Audience measurement — With your consent, LeadCo and Google Analytics use cookies to count visits. Declining changes nothing about your visit. You can change your mind at any time via “Manage cookies” at the bottom of the page. Learn more

AEO and GEO: how to get cited by ChatGPT, Perplexity and Google AI

How AI assistants pick their sources, which robots to allow, what llms.txt is really worth and how to write pages that get cited: a method for small businesses.

Published
Updated
9 min

Key points

  • AEO and GEO cover the work that makes your pages readable, trustworthy and citable by AI assistants: Google's AI Overviews and AI Mode, ChatGPT, Perplexity, Copilot.
  • These assistants rely on search indexes (Google's, Bing's or their own): without solid classic SEO, there is no citation.
  • Allow their search robots in your robots.txt; blocking training robots is a separate choice, with no effect on search answers.
  • Write direct, dated and sourced answers, and keep the same information on your website and your profiles.
AEO (answer engine optimisation) and GEO (generative engine optimisation) cover what helps your pages get picked up and cited in AI assistants' answers, on top of appearing in regular results. In Belgium, Google has shown AI Overviews above some results since March 2025, and its AI Mode, a conversational search, has been available in French and Dutch since October 2025. ChatGPT, Perplexity, Copilot and Claude also answer your future clients' questions, with links to their sources.

The good news: this is not a new profession, but an extra layer on top of good SEO.

AEO, GEO, SEO: what's the difference?

  • SEO: getting your pages into search engine results.
  • AEO: making your page the answer, whether it is read out by a voice assistant, shown as a snippet or written up by an AI.
  • GEO: being cited as a source in a generated answer. The term comes from an academic paper published in November 2023 (Aggarwal et al., "GEO: Generative Engine Optimization"), presented at the KDD conference in 2024.

In practice the three overlap heavily, and SEO remains the foundation.

How AI assistants find their sources

AssistantWhere its sources come fromRobot that matters for search
Google AI Overviews and AI ModeGoogle's indexGooglebot
Gemini (app)Google's indexGooglebot; Google-Extended decides whether your pages are used for Gemini
ChatGPT (search)its own index and partners, including BingOAI-SearchBot
Perplexityits own indexPerplexityBot
Claudeits own web searchClaude-SearchBot
Microsoft CopilotBing's indexBingbot
There are also training robots, which collect pages to train future models: GPTBot (OpenAI), ClaudeBot (Anthropic), Google-Extended (Google), Applebot-Extended (Apple), CCBot (Common Crawl). And there are agents acting on a user's request (ChatGPT-User, Perplexity-User, Claude-User), when someone asks the assistant to open a specific page. OpenAI notes that robots.txt rules may not apply to those visits, since a person triggers them.
Above all, remember this: blocking GPTBot does not remove you from ChatGPT, and blocking Google-Extended does not remove you from Google's AI Overviews or AI Mode, which depend on Googlebot.

Your robots.txt: two simple set-ups

For a small business that wants to be found everywhere, the simplest option is a single group for all robots. That is what leadco.be does:
User-agent: *
Allow: /
Disallow: /admin

Sitemap: https://www.example.be/sitemap.xml

If you want to be cited in answers without your texts being used to train models, add a group that refuses the training robots:

User-agent: *
Allow: /
Disallow: /admin

User-agent: GPTBot
User-agent: ClaudeBot
User-agent: Google-Extended
User-agent: Applebot-Extended
User-agent: CCBot
Disallow: /

Sitemap: https://www.example.be/sitemap.xml
Watch out for a common trap: a robot that finds a group with its own name ignores the general group. If you create a group for a robot you allow, copy into it the Disallow lines that protect your private areas. And note that Google-Extended also controls whether your pages are used in Gemini app answers: blocking it is a real choice, not a formality.

llms.txt: useful, but not magic

The /llms.txt file, proposed in September 2024 by Jeremy Howard (Answer.AI), summarises a website in Markdown for language models: a short description and links to the important pages (specification at llmstxt.org). Many documentation sites have adopted it.

Its status in November 2026 deserves an honest description: Google has said it does not use it for its search engine, and no major assistant has confirmed using it to choose sources. It is mostly useful to agents and tools that read it when asked to. It costs little to create, but it is not a visibility lever in itself.

LeadCo publishes one, generated from the same data as the website: leadco.be/llms.txt. An example for a fictional business:
# Example Bakery

> Artisan bakery in Ixelles (Brussels): sourdough bread,
> pastries and cakes to order. Open Tuesday to Sunday.

## Main pages

- [Our bread](https://www.example.be/en/bread): the range and baking days
- [Order a cake](https://www.example.be/en/order): lead times and form
- [Practical info](https://www.example.be/en/contact): address, hours, access

Writing pages that AI assistants cite

The GEO study tested several ways of rewriting pages. Adding cited sources, attributed quotations and statistics increased visibility in generated answers, by up to 40% in their tests; stuffing pages with keywords did not help. In practice:
  • A direct answer up front: one or two sentences that answer the question, before the details.
  • Self-contained sections: each heading covers one question and can be read on its own.
  • Verifiable facts: dates, figures, names of laws or tools, links to official sources.
  • Questions and answers based on your clients' real questions.
  • Tables and lists for comparisons and steps.
  • A visible update date, on pages that are genuinely updated.
  • The same information everywhere: identical services, area, hours and contact details on your website, your Google profile and your directory listings.

