Why LLMs cite some sources and not others
Large language models learn from training texts but in their current operation also use real-time search: Perplexity, ChatGPT with browsing, Gemini. What determines whether they cite a source is: the perceived authority of the domain, the clarity and specificity of the answer, the format of the content and the presence of structured data that clearly identifies who the entity is and what it claims.
The key is not content volume but density of useful information per paragraph. A 600-word article with a direct answer in the first 3 sentences outperforms a 3,000-word article without clear structure in terms of citability.
The 7 GEO techniques that work
These are the levers with the greatest impact on AI engine visibility, ordered from highest to lowest return:
- Direct answer in the first 2-3 sentences. LLMs extract the most informative part of the text. If the article starts with storytelling or context, the useful answer won't be cited.
- FAQPage schema. Implement schema.org/FAQPage with explicit questions and answers. It's the format LLMs process best for extracting definitions.
- Well-defined Organization schema. Use schema.org/Organization so LLMs can identify who you are, what you do and what your authority on the topic is.
- Semantic lists and tables. Content in `
- ` and `` is more citable than dense paragraphs. Lists simulated with dashes in plain text don't provide the same signal.
- Question-answer format headlines. "What is X?" as an H2 followed by the direct definition in the first paragraph.
- Data and statistics with sources. Specific figures with attribution get cited more than generic claims, for a simple reason: a model can attribute "43% according to X", but it cannot attribute "many companies".
- Canonical and semantic URL. LLMs with browsing need to identify the source URL. Clean URLs without parameters and with keywords in the path.
GEO vs. SEO: what changes and what stays the same
GEO doesn't replace SEO, it extends it. Google ranking remains a priority because many LLMs with real-time search prioritize sources well-positioned in search engines. What changes is the content format: from optimizing for keywords and backlinks to optimizing for extraction and attribution.
Good GEO content has the properties of a good Google snippet, direct answer, concise, with data, plus explicit structured data and clear identification of the authoring entity.
How to measure whether your GEO strategy is working
Unlike SEO, there's no official LLM visibility index. The practical metrics are: regular manual tests on ChatGPT, Gemini and Perplexity with your business's key queries; brand mention monitoring on social listening platforms; and tracking referral traffic from AI platforms, which starts appearing in Google Analytics with the tag "chatgpt.com" or "perplexity.ai" as referrer.
At Elevatec we use a monthly test protocol: 15 target queries across 4 AI engines, recorded in a tracking sheet with date, exact query, engine and result. It's the closest thing to a rank tracker for GEO that exists today.
Frequently asked questions
Results in LLMs are slower than in SEO because they depend on re-indexing or the model's training cycle. With well-executed content strategies, the first results are typically seen within 2 to 4 months. Improvements in real-time search: Perplexity, ChatGPT with browsing, can be seen within weeks.
Both. The initial priority is usually optimizing existing content: adding FAQPage schema, reformulating opening paragraphs to give direct answers and adding specific data with attribution. Then a new content strategy reinforces the entity's authority on the topic.
GEO is especially powerful for small companies with specific expertise. A niche well-covered by a small company has a higher probability of being cited than generic content from a large brand. Specificity and topical authority matter more than company size.