How to Get Cited by ChatGPT: The Complete Guide

To get cited by ChatGPT, you need three things in place: crawler access for OAI-SearchBot and ChatGPT-User, top-10 Bing rankings for the queries ChatGPT fans out to, and 40-80 word passages that answer those queries directly. ChatGPT Search, launched October 31, 2024, retrieves from the Bing index plus OpenAI's own crawl, then displays citations as inline source chips. Miss any of the three and the other two cannot compensate.

How does ChatGPT choose which sites to cite?

When a user asks a question that needs current information, ChatGPT decomposes it into multiple Bing-flavored sub-queries — a process called query fan-out. It retrieves candidate pages for each sub-query, extracts the passages that answer them, and cites the pages whose passages survive into the final answer. Established domains with corroborated facts win most of these contests: a claim that appears on your site, on G2, and in a news article gets synthesized; the same claim on your site alone often gets discarded.

The GEO research by Aggarwal et al. (KDD 2024) quantified what tips the balance at the passage level: adding quotations, statistics, and source citations lifted generative-engine visibility 30-40%, while classic keyword stuffing did nothing.

Which crawlers control ChatGPT access?

OpenAI operates three distinct bots, and each has a different job. Treat them separately in robots.txt.

A common self-inflicted wound: a blanket Disallow: / for all AI bots added in 2023 to block GPTBot, which silently also blocks the two retrieval agents. Audit your robots.txt and WAF rules before doing anything else.

The citation playbook, in priority order

1. Open the gates. Allow OAI-SearchBot and ChatGPT-User in robots.txt, and confirm your CDN or WAF (Cloudflare in particular) is not challenge-blocking them. Verify with server logs, not assumptions.
2. Win Bing first. Register in Bing Webmaster Tools, submit sitemaps, and enable IndexNow for instant indexing pings. ChatGPT's fan-out queries hit Bing, so a page outside Bing's top 10 rarely enters the candidate pool.
3. Serve full HTML without JavaScript. OAI-SearchBot and ChatGPT-User do not reliably execute JS. Client-side-rendered content is invisible; use SSR or prerendering for anything you want quoted.
4. Write extractable passages. Every H2 should be a question users actually ask, followed by a 40-80 word self-contained answer. ChatGPT lifts passages, not pages.
5. Build corroboration. OpenAI licenses Reddit data (deal announced May 2024), and review sites feed the consensus check. Earn mentions on Reddit threads, G2, industry publications — anywhere a second source can confirm your claims.
6. Track and iterate. Run a fixed prompt set weekly and log which domains get cited for your category. Citation tracking in Menra automates this sampling so you can see whether steps 1-5 are actually moving your citation rate.

What separates cited pages from ignored ones?

Cited pages tend to be specific rather than broad. A glossary entry defining one term, a statistics page with sourced numbers, or a comparison table beats a sprawling pillar page, because retrieval operates on chunks. Engines cite pages, not domains — which means a well-structured supporting page can out-cite your homepage, and a small site with perfect passages can out-cite a large site with vague ones.

Freshness matters at the margin. ChatGPT-User fetches live pages when browsing triggers, so a visibly updated page with a real dateModified change gets its newest facts into answers. Cosmetic date bumps without content changes do not help and can erode trust signals.

Start with the crawler audit today, get Bing indexing confirmed this week, and treat passage rewriting as an ongoing editorial standard rather than a one-time project. Teams that follow this sequence typically see first citations on long-tail prompts before they see them on head terms — that is the expected order, not a failure. For the broader strategy behind this playbook, see what GEO is and why it works.