Structured data: what still matters in 2026

The schema.org vocabulary describes your pages to search engines and assistants. Some uses have changed in recent years:

TypeWhat it does today
Organization, LocalBusiness (ProfessionalService, Restaurant…)your business identity: name, address, hours, logo, links to your profiles. The foundation.
WebSitethe site name shown by Google. The search box in results (sitelinks search box) has been gone since November 2024.
BreadcrumbListthe breadcrumb trail, still shown on desktop
Article, BlogPostingauthor, publication and update dates
Product, Offerprice, availability and reviews on product pages
FAQPageread by search engines and AIs, but since 2023 Google only shows these rich results for official government and health sites
HowTono longer shown by Google since 2023
To check your markup: validator.schema.org, the Rich Results Test and URL Inspection in Search Console.

E-E-A-T: show who is speaking

Since December 2022, Google's search quality rater guidelines talk about experience, expertise, authoritativeness and trustworthiness (E-E-A-T). AI assistants, which rely on the same indexes, favour the same signals. For a small business:
  • a clearly identified company: an "About" page, enterprise number, real address;
  • articles signed (by a person or by the company) and dated;
  • proof: past work, customer reviews, concrete examples;
  • sources cited for figures and rules;
  • mentions elsewhere: press, partners, trade bodies.

What works, assistant by assistant

Google: AI Overviews and AI Mode

Everything starts with classic SEO: the pages cited are pages Google can crawl, understand and trust. AI Mode often breaks a question into several sub-questions before answering: a page that covers a topic and its related questions well stands a better chance of being used. Clicks and impressions from these answers are counted in the Search Console Performance report.

ChatGPT

Let OAI-SearchBot in, and look after your presence on Bing, one of its sources: register in Bing Webmaster Tools and send IndexNow notifications. Links clicked from ChatGPT usually carry the utm_source=chatgpt.com parameter.

Perplexity

Let PerplexityBot in. Perplexity readily cites recent, dated pages with figures.

Microsoft Copilot

Copilot relies on Bing: Bing Webmaster Tools, plus Bing Places for a local business.

Claude

Let Claude-SearchBot in, Anthropic's search robot (ClaudeBot is used for training).

IndexNow: tell Bing without waiting

IndexNow, launched in 2021 by Microsoft Bing and Yandex, lets you instantly notify participating search engines (Bing, Yandex, Seznam, Naver…) of a new or updated page. Google does not use it. Most CMSs have a plugin or a setting to turn it on; leadco.be uses it to flag new and updated pages.

Measuring your visibility in AI assistants

  • Google Analytics: visits from chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com or claude.ai show up as referrals. Create a custom channel group "AI assistants" with a rule such as chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai. See our Google Analytics guide.
  • Search Console: appearances in AI Overviews and AI Mode are included in the Performance report.
  • Manual checks every month: ask the assistants your clients' questions ("which plumber in Schaerbeek?", "how much does a showcase website cost in Belgium?") and note the sources cited.
  • Specialised tools: they exist, often paid, to track citations at scale.

An 8-step action plan

  1. Check robots.txt: search robots allowed, a deliberate choice for training robots, sitemap declared.
  2. Register the site in Google Search Console and Bing Webmaster Tools, and turn on IndexNow.
  3. Add basic structured data: Organization or LocalBusiness, BlogPosting for articles.
  4. Rewrite key pages: a direct answer up front, self-contained sections, frequently asked questions.
  5. Add proof: dates, sources, past work, reviews.
  6. Align your information across your website, Google profile, Bing Places, Apple and directories.
  7. Publish an llms.txt if you like, without expecting miracles.
  8. Measure: an "AI assistants" channel in Google Analytics and monthly manual checks.
Our SEO service covers these steps. For the basics, start with what SEO is.

Frequently asked questions

Are AEO and GEO the same thing?

Almost. AEO targets any kind of direct answer (Google snippets, voice assistants, AI). GEO specifically targets answers written by generative AI. The methods overlap heavily.

If I block GPTBot, do I disappear from ChatGPT?

No. GPTBot collects pages to train OpenAI's models. For ChatGPT search, OAI-SearchBot is what counts: you can block one and allow the other.

Is an llms.txt file required?

No. It is an optional convention, which Google says it does not use for search. It can help agents and tools that read it, but it replaces neither SEO nor structured data.

How long before I get cited?

It depends on the assistant and the competition. Engines that search in real time (ChatGPT, Perplexity) can cite a page as soon as it is indexed and relevant. For Google, timelines follow classic SEO: several months on contested topics.

Is classic SEO becoming useless?

Quite the opposite. Assistants rely on search indexes and on criteria close to Google's: accessibility, structure, quality, trustworthiness. A site that is invisible in results is almost always invisible in answers.

How can I tell whether an AI cites my website?

Look at referrals in Google Analytics (chatgpt.com, perplexity.ai…) and regularly ask assistants your clients' questions, noting the sources they show.

Search engine optimisation (SEO)

Technical audit, keyword strategy, on-page optimisation and local SEO to show up on Google in Brussels and across Belgium.

See the service
LeadCo

LeadCo

Web and digital agency in Brussels

We build websites and online shops for freelancers and small businesses, then make them visible on Google and in AI assistants. Based in Schaerbeek, we work across Belgium.

Discover our services