GEO: how to get cited by ChatGPT and AI

What is GEO?

GEO (Generative Engine Optimization) refers to all techniques aimed at optimizing a website's visibility in responses generated by artificial intelligence models. While traditional SEO aims to appear in Google's 10 blue links, GEO aims to be cited, referenced, or recommended by ChatGPT, Claude, Perplexity, Google Gemini, and other AI assistants.

The stakes are considerable: more and more users ask their questions directly to an AI model rather than a traditional search engine. According to recent estimates, AI-driven searches already represent 15 to 20% of discovery traffic for certain sectors. If your site isn't visible in these AI responses, you're losing a growing share of potential traffic.

GEO relies on technical signals that AI models use to understand, evaluate, and cite websites. These signals include JSON-LD structured data, the llms.txt file, quality meta descriptions, Open Graph tags, and crawl authorization for AI bots. TeckBlaze measures all these signals and calculates an overall GEO score for each audited site.

Unlike traditional SEO where Google's algorithm is a black box, GEO is more transparent: AI models primarily use textual content, structured data, and metadata to generate their responses. Optimizing for GEO simultaneously improves your traditional SEO.

The llms.txt file

The llms.txt is a text file placed at the root of your site (example.com/llms.txt) that provides AI models with structured information about your business and content. It's the equivalent of robots.txt for search engines, but optimized for LLMs (Large Language Models).

The llms.txt file should contain at minimum: your company or brand name, a concise description of your activity, and a list of the most important pages on your site with a brief description of each. AI models that support llms.txt use it as the primary source of information about your brand.

Here's an example llms.txt structure: start with a title line (# TeckBlaze), followed by a description (> Complete SEO audit tool with 55+ checks), then list your important pages with links and descriptions. The format is simple and readable by both humans and machines.

TeckBlaze checks for the llms.txt file during every audit. The absence of this file reduces your GEO score. Our engine also verifies that the file contains substantial information: company name, description, and references to important pages.

AI bots in robots.txt

Your robots.txt file controls which bots can explore your site. To be visible in AI responses, you must allow the bots of major AI models to crawl your content. Blocking these bots prevents AI models from indexing your content and thus from citing you.

The most important AI bots to know are: GPTBot (OpenAI's crawler for ChatGPT), ChatGPT-User (ChatGPT's user bot), ClaudeBot and anthropic-ai (Anthropic's crawlers for Claude), PerplexityBot (Perplexity's crawler), Google-Extended (the bot for Google's AI services like Gemini), and CCBot (Common Crawl's bot used by many models).

TeckBlaze analyzes your robots.txt and identifies each AI bot individually: allowed, blocked, or not mentioned. The report clearly indicates which bots have access to your content and recommends necessary changes. Blocking a specific AI bot can be a strategic choice (some companies block AI scraping to protect intellectual property), but for most sites, allowing AI bots is beneficial.

Also ensure your robots.txt references your XML sitemap with the Sitemap directive. This helps AI bots discover all your important pages in a structured manner.

JSON-LD and meta description for AI

JSON-LD structured data is the number one signal for AI response precision. AI models extract structured information (company name, author, date, content type) directly from JSON-LD to generate precise, well-attributed responses. A site with complete Organization, Article, and FAQPage schemas is significantly more likely to be cited.

The meta description plays a specific role in GEO: it is often used by AI models as a page content summary. A well-written meta description of 150-160 characters, containing action words and a clear value proposition, facilitates AI citation. TeckBlaze evaluates meta description quality for GEO by checking for action words and optimal length.

Open Graph tags (og:title, og:description, og:image, og:url, og:site_name) provide AI models with a structured title-description pair that facilitates understanding and citation of your content. TeckBlaze measures OG tag completeness as part of the page-level GEO score, representing 20% of the score.

GEO score: how TeckBlaze calculates it

TeckBlaze calculates a combined overall GEO score from site-level signals (40%) and page-level signals (60%). Site-level signals include: llms.txt file presence (+25 points), no AI bots blocked in robots.txt (+25 points), XML sitemap existence (+20 points), sitemap coverage relative to crawled pages (+10 points), and Sitemap directive in robots.txt (+10 points).

Page-level signals include: JSON-LD structured data presence and completeness (+30%), meta description quality (+20%), and Open Graph tag completeness (+20%). The remaining score (30%) is attributed to the overall quality of the page's textual content.

A GEO score above 70 indicates your site is well optimized for generative search engines. A score below 40 means you're missing significant AI visibility opportunities. The TeckBlaze audit report provides specific recommendations for each missing or insufficient signal.

Advanced GEO strategies

Beyond technical signals, content plays a crucial role in GEO. AI models favor content that clearly answers specific questions, is well-structured with H1-H6 headings, contains factual data (numbers, dates, statistics), and demonstrates subject expertise (E-E-A-T).

Create content in question-answer format (FAQ) because AI models are trained to answer questions. FAQ pages with FAQPage schema have a double advantage: rich snippets in Google and citations in AI responses. This blog page you're reading uses this strategy.

Finally, keep your content up to date. AI models tend to favor recent, current content. Publication and modification dates in your Article schema help models evaluate your content's freshness. TeckBlaze automatically detects pages without publication dates in their structured data.

FAQ