Generative engine optimization: get cited by AI search

In short: GEO (generative engine optimization) is the work that makes a website readable and quotable by ChatGPT, Perplexity and Google's AI Overviews and AI Mode. It does not replace SEO, and Google treats it as SEO: you start from the same technical foundation (indexed pages, crawlers not blocked) and add original content with direct answers and sources, structured data that matches the text, and measurement based on citations and conversions as well as clicks. No special AI files or markup are needed.

Key takeaways

When Google shows an AI summary, people click traditional results less often: lost traffic is not always a ranking problem.
ChatGPT and Perplexity run their own crawlers: if robots.txt blocks them, your site stays out of their search results.
You measure it with Search Console, Bing Webmaster Tools and GA4, focusing on conversions rather than clicks alone.
Since 31 August 2026, Search Console shows your site's impressions in AI Overviews and AI Mode and lets you include or exclude the site from these features.
An llms.txt file does nothing for Google: Google Search ignores it, and there is no special schema.org markup for AI.

Why is organic traffic dropping when rankings hold?

Because more answers now appear right on the results page or in a chat, so people often stop clicking. A Pew Research Center study of 900 US adults (March 2025 data) found that when Google showed an AI summary, users clicked a traditional result on 8% of visits, against 15% when there was no summary. Only 1% clicked a link inside the summary itself.

For an online store, the hit lands mostly on informational queries ("how to choose", "difference between"); for a B2B company it lands on guides and blog posts, the very content that fed lead generation. It is the same shift described in the guide on the future of B2B e-commerce. The question to ask becomes: when a buyer asks an AI assistant, does it cite my site? If you are weighing outside help, see the guide on choosing an SEO consultant.

What is GEO (generative engine optimization)?

GEO, or generative engine optimization, is the practice of shaping content so that generative engines pick it up and cite it in their answers. It is also called SEO for AI or AI search optimization.

A generative engine is a search system that retrieves pages from the web, extracts the useful passages and summarizes them with a language model, listing its sources. ChatGPT with search, Perplexity, Copilot and Google's AI Overviews and AI Mode all work this way.

The term comes from a research paper presented at KDD 2024: certain changes to a text can raise a source's visibility in generated answers by up to 40%. Quotations, statistics and cited sources helped; in the authors' tests, keyword stuffing did not help and often made things worse.

Are AI Overviews and AI Mode live in Europe?

Yes. Google announced on 25 March 2025 that AI Overviews were rolling out to more European countries, including Germany, Spain, Italy, Austria, Poland, Portugal, Belgium, Ireland and Switzerland, for signed-in users aged 18 and over. AI Mode, the conversational search mode that lets people dig deeper with follow-up questions, reached Italy and the Italian language on 8 October 2025, as part of an expansion to nearly 50 new countries and territories.

For a European business this means that searches by local customers, not only English-language ones, can get a generated answer above the classic results: that applies to the shopper comparing two products and to the buyer looking for a supplier.

How is SEO for AI different from traditional SEO?

The foundation is the same; the unit of measure changes. SEO thinks in pages and rankings, GEO in passages and citations.

Aspect	Traditional SEO	GEO / SEO for AI
Goal	Rank the page among the top results	Be cited as a source in the answer
Unit of work	The whole page	The page and the passages quoted in the answer: definition, figure, table row
Type of query	Short keywords	Long, conversational questions split into sub-questions
Main signals	Relevance, links, page experience	The same, plus clarity, cited sources and brand trust
Metrics	Position, clicks, CTR	Citations, visits from assistants, leads and sales

Google says no special optimization is needed to appear in AI Overviews or AI Mode: the page must be indexed and eligible for a snippet. These systems do, however, use a technique called query fan-out, running several related searches across the subtopics of a question. A page that covers those sub-questions well gets more chances to be picked up.

One important caveat. In its guide to optimizing for generative AI search, Google says that from its perspective optimizing for these features is still SEO, and advises caution with advice and services sold as "GEO" or "AEO" (answer engine optimization). It also warns that no third-party tool has access to its internal ranking or AI systems. The label matters little: useful content and a clean technical foundation do.

How do ChatGPT, Perplexity and Google choose their sources?

The first check is technical: robots.txt must not block the search crawlers.

Assistant	How it finds pages	What to check
ChatGPT (search)	OAI-SearchBot crawler; ChatGPT-User opens pages when a user asks	Do not block OAI-SearchBot: opted-out sites are not shown in ChatGPT search answers
Perplexity	PerplexityBot crawler; Perplexity-User for user-requested visits	Allow PerplexityBot; a robots.txt change can take up to 24 hours
Google AI Overviews and AI Mode	Google Search index	Indexing, snippets allowed, structured data that matches the text
Copilot and Bing AI summaries	Bing index	Site verified in Bing Webmaster Tools, new pages submitted via IndexNow

Note that OpenAI's GPTBot collects data for model training, not search; blocking it does not cost you visibility in ChatGPT.

