Contents

To appear in AI assistant answers, a site must be reachable by their crawlers, serve text as HTML without JavaScript and answer questions directly and verifiably: a clear answer in the first paragraph, question-style headings and facts with sources. Much of this is simply good SEO, with extra focus on structure and clarity that make text easy to quote.

Here is how AI assistants find sources and what to do on your site so they choose it.

How do AI assistants find sources for their answers?

AI assistants with search access find sources much like search engines: crawlers index sites, and when answering, the model receives several retrieved pages and composes a response from them, often with links. Knowledge baked into the model during training also shapes answers, but a website cannot influence that quickly.

What this means for your site:

  • pages must be reachable by crawlers — both search engines' and AI services';
  • text must be findable for queries that resemble the user's question;
  • the text must contain a passage that answers the question directly and is easy to quote;
  • the source must be trustworthy: a clear author, a date and links to primary sources.

How do you open your site to AI crawlers?

Open your site to AI crawlers through robots.txt and server-side rendering: allow the crawlers you want and make sure they receive text without running JavaScript. Many of these crawlers do not execute scripts, so a page assembled in the browser is empty to them.

Major AI companies publish their crawler names and separate them by purpose: training models versus fetching sources at answer time. You can allow the search crawlers and block the training ones — they are separate lines in robots.txt:

# Crawlers that fetch sources for answers — allowed
User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

# Collecting data for model training — your call
User-agent: GPTBot
Disallow: /

Sitemap: https://example.com/sitemap.xml

Check each company's documentation for current crawler names — they change and new ones appear. We explained why server HTML matters in SEO for JavaScript websites.

A computer screen showing an interface menu Photo: Unsplash

How do you write text that AI assistants quote?

Text that gets quoted answers the question in its first sentences, splits the topic into question-style sections and is specific: numbers, criteria, steps. A model would rather take a clear, self-contained paragraph than assemble an answer from a long introduction.

Technique What it looks like Why it works
Answer first The first paragraph answers the headline question A ready quote for a short answer
Question headings "How much does website support cost?" Matches how users ask
Each section opens with the answer Conclusion first, details after Every section can be quoted on its own
Tables and lists Comparisons, criteria, steps Structure is easy to extract and restate
Facts with sources A number linked to its primary source Increases trust and verifiability
FAQs A block of short answers Covers related queries

This blog follows the same rules: every article opens with a short answer, and sections are headed with questions.

What to avoid

  • long introductions about "the importance of the topic" before the answer;
  • vague statements without criteria or numbers;
  • invented statistics without sources — harder to verify and damaging to trust;
  • key information only in images or video with no text version.

Which technical signals help?

The same technical signals as classic SEO help: fast pages, a clean heading structure, schema.org structured data, current dates and a sitemap. You can also add an llms.txt file — a short guide to your site for language models.

What to check:

  1. All public pages are in the sitemap with a last-modified date.
  2. Article (BlogPosting), organisation and FAQ (FAQPage) markup matches the visible text.
  3. Articles show publication and update dates.
  4. Outdated content is refreshed or labelled — models prefer fresh sources.
  5. An llms.txt file at the site root describes the site and links to key sections.

The full technical checklist is in our technical SEO checklist.

How can you tell your site appears in AI answers?

It is harder to tell than with regular search: most AI services have no proper webmaster console yet. But indirect signs show the trend.

What to track:

  • referrals from AI services in your web analytics, by traffic source;
  • AI crawler visits in your server logs;
  • assistants' answers to key questions in your field — checked manually on a schedule;
  • brand mentions in answers, even without a link.

In short

  • AI assistants with search pick sources from retrieved pages — your site must be reachable by their crawlers.
  • Serve text as server HTML: many crawlers do not execute JavaScript.
  • Answer in the first paragraph, use question headings and add tables and lists.
  • Back facts with sources, show dates and refresh outdated content.
  • Track referrals from AI services and their crawlers' visits in your logs.

FAQ

  • Should we block AI companies' crawlers?

    It is a business decision. Blocking AI search crawlers removes your chance of being cited with a link. Blocking training-data crawlers usually does not affect visibility in search-backed answers.

  • How is AI optimisation different from regular SEO?

    The foundation is the same: accessibility, speed, structure and useful content. The difference is the emphasis on clear, self-contained answers that can be quoted and on verifiable facts.

  • Can appearing in AI answers be guaranteed?

    No. As with search, you can only improve your chances: make the site accessible, the text useful and clear, and the information current and backed by sources.