1. A user question can become a web search
ChatGPT does not attach web citations to every response. OpenAI says it may search automatically when a question would benefit from current information; a user can also start a search manually. Its Help Center explains that queries may be rewritten into targeted queries, with follow up queries possible after initial results. The service may also draw on content provided directly by partners. OpenAI’s announcement of ChatGPT Search describes a conversational search experience that connects questions with relevant web sources.
For the wider retrieval concepts, see vector embeddings and SEO and the How AEO works hub.
2. What a citation and Sources panel tell you
In a response that uses web search, a citation is a clickable route from part of the answer to a source page. The Sources control, when available, groups cited sources and other relevant links. This helps readers inspect the underlying material; it does not certify that every sentence is fully supported or correctly interpreted. OpenAI cautions that search results and citations can be incomplete, outdated, or incorrect. Open the source, check its publication or update date, and compare the cited passage with the claim.
A visible link documents attribution in that response; it does not guarantee referral traffic.
3. Citation, source listing, mention, and click are different signals
Teams often compress different observations into the phrase ‘AI visibility.’ Keep them separate. A page can be cited inline, appear among additional Sources links, be named without a clickable page citation, or be absent. A reader may click through or simply use the answer. These outcomes distinguish whether a source was linked, a brand mentioned, or a visit reached the site. When reporting, record the exact prompt, date, locale or account context when known, answer text, cited URL, and whether the link was opened.
Label the observation rather than extrapolating beyond it.
4. What publishers can control: search crawler access
OpenAI identifies OAI SearchBot as the crawler used to surface websites in ChatGPT search. Its crawler documentation says sites that opt out of OAI SearchBot will not be shown in ChatGPT search answers, though they may still appear as navigational links. OpenAI recommends allowing this bot in robots.txt and allowing requests from its published IP ranges. A robots change can take about 24 hours to affect its search systems. These are eligibility and access controls, not a promise of selection, prominence, or citation.
Read OpenAI’s crawler documentation before editing directives. Then check the live robots.txt, CDN or WAF rules, server responses, and any authentication wall. A crawler allowed by robots.txt can still be blocked by a firewall or receive an error page, so verify the actual path to the content rather than relying on one configuration file.
5. Search access is separate from model training controls
OpenAI distinguishes OAI SearchBot from GPTBot. The former is for surfacing websites in ChatGPT search; GPTBot crawls content that may be used to train OpenAI’s generative AI foundation models. OpenAI says these robots.txt settings are independent, so a publisher can allow OAI SearchBot while disallowing GPTBot. ChatGPT User is another user agent associated with certain user initiated visits; OpenAI says it does not determine whether content may appear in Search. This matters when a client asks whether ‘blocking ChatGPT’ will stop a page from being cited.
There is no single switch that covers every purpose. Review the documented user agent and intended use, agree the policy with the site owner, and test the exact robots rules. Do not treat training preferences as search opt outs or infer search eligibility from a ChatGPT User request.
6. Make the page useful if it is retrieved
After access, the editorial task is to make the page a dependable source for the question it genuinely answers. State the scope, define specialized terms, separate facts from recommendations, and place qualifications near the claims they limit. Cite primary evidence for consequential assertions and name dates, versions, locations, or exceptions when those details change the answer. Avoid padding a page with repeated variants of a question: it gives readers little extra value and cannot ensure retrieval.
For example, a software agency writing about a product’s data export policy should distinguish the documented export steps from its own advice, link the official policy, and state when that policy was checked. A concise, self contained passage helps a human verify the answer; it is not a disclosed OpenAI ranking signal. The answer first content guide covers structuring a direct response without stripping away its evidence and context.
7. A practical agency workflow for checking eligibility and citations
Treat this as a repeatable diagnostic, not a citation building trick. First, identify a small set of real questions clients and customers ask, including questions where freshness or source verification matters. Second, inspect the canonical page: is it public, indexable as intended, returning a healthy response, and readable without an account? Third, review robots.txt and network controls for OAI SearchBot, then confirm the client’s separate GPTBot preference.
Fourth, test a query in ChatGPT Search and save the complete response with its citations and Sources list; repeat on different days or contexts before describing a pattern. Fifth, open each cited URL and check whether it supports the relevant claim. Finally, compare observations with ordinary site analytics, remembering that not every answer view or citation becomes a measurable visit. For platform comparison, see how Perplexity indexes content and how Gemini retrieves answers.
8. Report evidence, not a promised outcome
An agency can responsibly report what it observed: for example, ‘On 12 May, this test prompt produced an answer that cited the client’s guide in the Sources panel.’ It should not turn that observation into ‘the client now ranks in ChatGPT’ or predict how often the citation will appear. Responses can vary with query wording, timing, location, product changes, and available sources; OpenAI does not provide a fixed publisher facing formula that would justify a guarantee.
Keep a dated log of the prompt, response, cited URL, and evaluation criteria, and distinguish direct citations from brand mentions. Use findings to identify factual gaps, inaccessible pages, or confusing explanations not to claim a placement that cannot be verified. The ChatGPT Search optimization guide offers a broader publisher readiness view, while testing AEO readiness helps organize repeatable checks.
Key takeaways
- ChatGPT may search automatically or on request; only search enabled answers may show web citations.
- Citations and Sources links aid verification but can be incomplete, outdated, or incorrect.
- OAI SearchBot governs search visibility; GPTBot and ChatGPT User have different documented purposes.
- Crawler access supports eligibility, not a promise of selection, citation, placement, or traffic.
- Measure dated, reproducible observations and keep citations, mentions, and visits distinct.
Frequently asked questions
Does every ChatGPT answer cite websites?＋
No. ChatGPT may search when current information would help, and users can choose Search. Responses that use web search may include citations; an answer that does not use web search may not show web citations. The interface and availability can change, so confirm the current behavior in the product.
Does allowing OAI SearchBot guarantee that my page will be cited?＋
No. OpenAI documents OAI SearchBot as the crawler for surfacing websites in ChatGPT search and says opting out prevents a site from appearing in search answers. Allowing it removes that particular access barrier; it does not guarantee that a page will be selected, cited, or shown prominently.
Can a site allow ChatGPT Search crawling but block training crawling?＋
OpenAI documents the OAI SearchBot and GPTBot robots.txt settings as independent. A publisher can allow search crawling while disallowing GPTBot for training related crawling. Review the current official documentation and the site owner’s policy before changing either rule.
Can an agency promise a ChatGPT citation after an optimization project?＋
No reliable promise follows from the published crawler controls or search documentation. An agency can improve access, clarity, evidence, and measurement, then report dated observations. It should not guarantee a citation, rank, frequency, or traffic outcome.