GEO: how to structure content so ChatGPT and other AIs cite it

How does an AI choose the passages it cites? (“extractable” content)

An answer engine doesn't read a page like a human: it looks for passages it can extract, summarise and attribute to a source. A clear passage that answers a precise question without needing the surrounding context is far easier to reuse than a long narrative paragraph. Structuring content for GEO is therefore mostly about making it "extractable".

How to structure content to get cited by AI: the 5 rules

Rule 1: give the answer in the first sentences

For each question posed by a subheading, the first sentence or two should contain the complete answer. The rest of the paragraph adds nuance, an example, the source. Many articles do the opposite: a bit of suspense, then the answer at the end. That works to keep a reader, but not to get cited.

Rule 2: phrase subheadings as real questions

Subheadings phrased like the questions people actually ask ("How long does it take to…?") match how users query an AI. A vague subheading ("Going further", "Our approach") tells an engine nothing about what the paragraph contains.

Rule 3: write self-contained blocks with a verifiable fact

Each block should make sense on its own: define the term it discusses, name the subject rather than saying "it" or "this", and contain at least one verifiable fact (a dated and sourced figure, a definition, a step). Comparison tables and numbered lists are particularly easy to pick up.

Rule 4: cite your sources and identify the author (E-E-A-T)

A figure quoted without a source is hard to reuse with confidence. So I state the origin and year of every statistic, and I identify the author with her role and experience. This serves readers as much as engines: it's the credibility principle (E-E-A-T) Google describes in its guidance on helpful, reliable content.

Rule 5: add a useful FAQ with real questions

An FAQ section with real questions and two-to-three-sentence answers gives you ready-to-cite blocks. FAQPage markup isn't required to appear in Google's AI features (Google Search Central): what matters is the quality of the answers. I write the FAQ from the questions clients genuinely ask, not from questions made up for SEO.

Which technical points to check to get cited: readable HTML, AI crawlers, llms.txt?

An AI can only cite what it can read. I check that the main text is present in the HTML (not loaded only by scripts), that robots.txt doesn't accidentally block the AI crawlers I want to let through, and that the page is fast and stable. The llms.txt file, proposed as a convention to guide AIs, isn't needed to appear in Google's AI features, and Google states you don't need to create any specific file or markup (Google Search Central): I treat it as optional, not a priority.

Table — classic content vs content structured for GEO

How to check that content is cited by ChatGPT and AI?

I regularly put the questions targeted by the content to several AI assistants and note whether the site is cited, and with what wording. Results vary from one answer to the next, so I look at the trend over several weeks rather than a one-off test. The full method is covered in the article on measuring AI visibility.

Keep reading: GEO, measuring AI visibility and tools

→ GEO: the new SEO for the ChatGPT era — why it matters now.

→ GEO and AI visibility, my services — the overview page on the topic.