How does ChatGPT decide which sources to cite?

ChatGPT (with browsing) retrieves pages whose content is chunk-shaped, factually dense, and topically specific to the prompt, then prioritizes higher-authority domains when ranking quotable sources. Pages without extractable structure are skipped even when they rank well in Google.

Does ChatGPT favor pages with FAQ schema?

FAQPage schema helps ChatGPT identify question-answer chunks cleanly. It is not the only retrieval signal — extractable copy and outside corroboration matter at least as much — but it lowers the parsing cost for AI engines.

Will ChatGPT cite a page that is only days old?

Yes when ChatGPT is browsing. The browsing-grounded mode retrieves live pages, so a freshly published canonical answer page can appear in citations within days of being crawled.

Does adding llms.txt make ChatGPT cite a brand?

There is no published evidence that ChatGPT weights llms.txt in retrieval. Ship it as cheap hygiene, but the citation lift comes from the canonical answer pages, schema, and corroboration — not the file itself.

How is this different from SEO?

SEO targets Google ranking. ChatGPT citation targets retrieval inside an AI-generated answer. They overlap (both reward clarity, structure, and authority) but they are not the same channel. A page can rank well and still never be quoted.

WhoCites

How to Get Cited by ChatGPT

Getting cited by ChatGPT requires a quotable public source, not just a crawlable website. WhoCites measures whether ChatGPT cites the brand today and which source gaps explain the miss.

Last updated 2026-05-21

Run an AI visibility scan

Diagnostic answer

How do I get cited by ChatGPT? is not a mystery problem. It is usually a retrieval-surface problem: public pages either give AI systems enough clear evidence to quote, or they leave the engine to choose a better-documented competitor.

Symptom

ChatGPT answers the category question but cites competitors, directories, broad SEO guides, or no brand-owned page at all.

Likely cause

The brand has pages that humans can read, but not pages ChatGPT can safely quote: the answer is buried, the page lacks schema, facts are vague, and third-party corroboration is thin.

What to check

Check the public evidence layer before changing the product. The highest-signal checks are crawl access, answer-shaped copy, schema, citations, and whether the same buyer question is answered on one canonical URL.

The exact question appears in an H1, H2, FAQ entry, and first-answer paragraph.
The first 60-100 words answer the question before selling the product.
Visible facts include price, engine count, methodology, dates, and what the scan measures.
FAQPage and Service/WebApplication schema describe the same facts visible on the page.
The page is indexed or inspectable in Google Search Console and linked from the homepage.

What WhoCites measures

WhoCites does not guess from metadata alone. It runs the category prompts against live AI and search sources, then ties the result back to the pages and signals that explain the miss.

Whether ChatGPT names or cites the brand for category and recommendation prompts.
Which source URLs ChatGPT exposes when it chooses other brands or generic guides.
Whether the brand's pages have enough surface signals to be cited after fixes.
How ChatGPT visibility compares with Claude, Gemini, Copilot, Perplexity, Grok, and AI Overviews.

What to fix next

The next fix should improve the source material AI can retrieve, not just the words on a landing page.

Rewrite the first answer block as a neutral, self-contained explanation.
Add a short diagnostic table covering symptom, cause, check, and fix.
Publish a methodology or proof page that the answer page can cite internally.
Submit the changed URL through Search Console and IndexNow before re-scanning.

How to get cited by ChatGPT

To get cited by ChatGPT, publish one canonical answer page per buying-intent question, put the answer in the first ~60 words (BLUF), add FAQPage and Service/WebApplication schema, include verifiable facts per paragraph, and earn outside corroboration. WhoCites measures whether ChatGPT currently cites the brand and lists the missing page-level signals for $49.

Write for chunk retrieval, not article flow

AI engines extract chunks. Each paragraph should be self-contained — no 'as mentioned above', no narrative transitions. A reader landing mid-page should still get a complete, citeable thought from any single paragraph.

Add the schema ChatGPT can parse

FAQPage schema flags question-answer chunks. Service or WebApplication schema describes the offering. BreadcrumbList shows site structure. HowTo schema describes step-by-step processes. None of these guarantee citation, but they lower the parsing cost for ChatGPT and other engines.

Build verifiable factual density

Aim for roughly one concrete, citable fact per 150-200 words: engine count, price, methodology specifics, glossary-linked terminology. Invented stats, awards, and reviews are not just unethical — they get filtered out by trust signals and degrade overall citation rates.

Earn outside corroboration

ChatGPT favors brands whose claims are confirmed by third-party sources. Real press coverage, podcast appearances, listing on credible directories, and citations in independent articles all reinforce the retrieval signal. Fake authority backfires.

How WhoCites measures ChatGPT citation specifically

WhoCites runs buying-intent prompts in ChatGPT and records whether the brand is mentioned, where it ranks in the response, which competitors are cited instead, and which source URLs ChatGPT exposes. The same prompts run against 6 other engines so the result is cross-engine, not ChatGPT-only.