Anticode Open builder

How to get your website cited by ChatGPT and AI search

AI assistants cite pages they can fetch, understand and trust to answer a specific question. That means letting their search crawlers in, putting clear, quotable facts in the HTML itself, and being the most specific source on something. There's no separate trick: Google says its AI features need nothing beyond ordinary SEO, and the other assistants reward the same basics.

Updated · By the Anticode team

How an answer finds your page

Working out how to get your website cited by ChatGPT and AI search starts with how those answers are put together. When ChatGPT, Perplexity, Claude or Google's AI Overviews answer with links, they generally run a web search behind the scenes, fetch a handful of results and write an answer from them, citing the pages they drew on. To be cited, your page has to be in the index the assistant searches, reachable by its fetcher, and contain a passage that answers the question better than the alternatives. "Generative engine optimization" is the name for this, but in practice it overlaps almost entirely with good search practice.

Step 1: let the right crawlers in

Each company runs separate bots for separate jobs, and robots.txt can treat them differently. OpenAI's crawler documentation explains that OAI-SearchBot gathers pages for ChatGPT search, GPTBot collects training data, and ChatGPT-User fetches a page when a person asks about it. Blocking GPTBot doesn't remove you from ChatGPT search; blocking OAI-SearchBot does. Google's AI features guidance says pages appear in AI Overviews only if they're indexed and eligible for a snippet, so a nosnippet or noindex keeps you out.

  1. Open /robots.txt on your site and read every User-agent block.
  2. Decide per bot: you can allow search fetchers and refuse training crawlers if you prefer.
  3. Make sure nothing in your hosting or firewall blocks crawlers by user agent without you knowing.
  4. Confirm the site is indexed by ordinary search engines first; several assistants rely on them. Submitting the site to Google Search Console shows you that.

Step 2: write passages worth quoting

An assistant lifts a sentence or two, so the sentence has to stand on its own. Compare "We offer competitive rates for all your needs" with "A standard boiler service takes about an hour and includes a flue gas analysis and a written report." The second can be quoted without the rest of the page.

  • Answer the question in the first line under each heading, then explain.
  • Use the nouns people use: the product name, the place, the model number.
  • Put facts in text, not only in images, PDFs or interactive widgets. A price list that's a photo is invisible to most fetchers.
  • Keep one topic per page so the page is the obvious source for it.
  • Show dates and sources for anything that changes or that you didn't measure yourself. Assistants and readers both weigh a page that says where its numbers come from.
  • Write an honest about page and contact details. Pages from an identifiable business are easier to trust than anonymous ones.

Step 3: add the machine-readable files

A sitemap.xml helps search engines find every page. An llms.txt file, a proposed standard for a Markdown summary of a site with links to its key pages, gives AI tools a short map. Neither replaces the content: Google says it needs no special AI text files for its own features, and support for llms.txt varies between tools. It's cheap to have and does no harm. The llms.txt explainer covers the format.

What Anticode does here

Every published Anticode site serves its own robots.txt, sitemap.xml and llms.txt. Design-engine pages are plain HTML with no scripts, so the words are in the page any fetcher receives. Anticode's analytics are counted without cookies and list referrals from AI assistants as their own source, and count AI and search crawlers separately from people, so you can see whether assistants are reading the site and whether anyone arrives from them. Analytics count views per site, not per page, so they won't tell you which page was cited.

Checking whether it's working

  1. Ask several assistants the questions your customers ask, with search turned on, and note which sources they cite.
  2. Ask about your business by name and check the facts they give back. Wrong hours or prices usually trace to an outdated page or listing somewhere.
  3. Watch AI referrals and crawler visits in your analytics over a few weeks rather than days.

Limits and honest expectations

No one can guarantee a citation: each assistant picks sources its own way and changes how it does so. Avoid services that promise placement in AI answers, and avoid writing pages aimed at bots rather than people, such as hidden text or lists of questions with thin answers. Those tend to make a page less trustworthy to the readers you actually want.

Questions people ask

Is generative engine optimization different from SEO?

Mostly not. Crawlable pages, specific answers and a trustworthy source help in both. The main extra step is checking robots.txt rules for each AI company's search crawler.

Should I block AI crawlers?

That's a choice about training, not visibility. You can refuse training crawlers such as GPTBot and still allow search fetchers such as OAI-SearchBot.

Do I need structured data for AI search?

Google says no special schema is needed for its AI features. Clear text on the page does most of the work.

Sources

  1. Overview of OpenAI crawlers (OpenAI docs), checked
  2. AI features and your website (Google Search Central), checked
  3. The /llms.txt file (llmstxt.org), checked