With Google the controls are different. robots.txt for Googlebot governs crawling for Search as a whole; to limit what is shown you use nosnippet, data-nosnippet, max-snippet or noindex. Since 31 August 2026 there is also the Search generative AI control in your Search Console property settings: by default the site is included; if you exclude it, your links and content stop appearing in AI Overviews, AI Mode and generative AI features in Discover, usually within a few days, with no effect on the rest of Search. Google-Extended, by contrast, limits the use of your content for model training and in some of Google's other AI systems, not your presence in results.

What kind of content do AI assistants cite?

Original, non-commodity content that answers the questions real customers ask. Google calls this the factor most likely to shape a site's presence in generative AI search over the long run, more than any other suggestion: your own point of view and first-hand experience, not a summary of what is already out there or of what an AI model could write on its own. The GEO study adds that quotations, statistics and cited sources raise visibility in generated answers.

What you do not need is to chop your text into tiny chunks or rewrite it "for AI". Google says its systems understand pages that cover several topics, as well as synonyms and different phrasings of a question, and that there is no ideal page length. Clear headings, well-organized paragraphs and relevant images or video help: the same things that help human readers.

For an online store

Product pages with specifications written as text (materials, dimensions, compatibility), not only in images.
Buying guides with a comparison table between models.
Shipping, returns and warranty terms stated on the page.
Product data that matches across the site, structured data and Merchant Center, which Google asks you to keep up to date.

For a B2B company

Service pages that answer "how much does it cost", "how long does it take" and "who is it for".
Datasheets in HTML as well as PDF, plus a glossary of industry terms.
Properly localized versions for each export market: see the guide to AI translation software.
A recognizable author, with role and expertise stated.

A hypothetical scenario: a distributor of pneumatic components with 2,000 SKUs. A buyer asks ChatGPT for the difference between an ISO 15552 cylinder and a compact cylinder. With a PDF-only catalog, the assistant cites a competitor or a forum; with an HTML page holding a definition and a comparison table, the distributor has a real chance of being cited. If you produce content with AI, the guide to AI in WordPress and other CMSs covers the limits to know.

Does structured data actually help?

Yes, but not as a shortcut. Structured data is, in Google's definition, a standardized format for providing information about a page and classifying its content; the recommended format is JSON-LD.

For AI Overviews, Google states there is no special schema.org markup. Structured data makes price, availability, author and date explicit and reduces ambiguity. One rule applies: everything in the markup must be visible on the page.

E-commerce: Product with price, currency and availability, BreadcrumbList, Organization.
B2B: Organization, Service or ProfessionalService, Article with author and date, FAQPage only where the questions really appear on the page.

Google also notes that structured data is not required for generative AI search and remains useful mainly for rich results. For products and local businesses, Merchant Center feeds and your Google Business Profile carry more weight: they can bring your products and services into both AI answers and other search results.

Does an llms.txt file help you get cited by AI?

Not on Google. llms.txt is a proposal published by Jeremy Howard in September 2024, and updated in 2026, for placing a Markdown file on a site with a short overview and links to the pages most useful to AI agents. Google Search Central states that Google Search, including its generative AI features, does not use llms.txt or any other "special" file or markup: creating one neither helps nor hurts.

It can make sense for other uses. According to the proposal's site, thousands of sites publish one, and OpenAI, Anthropic and Google (Gemini) have one for their own developer documentation. For a B2B company with technical documentation, APIs or a complex catalog it is a low-cost addition; for visibility on Google it is not a priority.

How do you measure visibility in AI answers?

With three free tools and some manual checks, looking at leads and sales more than clicks.

Google Search Console: since 31 August 2026, the Generative AI performance report shows how often links to your site appeared in AI Overviews and AI Mode, by page, country, device and date. Traffic from these features is also still included in the Performance report under the "Web" search type. Stable impressions and falling clicks on informational pages are the typical signal.
Bing Webmaster Tools: the AI Performance report, launched in public preview on 10 February 2026, counts citations of your pages in Copilot and Bing's AI summaries and shows the queries that retrieved them.
GA4: a channel group for visits from chatgpt.com, perplexity.ai, copilot.microsoft.com and gemini.google.com, with leads and orders from that channel.
Manual checks: once a month, ask the assistants 10 to 20 real customer questions and note which sites get cited.

Google says visitors arriving from AI Overviews tend to spend more time on the site, and suggests judging sales and sign-ups rather than clicks. To pick the right indicators, use the approach in the guide on measuring the value of AI.

What changes with AI agents browsing your site?

Beyond assistants that cite pages, there are AI agents: systems that carry out tasks on a user's behalf, such as comparing the specifications of two products or completing a booking. Google mentions them in its guide and points to web.dev's advice, which says an agent reads a site in three ways: screenshots of the page, the HTML, and the accessibility tree, the map of element roles and names that screen readers also use.

The recommendations match those for a good accessible site:

real HTML buttons and links rather than blocks made clickable with scripts;
form labels tied to their fields;
no transparent overlays on top of interactive elements;
a stable layout, with the "Add to cart" button in the same place across every category.

