Traditional search returns a ranked list of links. You optimize to get near the top of that list — and users decide whether to click. AI search is different. ChatGPT, Perplexity, and Gemini read multiple sources and synthesize a single answer. Either your content gets quoted in that answer, or it doesn't appear at all.
The decision of whether to cite your content is made by the AI model, not by a ranking algorithm. Understanding what drives that decision is the core of GEO.
Step 1: Make sure you can actually be crawled
This sounds obvious, but it's the most common failure point. If GPTBot, PerplexityBot, or Google-Extended are blocked in your robots.txt — or intercepted by a Cloudflare firewall rule — the AI can't read your content at all. No crawl, no citation.
Check your robots.txt at yourdomain.com/robots.txt and confirm none of the major AI crawlers are listed under a Disallow: / rule. If you're not sure, run a free GEO audit — it checks all 14 AI crawlers in about 45 seconds.
Step 2: Write content that's easy to quote
Research from Columbia University and IIT Delhi identified the content attributes that most reliably increase AI citation frequency. The pattern is consistent: AI engines prefer content that makes specific, self-contained, quotable claims.
Use statistics and concrete data
Vague statements get paraphrased or ignored. Specific data points get quoted directly. Compare:
- Weak: "AI search is growing rapidly and affecting how people find information."
- Strong: "Perplexity served over 500 million queries in 2024, up from roughly 50 million in 2023."
If you have original research, survey data, or proprietary metrics — publish them. That's the kind of content AI engines are most likely to cite by name.
Write quotable paragraphs
Dense, multi-topic paragraphs are difficult to extract cleanly. Aim for paragraphs that make one clear point and could stand alone as a quote. Short enough to cite, substantive enough to be worth citing.
Use first-person authority signals
"In our analysis of 500 audited sites…" carries more weight than "research suggests…". If you have direct experience or proprietary data, write from it explicitly. AI models are trained to prefer sources that demonstrate first-hand expertise.
Step 3: Establish your brand as a known entity
AI models don't just read your content — they cross-reference it against knowledge graphs. If your brand exists in Wikidata, has a well-structured LinkedIn company page, and appears consistently across the web, AI models have a stable entity to attach your content to. That makes citation with attribution far more likely.
Three concrete actions:
- Create a Wikidata entry for your organization if one doesn't exist. This is free and takes about 20 minutes.
- Add
sameAsto your Organization schema — link your website's JSON-LD to your Wikidata URL, LinkedIn, and other authoritative profiles. - Keep your LinkedIn company page complete — description, industry, website URL, and founding date all contribute to entity confidence.
Step 4: Add structured data (Schema.org)
Schema markup is machine-readable metadata that tells AI engines what type of content they're reading and who produced it. The most impactful schemas for AI citability:
Organization— establishes your brand identity, contact info, andsameAslinksArticleorBlogPosting— identifies authored content with a clear author, publication date, and topicFAQPage— Q&A content is extremely citation-friendly; AI engines frequently pull from FAQ schemas verbatimHowTo— step-by-step instructional content that AI engines can cite as a structured source
All schema should use JSON-LD format (a <script type="application/ld+json"> block in the page head). It's the format Google and most AI platforms prefer.
Step 5: Create an llms.txt file
llms.txt is a new standard (proposed by Jeremy Howard) that provides a concise, AI-readable summary of your site at yourdomain.com/llms.txt. Think of it as a robots.txt for AI models, but instead of access rules, it contains a structured description of what your site is about and which pages are most important.
Not all AI engines check for it yet, but adoption is growing — and it's a 10-minute setup that signals you're AI-ready.
What to prioritize
If you only have time for two changes: fix your crawler access (robots.txt) and add an Organization schema with a sameAs link to Wikidata. These two changes address the most common reasons sites are invisible to AI search.
To see exactly which signals are suppressing your citability, run a free GEO audit. You'll get a breakdown across all six dimensions in about 45 seconds.