentic commerce.

Ads, where they exist, are a separate channel, explained under ChatGPT ads and AI search advertising. As of October 2026, this is the picture:

• ChatGPT. It has shown ads in Spain since 24 August 2026, on the Free and Go plans only. In the European Economic Area they are contextual, not personalised.
• Google. In Spain, ads can appear above or below an AI Overview. Inside it they run only in twelve English-language markets, with neither Spain nor the EU among them, and in AI Mode they are still in testing in the United States.
• Perplexity, Claude and Gemini. These are not paid channels: Perplexity wound down its ads programme (as reported in February 2026), Anthropic pledged to keep Claude ad-free and the Gemini app carried no ads as of mid-2026.

For a business in Spain, GEO means checking three things: that AI can read your site, that it describes you accurately and that it cites you when someone asks about what you sell. It does not require rebuilding everything, only putting what you have in order and filling the gaps, starting with search engine optimisation (SEO).

Language matters. Assistants tend to answer in the language of the question, and an international client asks in English or German. If your content exists only in Spanish, or as a machine translation, you compete at a disadvantage. In markets such as Mallorca, the English version is not an extra. Five steps to start with:

• Ask the assistants what your customer would ask, in each language, and note who appears.
• Check that robots.txt, your CDN and your firewall are not blocking AI search crawlers.
• Review the AI reports in Search Console and Bing Webmaster Tools to set a baseline.
• Turn what only you know into content with clear answers.
• Measure every month: the trend matters more than any single answer.