How ChatGPT chooses sources

When a question needs current information, ChatGPT runs a search, retrieves candidate pages from a web index, and writes an answer with inline links and a sources panel. It is not reading your rankings; it is retrieving and quoting passages it can trust.

OpenAI uses distinct crawlers: OAI-SearchBot surfaces and links pages in ChatGPT's search results, ChatGPT-User fetches a page live when a prompt sends ChatGPT to it, and GPTBot is for model training. Block OAI-SearchBot or ChatGPT-User in robots.txt and your pages can be left out of the cited sources — so the first requirement is simply to let those crawlers reach and index your content.

Visual explanation

How a source can enter a cited ChatGPT answer

How a source page can become an AI-answer citation A simplified four-stage model shows crawl and index access, retrieval as a candidate source, passage-level evidence and context, and selection in a cited AI answer. Publishers can improve the page, but the engine controls selection. SIMPLIFIED DOCUMENTATION-BASED MODEL · NO CITATION GUARANTEE PUBLISHER-CONTROLLED INPUTS ENGINE-CONTROLLED SELECTION 01 Discoverable page • Crawl allowed• Indexable HTML• Clear entity + topic 02 Candidate source • Relevant query match• Fresh page context• Connected site structure 03 Supported passage • Direct answer• Traceable source• Claim boundary visible 04 Cited answer Source page cited The engine decides whether, when, and how to cite.
A simplified public-documentation model. Search and retrieval systems change, and no page-level tactic guarantees a citation.

Why ranking first is the wrong goal

Answer engines are probabilistic. Ask the same question twice and the cited sources can change. There is no fixed first position to win and no reliable way to guarantee one, so promises of a number-one spot in ChatGPT should be treated with caution.

The realistic goal is share of citation: how often your brand and pages appear across the questions your buyers ask. That is something you can influence, measure, and grow over time.

Be discoverable first

The search in ChatGPT leans on a web index that overlaps heavily with Bing, alongside its own crawler. So Bing visibility matters as much as Google here. Verify the site in Bing Webmaster Tools, submit a sitemap, and make sure new pages get indexed quickly.

Allow the AI crawlers in robots.txt rather than blocking them by default, and keep critical content in the HTML rather than only rendering it with JavaScript. Content the engine cannot fetch or parse cannot be cited.

Make content easy to cite

Answer engines lift passages, not pages. Put a direct answer near the top of each section, use question-shaped headings, and state facts plainly. Specific figures, clear definitions, and short lists are easier to quote than long, hedged prose.

Keep pages current. Answer engines tend to prefer recently updated sources, so maintained pages with honest publish and update dates have an advantage over stale ones.

Build off-page signals

For most sites, and especially B2B software, citation is won off the page. A consistent brand entity, mentioned across credible sites, makes you a source the model reaches for. Third-party review platforms and current roundup lists are frequently cited for product and vendor questions.

Presence on high-trust sources — reviews on platforms like G2 and Capterra, authentic discussion in communities, a well-sourced reference page — often matters more than any single change on your own site. Being named in many trusted places is the signal answer engines reward.

What does not work

Schema markup alone does not earn citations. It helps machines parse a page, but it is not a lever on its own. Keyword stuffing and padding length tend to hurt, because the model prefers concise, trustworthy answers.

Ranking first in Google does not guarantee a ChatGPT citation either. Only about 12% of the URLs cited by AI assistants like ChatGPT also rank in Google's top 10 (Ahrefs, 2025) — the overlap is weaker than most people assume, which is why Bing indexing and off-page trust deserve their own attention.

How to measure it

Pick a fixed set of questions your buyers would ask, run them across ChatGPT and other answer engines on a regular schedule, and record whether your brand is mentioned and whether your pages are cited. Compare against a few competitors to get a share-of-citation trend.

Reconcile that with referral traffic from AI sources in your analytics. Because outputs vary from run to run, trust the trend over several checks rather than any single answer.

How we sourced this guide

This is practitioner guidance, not a study. It reflects how Gadex runs content production across the sites it operates, checked against the current public documentation of the search and answer engines it describes. There is no proprietary dataset behind it and no sample to cite.

Where a platform’s behaviour is stated, it is drawn from that platform’s own documentation at the time of the update shown above — search and AI systems change without notice, so verify anything you are about to act on. Where a claim could not be supported, it is framed as our practice rather than as fact. See our methodology and editorial standards for how pages are researched, reviewed, and corrected.

Next step

Want Gadex to run this on your site?

We turn guides like this one into a running content pipeline — researched, fact-checked, and shipped to your CMS.

See the SEO content service Book a 15-minute call

Published by Gadex, the SEO and AI-search content service operated by ALPHAOSCAR EURL. Every guide follows our editorial standards — sourced, fact-checked, and reviewed before publishing. About Gadex.