What technical GEO measures should you implement to be cited by AIs?

Which GEO technical optimizations are essential?

Allow AI crawlers to access the site

If AI crawlers cannot explore your site, your pages cannot be included in their real-time answers. Not all these bots serve the same purpose: search crawlers, like OAI-SearchBot (ChatGPT), PerplexityBot, or Claude-SearchBot, feed answers and citations, while others, like GPTBot or ClaudeBot, collect data for model training. To be cited, you must first allow search crawlers through.

Check that your robots.txt file does not block them, then review your firewall and CDN rules: some anti-AI settings block these bots by default and may even block Googlebot. For a comprehensive guide on this topic, see our article on AI crawlers.

Serve content readable without JavaScript

Several AI crawlers do not execute JavaScript, or execute it only minimally. If your main content only appears after asynchronous loading or client-side rendering, it may remain invisible to these bots. The raw HTML must contain the core of your text, headings, and structural information. Sites built in React, Vue, or Angular should favor server-side rendering (SSR) or static prerendering. This requirement aligns with classic SEO, but it is even more crucial for GEO: the bots feeding ChatGPT or Perplexity do not all render pages like a browser.

Mark up your pages with JSON-LD structured data

Structured data helps machines understand your content. It does not guarantee a citation, but it helps AIs identify the page type, author, publication date, or mentioned entities. Here are the priority types:

• Organization: to present your company, logo, and contact details
• Article: for editorial content, with author, publication date, and main image
• FAQPage: for Q&A pages, similar to AI answer formats
• Product: for product pages, with price, availability, and reviews
• Person: for author pages or biographies

The direct impact of these types on citations is hard to isolate, and no public study reliably quantifies it to date. Still, they remain a low-cost best practice that complements your content work: clean markup does not replace content optimization for LLMs.

Which GEO technical optimizations provide real added value?

Clean semantic HTML

Well-structured HTML code helps AIs understand your content hierarchy. Use heading tags (h1, h2, h3) in order, without skipping levels. Wrap your articles with the article tag and your thematic sections with section. Tables should use table, thead, tbody, and td tags, not be inserted as images. Lists should use ul, ol, and li. These may seem basic, but many sites still neglect them, even though they make it easier to extract information for answers.

Markdown versions of pages

Some AI agents and tools process Markdown more easily than HTML, which is heavier with code. Offering a Markdown version of your main pages can therefore facilitate their processing. This practice is still marginal, and its impact on citations is unproven. It is mainly worthwhile if you already produce your content in Markdown before converting it to HTML. For most sites, clean HTML is enough.

Healthy crawl architecture

AI crawlers, like those of search engines, follow internal links and explore your site based on depth. An up-to-date XML sitemap guides them to your important pages. Consistent canonical tags prevent dilution between duplicate versions. Fast server response times allow bots to crawl more pages per visit. Finally, your key content should be accessible within a few clicks from the homepage.

llms.txt, agent standards: overrated GEO techniques?

llms.txt: a promise without proof

The llms.txt file is often presented as a “robots.txt for ChatGPT.” This comparison is misleading. Robots.txt controls crawler access, whereas llms.txt is an editorial table of contents: it lists your important pages and their content for agents and LLMs. It is a community convention, not an official standard, and there is no public evidence yet that it improves citations. The current version is documented on llmstxt.org. Implementing it is low effort, but do not expect measurable results.

New agent-oriented standards

Several initiatives are emerging to facilitate interaction between websites and autonomous agents: content negotiation protocols, standardized API formats, and specific metadata. These standards are still experimental, and none has reached sufficient adoption to justify prioritizing investment. Monitor their development, but do not make them a top priority.

False good ideas

Some practices circulate without foundation. Creating hidden pages reserved for AI crawlers is cloaking and exposes you to search engine penalties. Adding meta tags “for AIs” without official documentation is pointless. Writing content hidden from humans but visible to bots is a risky technique with no proven benefit. Focus on recognized optimizations instead.

How to check the proper implementation of your GEO technical optimizations?

• Check for AI crawler visits in your server logs: look for user-agents such as OAI-SearchBot, ChatGPT-User, PerplexityBot, Claude-SearchBot, and GPTBot, and analyze how often they visit.
• Test HTML rendering without JavaScript: disable JavaScript in your browser or fetch the raw source code (using curl, for example) to ensure the main content appears.
• Validate your structured data using Google’s Rich Results Test or the Schema.org validator.
• Check your CDN and firewall settings: make sure no rule blocks search crawlers or Googlebot.
• Monitor server response times and XML sitemap availability to ensure smooth crawling.

These checks verify your technical setup. To assess your visibility and areas for improvement more broadly, a GEO audit reviews all the relevant factors.

The prompt tracking feature of Search AI then lets you check whether your citations increase after these adjustments.