What gets cited in AI answers: a checklist for publishers

AI search engines and assistants answer questions by retrieving passages from the web and synthesising them, often with citations. Being one of those citations is becoming as important as ranking in a list of links. There is no secret formula, and anyone who promises one is guessing. But the systems’ design and the research so far point to practical steps that help.

Use this checklist to review a page or a whole site.

1. Can AI systems reach your content?

• Pages are crawlable and indexable. AI features in search engines rely on the normal search index, so pages blocked by robots.txt, noindex or login walls are generally not used.
• You have made a deliberate choice about AI crawlers. Several AI companies use separate crawlers for training and for live retrieval, and they can be allowed or blocked individually in robots.txt. Blocking retrieval crawlers can remove you from those assistants’ answers.
• Important content is in the HTML, not only rendered by JavaScript, hidden in images or locked inside PDFs.
• Pages load fast and do not time out.
• Your sitemap is accurate and up to date.
• Optional: an llms.txt file lists your most useful pages for AI systems. Support for it is still limited, but it costs little.

2. Does each page answer clear questions?

• One main topic per page, with a title that states it plainly.
• The answer comes first. Each section opens with a direct statement, followed by the detail.
• Headings read like questions or claims (“How the impact factor is calculated”), not like labels (“Overview”).
• Sub-questions are covered. Think of what a reader would ask next, and answer it on the page.
• Passages make sense on their own. AI systems often retrieve a single paragraph. Avoid “as mentioned above” for essential information.

3. Is the content structured for extraction?

• Lists for steps and options, and tables for comparisons. These are easy to parse and quote.
• Definitions are explicit: “X is …”.
• Structured data (schema.org) matches the visible content: Article, Organization, Person, Product, FAQPage or HowTo, where appropriate.
• Consistent terminology. Use the same name for the same thing throughout.

4. Is it backed by evidence?

• Facts have sources, and primary sources are linked.
• Numbers, dates and named examples replace vague claims. In the 2023 GEO study, adding statistics, quotations and citations increased visibility in generated answers.
• Dates are visible: published and updated.
• Original contributions are clear: your own data, analysis, examples or experience. Content that only repeats what is everywhere else gives an AI system no reason to choose you.

5. Are entities and authorship clear?

• The organisation behind the site is described on an About page and in Organization markup.
• Authors are identified, with relevant expertise, where appropriate for the topic.
• People, companies, products and places are named precisely and consistently. Link to authoritative references such as official sites and Wikidata where useful.
• Your brand is described consistently across your site and profiles. AI systems compare sources, and contradictions reduce confidence.

6. Do others reference you?

• Your content is cited and linked by other credible sites in your field. Mentions and links remain strong signals of authority for both classic and AI search.
• You participate where your audience asks questions, such as industry publications, forums and conferences, with genuinely useful contributions.

7. Are you measuring the right things?

• Track AI visibility: check regularly how assistants answer the key questions in your field, and whether and how they cite you.
• Track referrals from AI assistants in your analytics.
• Don’t judge only by clicks. Being cited builds awareness even when users do not click.

What not to do

• Do not write for machines instead of people. Keyword stuffing did not help in the research, and search engines treat scaled, low-value content as spam.
• Do not hide text or create AI-only pages. Showing different content to crawlers and users violates search engine policies.
• Do not expect quick wins from tricks. The systems change constantly. Clear, sourced, well-structured content that answers real questions has been the durable strategy in classic search, and the evidence so far suggests the same holds for AI search.