Short answer: for ChatGPT, Gemini or Perplexity to cite you, your site must be accessible to their crawlers, explain clearly who you are and what you offer, answer specific questions accurately and be backed by reliable mentions. There are no guarantees, but there is a checklist.
How these systems decide what to cite
Assistants with search consult web pages, extract passages and compose an answer citing the sources that best explain the topic. They favour pages that are crawlable, clear, well structured and backed by other sources. This is not a trick: it is doing content and technical work well.
Technical checklist
- Allow the crawlers. In
robots.txt, review OAI-SearchBot and ChatGPT-User (OpenAI), Claude-SearchBot and Claude-User (Anthropic), PerplexityBot and Perplexity-User, and Googlebot and Bingbot. The “training” bots (GPTBot, ClaudeBot, Google-Extended) are a separate decision. - Check for network-level blocks. Some security services or CDNs block AI bots by default.
- Correct sitemap and canonical, with no duplicate pages or 404 errors on what matters.
- Performance: a fast site with readable HTML that does not depend on heavy scripts.
- Structured data: Organization, Service, FAQPage and BreadcrumbList.
- An llms.txt file with a summary of the company and links to key pages (a proposed convention; not every engine reads it).
Content checklist
- One page per service, sector and product, with a title and first paragraph that define it in one sentence.
- An “About us” page with legal name, location, team and contact details.
- Direct answers to real questions and an FAQ section.
- Your own data: cases, figures, methodology and examples others can cite.
- Freshness: a visible date and regular reviews.
Reputation checklist
- Consistent profiles (LinkedIn, Google Business Profile, YouTube) with the same name and description.
- Real reviews on Google and on platforms in your sector.
- Mentions in media, directories and relevant communities.
- A presence in Bing: ChatGPT and Copilot rely on its index.
How to check whether it works
Put together 10 to 20 questions your customer would ask and test them each month in ChatGPT, Gemini and Perplexity. Note whether you appear, in what position, with what description and who appears in your place. It is an imperfect measurement, but useful to see the trend.
Common mistakes
- Blocking AI bots “just in case” and expecting to be cited.
- Generic content with no data or experience of your own.
- A different description of the company in each place.
- Promising guaranteed results.
Frequently asked questions
Can you pay to appear in the answers?
Some platforms are testing advertising formats, but organic presence in answers is earned with content, structure and reputation.
How long until it shows?
Technical changes can be crawled in hours or days. Reputation and mentions take longer.
Should I allow crawling for training?
It is a business decision. Allowing search and on-demand access crawlers is what makes citations easier; the training crawler affects the use of your content in future models.
See how we apply it in What is GEO and in our SEO and GEO service, or ask us for an audit.
