Brands keep asking which search engine sits behind ChatGPT so they know where to rank. Wrong question. ChatGPT retrieves through more than one door, and the honest 2026 answer is a blend most people describe wrong in both directions.
Yes, and this part is well documented. Microsoft and OpenAI announced Bing search inside ChatGPT Plus on May 23, 2023, then OpenAI shipped a standalone ChatGPT search product on October 31, 2024, built on the same Bing-powered foundation. That's where the "ChatGPT runs on Bing" line came from, and for over a year it was close to the whole story.
It's no longer the whole story. Bing still supplies part of what ChatGPT retrieves - Microsoft's own Bing Webmaster Tools AI Performance report exists precisely because Bing indexation still feeds AI answer surfaces, ChatGPT included. But OpenAI has spent 2025 and 2026 building infrastructure that doesn't depend on Bing alone.
Not by design, and not by disclosure. OpenAI has never named Google as part of ChatGPT's search stack - the entire reason the Bing partnership exists is that OpenAI doesn't have a search deal with Google, and Google's index isn't licensed out to competitors.
That said, a technical investigation published by Search Engine Land in August 2026 found something worth naming carefully. Researchers examined the internal pipeline identifiers ChatGPT's client briefly exposed and found four values behind search results: labrador (OpenAI's own retrieval hub, covering web search, news, arXiv, Reddit, and YouTube), bright and oxylabs (scraped Google results purchased live through third-party data providers), and serp. None of these are documented by OpenAI, which speaks publicly only of "third-party search providers."
Read that precisely: this is industry-inferred, from one investigation reading undocumented pipeline flags, not something OpenAI has confirmed. It suggests Google-originated results can enter ChatGPT's paid tier indirectly, through scraping intermediaries, not through an OpenAI-Google index partnership. Treat it as a signal that the stack is messier than "Bing or Google," not as proof ChatGPT is secretly Google-powered.
The same investigation, plus OpenAI's own developer documentation, describes a layered pipeline rather than a single index lookup:
OpenAI's own crawler, OAI-SearchBot, is documented as the bot that indexes pages for ChatGPT search and its citations. It's a different bot from GPTBot, which OpenAI uses to gather model-training data - you can allow one and block the other in robots.txt, and many sites should.
This is where "does ChatGPT use Bing or Google" gets asked as shorthand for a bigger question: does every AI engine work the same way? No.
Through 2025, "get into Bing and you're covered for ChatGPT" was reasonable, if incomplete, advice - it was true enough for most sites because Bing carried most of the retrieval weight. In 2026 it's only partly true. Bing indexation still opens a door into ChatGPT. It's no longer the only one, and it isn't necessarily the primary one for a given query, tier, or user.
Treat Bing indexation as necessary and not sufficient. A page can rank well in Bing and still never surface in ChatGPT if OAI-SearchBot hasn't crawled it, if it's blocked from OpenAI's own retrieval layer, or if the query routes through a provider that never touches your page at all. This is the same shift we cover for the reverse case - a page reads well on Bing's own reporting and still says nothing about whether ChatGPT, Claude, Gemini, or Perplexity actually recommend you. See how to navigate AEO for the broader framework.
We don't debate which index ChatGPT prefers this quarter - the stack changes faster than anyone can document it, and OpenAI doesn't publish the split. Instead we ask the engines the real questions your buyers ask, on a fixed schedule, across ChatGPT, Claude, Gemini, and Perplexity, and read what each one actually says. Access and indexation explain why you can appear. Only the answers themselves tell you whether you do.