Does ChatGPT Use Google?
ChatGPT SEO starts with knowing what index it actually retrieves from: not Google's, but OpenAI's own crawl blended with scraped Google data.
ChatGPT does not query Google's search index directly; it retrieves primarily from OpenAI's own index, internally called Labrador, with two additional pipelines pulling scraped Google results through third-party data providers.
No, ChatGPT does not use Google's search index. It cannot query Google's ranking systems directly the way a browser search bar can. What it uses instead is a blend of OpenAI's own index and Google content that arrives secondhand, through paid data resellers, and the mix between those sources has been shifting all year.
The four pipelines nobody was supposed to see
Between May 21 and July 21, 2026, researchers tracking ChatGPT's raw network traffic found a field in its server-side events called result_source. It only ever carried four values before OpenAI removed the field in late July: Labrador, Bright, Oxylabs, and SERP.
Labrador is OpenAI's own index, built from its own web crawl plus specialized verticals for news, academic papers, Reddit, and YouTube. Bright and Oxylabs are scraped Google search results, bought live from third-party scraping providers rather than licensed from Google directly. The fourth, SERP, is a supplementary pipeline whose actual source the researchers did not publicly identify. None of the four is a direct, licensed line to Google's own ranking systems.
The clearest technical evidence that Labrador is not simply relabeled Bing or Google data: researchers found only about 1.5% overlap between Labrador's URLs and Bing's top 20 results for the same queries, and Labrador's titles ran up to 289 characters long against Bing's roughly 75-character cap, a formatting difference that would not exist if one system were just repackaging the other's output.
Why this shift happened
This didn't start as a strategy of independence. In April 2025 antitrust testimony in the DOJ's case against Google, OpenAI's head of ChatGPT product, Nick Turley, testified that OpenAI had approached Google about a search partnership and been rejected, and described an internal goal of serving about 80% of ChatGPT's traffic from OpenAI's own first-party index, calling it "a lofty goal" the company was "nowhere near close" to hitting. The Labrador pipeline identified over a year later is that same project having reached production, even if slower than OpenAI originally hoped.
A Search Engine Land analysis published August 17, 2026, covering 1,200 ChatGPT answers and roughly 88,000 search results, found free-tier "Think" mode pulling 74.7% of its results from Labrador, versus paid-tier "Thinking" mode, which leaned into scraped Google content instead at 75.3%. The split is not fixed, and OpenAI has run live A/B tests weighting its own index against third-party results for different slices of traffic.
What this changes for chatgpt seo
The practical implication is that indexing in Bing is no longer the whole story. How to Rank in ChatGPT covers the crawler-access basics: allow OAI-SearchBot, keep pages fetchable, state facts plainly. Those still apply, and now matter more, not less, because Labrador is populated by OpenAI's own crawl rather than inherited from Bing's index. A page invisible to OpenAI's crawler can be excluded from Labrador results even while ranking fine in Bing.
A GEO audit checks whether a site is actually crawlable by OpenAI's bots and currently surfaced across ChatGPT, Perplexity, Claude, and Gemini, rather than assuming Bing indexation alone still covers it.
Sources: Inside ChatGPT's retrieval stack: The index, cache, and pages it actually reads (Search Engine Land, Aug 17, 2026), ChatGPT built its own search index (Peec AI, Sep 14, 2026), Exec says OpenAI would buy Chrome if judge orders Google breakup (Courthouse News, Apr 2025)
Frequently asked
Does ChatGPT use Google's search index directly?
No. Reverse-engineered server-side event data captured between May and July 2026 showed ChatGPT's retrieval requests carrying a field with four possible values, and none of them was a direct Google API call. OpenAI does not have a licensing deal that hands it raw access to Google's index.
So does ChatGPT ever see Google's results at all?
Yes, indirectly. Two of the four retrieval pipelines identified in that research, labeled Bright and Oxylabs, are scraped Google search results purchased live from third-party data providers. Google content reaches ChatGPT through a reseller, not a partnership.
Did OpenAI ever try to build its own search index?
Yes, and the effort is further along than most sites assume. In April 2025 antitrust testimony in the DOJ's case against Google, OpenAI's Nick Turley said the company had approached Google about a search integration and been turned down, and described an internal goal of serving about 80% of ChatGPT's traffic from OpenAI's own first-party index, a goal he testified the company was 'nowhere near close' to hitting at the time. The index he described, since identified as Labrador, is now live in production.
Is Bing still part of the picture?
Less than it used to be, and maybe not at all for web results. The same research that found the Labrador pipeline also tested it directly against Bing and found only 1.5% of Labrador's URLs appearing in Bing's top 20, with no matching snippets between the two, concluding Labrador is not fed by Bing's index. The fourth pipeline, labeled SERP, has not been publicly identified as Bing or anything else. What's confirmed is that ChatGPT's 2024 launch leaned on a Bing partnership, and that reliance has clearly shrunk since.
What does this mean if I'm trying to get cited by ChatGPT?
Optimize for OpenAI's own crawler and cache, not just Bing indexation. If a page is blocked to OAI-SearchBot or GPTBot, it can be excluded from Labrador results even while sitting fine in Bing's index, because Labrador is populated by OpenAI's own crawl rather than borrowed from Bing.
More in GEO & AI Search
Real llms.txt Files, Annotated
Five real llms.txt examples from Vercel, Zapier, Cloudflare, Stripe, and Anthropic, fetched and counted, showing curated versus dumped in practice.
How to Generate a Correct llms.txt
An llms.txt generator can produce a file that fails the spec. Here is the one required line, the mistakes auto-generators make, and how to build a correct one.
AEO vs SEO vs GEO: The Three-Letter Problem
AEO vs SEO: AEO targets the answer inside a snippet or AI reply, SEO targets page rank, and no one agrees where AEO ends and GEO begins.
See how AI describes your brand
A free GEO audit takes about 90 seconds and tells you what ChatGPT, Claude, Perplexity and Gemini actually say about you.
Run a free GEO audit →Written by YaBaDo's AI agents and reviewed by a human before publishing. If you find something wrong, tell us and we will correct it.