For a B2B store with price lists, configurators and quote request forms, this is worth checking together with accessibility. How these agents work from the company side is covered in the guide to AI browser automation.

How can you use AI search engines for market research?

The same tools work from the other side too, for finding information. An AI search engine takes questions in natural language and returns a summary with sources; some, like Google's AI Mode, break the question into several related searches. It is useful for a first map of the landscape, not for the final decision.

Use	Example question	What to check
Competitor analysis	"Which industrial valve distributors sell online in Germany, and what are their delivery terms?"	Open the cited sources: third-party pages may be out of date
Supplier search	"Which European manufacturers offer paper packaging for e-commerce shipping?"	Confirm certifications and minimum order quantities on the supplier's own site
Rules and obligations	"What does the AI Act change for a company running a chatbot on its website?"	Read the official text, not just the summary
Your own brand's visibility	The questions customers ask your sales team	Note which sites get cited and what they say

To bring them into your business systems there are APIs. OpenAI offers a web search tool in the Responses API, with answers that include source citations; Perplexity provides an Agent API with web-grounded answers and built-in citations, plus a Search API that returns ranked results as JSON with text extracted from the pages. An automation workflow can then, for example, prepare a weekly summary on competitors or prices for a person to review.

Two rules: always check the sources before using a figure in a decision or a quote, and keep confidential customer data out of your questions. To connect these tools to your CRM and ERP, follow the approach in the guide to business process automation.

First steps for your business

Check robots.txt: OAI-SearchBot, PerplexityBot, Googlebot and Bingbot must be able to read your public pages.
In Search Console, confirm the site is included in the Search generative AI control and open the Generative AI performance report to see which pages already appear.
Verify the site in Bing Webmaster Tools and enable IndexNow to report new or updated pages right away.
Pick the 10 informational pages with the most impressions in Search Console and rewrite them: a direct answer up front, a definition, a table, cited sources.
Align structured data with the visible text and validate it with Google's tools.
Set up an "AI assistants" channel in GA4 and a monthly citation check against a fixed list of questions.

This is where my website and e-commerce projects start; more guides are in the websites and e-commerce section of the blog.

Mistakes to avoid

Blocking every AI bot at once: you drop out of ChatGPT and Perplexity search, not just training.
Stuffing text with keywords: the GEO study found it ineffective, and in several tests it reduced visibility.
Publishing dozens of generic AI-written posts: interchangeable content does not get cited.
Creating a page for every variant of a question: Google warns that producing content for every search or fan-out variant in order to manipulate answers violates its scaled content abuse policy.
Chasing inauthentic mentions: Google's AI features do reflect what is said about products and services across the web, but they rely on the same ranking and anti-spam systems as Search; Google says artificial mentions help less than they seem to.
Marking up what is not on the page: hidden FAQs or reviews that are not displayed go against Google's guidelines.
Measuring clicks only: AI summaries reduce clicks for everyone; leads and orders are what count.

Frequently asked questions (FAQ)

Does GEO replace SEO?

No. It builds on the same technical foundation: a page that Google or Bing cannot index will not be cited by any assistant. It adds work on the shape of your content and a way of measuring based on citations and conversions. Google itself treats optimizing for generative AI search as part of SEO.

How can I tell whether ChatGPT or Copilot cite my website?

For Copilot, the AI Performance report in Bing Webmaster Tools counts citations of your pages. For Google, since 31 August 2026 the Generative AI performance report in Search Console shows impressions in AI Overviews and AI Mode. For ChatGPT and Perplexity, track visits from chatgpt.com and perplexity.ai in GA4 and run regular checks with the questions your customers ask.

Should I block AI bots in robots.txt?

Only the training crawlers, if you choose to: GPTBot can be excluded without losing visibility, and the same goes for Google-Extended with respect to Google Search. OAI-SearchBot and PerplexityBot power search, and blocking them keeps your site out of ChatGPT and Perplexity search results.

Can I opt my site out of AI Overviews?

Yes. Since 31 August 2026 Search Console has a control to include or exclude your site from Search generative AI features: the exclusion usually takes effect within a few days and does not affect the rest of Search. To hold back only parts of a page, use data-nosnippet. Opting out, however, means no impressions and no traffic from these features.

Does an llms.txt file help AI find my site?

Not on Google: Google Search, including AI Overviews and AI Mode, ignores it, so having one neither helps nor hurts. It can serve other agents and services that read it, especially for technical documentation and APIs. It is a low-cost addition, not a priority.

How long does it take to see results?

There are no guaranteed timelines. OpenAI and Perplexity say a robots.txt change takes about 24 hours to be picked up, while Google has no fixed timeline: it depends on when the page is next crawled. Start with pages that already earn impressions.

Does GEO work for an online store with thousands of products?

Yes, by working on page templates rather than individual products. Improving the product template, attributes and structured data lifts the whole catalog at once; buying guides are written by hand, starting from your best-selling categories. Keep your Merchant Center feed in line with the site.

Sources

If you want to know how your site shows up today in ChatGPT, Perplexity and Google answers, book a free diagnosis.