How to get cited by ChatGPT

How does ChatGPT decide what to cite?

ChatGPT cites a page only when it searches. When it answers from what its model already holds, the reply carries no sources at all; when the question sends it to its search layer, it retrieves candidate pages, reads them, and links the ones it drew from. OpenAI's crawler documentation is explicit: a site opted out of its search crawler "will not be shown in ChatGPT search answers". An Ahrefs study of 1.4 million prompts, published in April 2026, found that about 88 percent of the URLs ChatGPT ended up citing came directly from its general search index.

Ranking gets you into the candidate pool; quotability gets you picked. The pages that win are roundups, comparisons, definitions and direct answers to one buyer question. Home pages rarely qualify, because a home page describes a company rather than answering anything. Why ChatGPT recommends your competitors shows the same mechanism from the losing side.

Which OpenAI crawlers need access?

Three OpenAI user agents matter, and OAI-SearchBot is the one that governs whether you can be cited. OpenAI documents OAI-SearchBot as the crawler behind ChatGPT search, GPTBot as its training crawler, and ChatGPT-User as the fetcher that visits a page on a user's behalf. Allow all three in robots.txt:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Allow: /

User-agent: ChatGPT-User
Allow: /

To stay out of training while remaining citable, disallow GPTBot alone and leave OAI-SearchBot open; OpenAI documents the two as separate controls.

A CDN or firewall rule stops a crawler before it reads robots.txt, so a clean robots.txt proves nothing on its own. That matters more from 15 September 2026, when Cloudflare's new defaults block crawlers it classes as Training and Agent on pages that display ads while leaving Search allowed, a change TechCrunch reported applies to new customers, new sites and every existing free plan. GPTBot does training work and ChatGPT-User acts for a user, so both can be caught even when robots.txt allows them; if your site sits behind Cloudflare, check its AI crawler settings. The two access tests in the Perplexity guide work for OpenAI's bots too.

What does a citable page look like?

A citable page carries a complete answer in the first sentences under a heading phrased the way the buyer asked. Search Engine Land's November 2025 audit of 15 domains, close to 2 million monthly organic sessions and 7,500 ChatGPT referral sessions found that 72.4 percent of the blog posts ChatGPT cited had an identifiable answer capsule, defined as a self-contained answer of roughly 20 to 25 words placed directly after the title or a question-shaped H2. More than nine in ten of those capsules contained no links at all.

The model quotes in one pass: it lifts the sentence that resolves the question and moves on, and it will not assemble an answer from a paragraph that opens with context and lands the point at the end. Write the answer first, then explain it, and keep links out of the capsule and in the sentences that follow. That structure is the core of what AEO is, and it is how this page is written.

Does schema markup get you cited?

No measured effect. JSON-LD helps engines parse a page, which is why CiteRank checks for it, and nothing more. Google's documentation on AI features states there is "no special schema.org structured data that you need to add" to appear in AI Overviews or AI Mode. Ahrefs' May 2026 study tracked 1,885 pages that added JSON-LD against 4,000 matched control pages: ChatGPT citations moved 2.2 percent and AI Mode 2.4 percent, differences the authors say "could easily be random noise", while AI Overviews citations fell 4.6 percent. Every page in the sample was already heavily cited, so the study is silent on pages that have never been parsed.

Treat schema as plumbing: add it so a crawler reads the same title, date and author the visitor sees, and be wary of anyone selling a structured data project as an AI visibility fix. The AI Overviews guide covers what Google does reward.

How long does it take and how stable is it?

An engine cannot quote a page it has not indexed, so expect days to weeks before a new page can appear, and expect the answer to keep changing after it does. The same Ahrefs prompt study found the median page ChatGPT cited was around 500 days old, so age alone is no barrier. On stability, a SparkToro study published in January 2026 ran 12 prompts 2,961 times through ChatGPT, Claude and Google's AI and found less than a 1 in 100 chance that two responses to the same prompt from ChatGPT or Google's AI would list the same brands.

So a single check tells you almost nothing, and nobody can promise a citation by a date. What you can measure is frequency: how often your name appears for the same question across repeated, dated checks. That is the number to watch as you improve your pages.

How do you check whether it worked?

Open a fresh ChatGPT chat with no history, ask a real buyer question the way a stranger would, and note who was named, in what order, and which sources were cited. Repeat the same question in two or three more fresh chats, then again a week later, because one run is a snapshot rather than a verdict. The full method is in how to check what ChatGPT says about your business.

CiteRank runs the same check every day across five engines and keeps a dated record of who was named and which sources were cited. Either way, start the record this week; a trend needs a baseline.