GEO for Deep-Tech Founders: How to Get Cited by LLMs

When someone asks an answer engine about your field, it may summarize other people's research and link to their sources. If you publish research, you want readers to find your original work and check what it actually says. Generative Engine Optimization (GEO) is the practice of preparing that work for discovery and citation by systems such as ChatGPT, Claude, Perplexity, and AI Overviews.

What a citation needs

Search engine optimization (SEO) aims to help a page appear in search results. GEO focuses on whether an answer engine can find and cite a claim from that page. Both benefit from readable HTML, clear structure, and identifiable authorship. A citation also needs a passage that stands on its own and points to evidence.

Prepare the source

Define an unfamiliar term near the top of the page. For epistemic memory, the explanation should tell a new reader what the system records and why that helps answer a question. A paragraph that needs no surrounding explanation is easier to quote accurately.

A question-and-answer section is useful when readers arrive with a specific question. Each answer should make sense on its own. FAQ structured data can describe that section to software; it does not guarantee that an engine will cite it.

Use tables for comparisons and parameters, and numbered lists for procedures. Keep units, conditions, and limitations beside the numbers so a copied passage carries them with it.

State the scope of every result. The OIDA pilot used about 42 times fewer input tokens for structured retrieval on ClearPath, while full context scored higher for overall answer quality. Publishing only the token reduction would leave the reader with an incomplete account of the study.

Use a consistent author name, affiliation, and set of links. Structured data such as Person, Organization, and ScholarlyArticle identifies who wrote a page and which work it describes. These records help readers and software distinguish the source from a summary of it.

An llms.txt file can provide a short map of important pages, with a description for each. Treat it as an additional index; support varies, and the pages still need to be accessible through the website.

Publish something readers can check

A dataset, benchmark, or paper gives a reader something to inspect. An explainer can make the methods easier to understand; a FAQ can answer common questions. Link both back to the original work, as in this article on epistemic infrastructure.

None of these changes guarantees a citation. They make the source easier to find, understand, and check.