ChatGPT Search does not have its own index. It uses Microsoft Bing — meaning your path to ChatGPT citations runs through Bing indexing, Bing ranking, and OpenAI's three-crawler framework. Most sites optimize for Google and assume ChatGPT will follow. It doesn't.
When a ChatGPT user asks a question that requires current information, ChatGPT queries Bing's search API. Bing returns a set of candidate URLs. ChatGPT browses those URLs in real time, extracts the relevant content, and synthesizes an answer — citing the sources it used.
This has two major implications:
This is the most commonly misunderstood technical detail in ChatGPT SEO. OpenAI has three distinct bots, each with a different purpose:
| Crawler | Purpose | Affects ChatGPT Search? |
|---|---|---|
| GPTBot | Model training data collection | No |
| OAI-SearchBot | ChatGPT Search index crawler | Yes — primary |
| ChatGPT-User | Real-time browsing during queries | Yes — secondary |
Most websites block or allow crawlers at the brand level: "I allow OpenAI crawlers." But robots.txt matching is by exact User-agent string. If your robots.txt only has GPTBot: Allow, OpenAI's search crawlers are unaffected — they use different User-agent strings and fall through to your default rule.
Check your robots.txt: Do you have explicit Allow rules for OAI-SearchBot and ChatGPT-User? If not, your site may be invisible to ChatGPT Search even if you've explicitly allowed GPTBot.
Submit your sitemap to Bing Webmaster Tools at webmaster.bing.com. Bing's index is significantly smaller than Google's — many pages indexed on Google are not in Bing. Check your Bing coverage and identify which key pages are missing.
Allow all three OpenAI crawlers explicitly. The minimal correct robots.txt entry:
User-agent: GPTBot
Allow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: ChatGPT-User
Allow: /
ChatGPT extracts answer passages from your pages. Structure content so that the first paragraph directly answers the question the page targets. Don't bury the answer after three paragraphs of background. ChatGPT's extraction is strongly biased toward page-opening content.
FAQPage schema tells ChatGPT exactly which questions your page answers and what the answer is. It's the cleanest extraction signal available. ChatGPT cites FAQ-structured content at significantly higher rates than unstructured prose.
ChatGPT's quality scoring includes E-E-A-T signals from the page. Include a visible author byline, an Article schema with datePublished, and a clear publication date. Pages with no author and no date score lower for trustworthiness.
ChatGPT browses pages in real time during answer generation. If your page takes more than 3 seconds to load or is heavily JavaScript-rendered, the browsing agent may time out before extracting content. Prefer fast, server-rendered HTML for content pages.
Neurobird's audit checks your robots.txt for all three OpenAI crawlers, your Bing indexability, and your answer-structure signals — free in 30 seconds.
Run free audit →