SEO and visibility in language models (GEO/LLM)
More and more people put their questions to language models instead of searching. That changes how companies get found — and what's worth doing with your content.
Quick overview
GEO (LLM visibility) is about showing up when the answer is generated by a language model instead of in a classic search listing – and the foundation is the same: write for humans, structure for machines.
How to get started:
- Structure: clear, readable HTML and quotable answers that can be lifted out as a standalone paragraph.
- Schema: structured data (Organization, Article, FAQ) plus an llms.txt at the root.
- Substance: content that actually answers questions, with the same facts stated consistently everywhere.
Remember: classic SEO is not dead – content that is clear, true and machine-readable shows up in both Google and AI answers.
Classic SEO is about ranking in a list of links. When the answer is instead generated by a language model there is no list — the model summarises. This is usually called GEO (generative engine optimisation) or LLM visibility.
The difference is bigger than it sounds. In a search listing you compete to be clicked; in a generated answer you compete to be cited. The model chooses which sources it leans on and rephrases them into a single answer. If you're not in there, you simply don't exist in the answer — there's no "second page" to browse to. That's why the focus shifts from ranking high to being the clearest and most trustworthy source on the question someone is actually asking.
What still applies
The good news is that much of the foundation is the same. Models are trained on and retrieve from the same open web that search engines index.
- Clear, readable HTML — structure that both search engines and models can interpret.
- Substance — content that actually answers questions, not just keywords.
- Structured data — schema (Organization, Article, FAQ) that makes the content machine-readable.
- Clear answers — questions and answers (FAQ) that are easy to cite.
What's new
What's added builds on the same foundation but targets how a model reads and trusts content. A model prefers to cite what's easy to lift out on its own and what's consistent enough that it dares to reproduce it.
- Become quotable — write answers that can be lifted out as a paragraph and stand on their own, without the reader needing the rest of the page to understand them.
- llms.txt — a simple file at the root that describes the site and points out the most important pages for AI tools.
- Consistency — the same facts (names, figures, offers) everywhere, so the model doesn't get unsure about what's true and therefore picks a different source.
- Breadth — programmatically generated pages can cover many questions, as long as each page actually has substance and isn't just filler.
This is exactly the infrastructure we build — from technical SEO and structured data to content that is readable for language models.
Classic SEO vs GEO (visibility in language models)
A lot overlaps, but the focus differs. Classic SEO optimises for a placement in a list; GEO optimises for being cited when a model summarises the answer. The table sets the most important aspects against each other.
| Aspect | Classic SEO | GEO (LLM visibility) |
|---|---|---|
| Goal | Rank high in a list of links. | Get cited in a generated answer where there is no list. |
| Unit that shows up | The page and its title in the search result. | A lifted-out paragraph or a phrasing from the text. |
| Content form | Keywords and headings that match search intent. | Clear questions and answers that can be lifted out on their own. |
| Technical foundation | Readable HTML, sitemap, links. | The same foundation plus structured data (schema) and llms.txt. |
| Trust | Authority and inbound links. | Consistent company info — the same facts everywhere. |
How to become quotable by language models
This is a concrete checklist. None of it is magic — it's the same principles that make content good for humans, put into a system so a model can trust and reproduce it.
- Clear HTML structure — logical headings (one h1, clear h2s) and paragraphs a model can interpret without guessing.
- JSON-LD schema — mark up Organization, Article and FAQ so the content becomes machine-readable.
- FAQ with quotable answers — write each answer so it stands on its own and can be lifted out as a paragraph.
- Substance — content that actually answers the question, not just repeats keywords.
- Consistent company info — the same name, figures and offers everywhere, so the model doesn't get unsure about what's true.
- llms.txt at the root — a simple file that describes the site and points out the most important pages for AI tools.
Want to see what this looks like in practice? Explore our solutions, see cases and platforms we've built, or get three free suggestions.