It was an open secret to some. But it's still a bombshell shaking Silicon Valley. According to The Information, OpenAI, Sam Altman's company behind ChatGPT, uses SerpAPI, a service specialized in scraping search results—to directly retrieve data from Google. And this despite an explicit refusal from Google, which declined in 2024 an official request for access to its index. In other words: OpenAI seeks to compete with Google on search… while exploiting its own results. A terrible admission of weakness for the leader of conversational platforms, especially as it is already weakened by numerous criticisms that accompanied the release of its latest language model, GPT-5.
OpenAI's secret weapon for conducting web searches: Google's search results page
The Information's investigation published on August 22, 2025, signed by Amir Efrati, Stephanie Palazzolo, and Natasha Mascarenhas reveals that ChatGPT relies on SerpAPI to answer questions in real time. This startup, based in Austin, Texas, well-known in the scraping world, circumvents Google's restrictions to provide customers with raw search results.
The revelation, widely picked up in the media and notably by Tom's Guide and TheFly, highlights this paradox: to replace Google, OpenAI remains in reality… dependent on Google.
Yet a clear refusal from Google
It is worth noting that in July 2024, OpenAI officially approached Google to access its Search API. An email, revealed during the US antitrust trial against Google, shows that the response was negative as early as August 2024.
In the background, Google fears that by granting access to its index, it would directly strengthen the quality of competing products and lose its strategic advantage. Court documents reveal that for Google, offering grounding capabilities (that is, anchoring search results to chatbot responses) to its rivals would improve their products. The American giant also insists that its Knowledge Graph, its Oneboxes, and its Related Questions (accessible internally via the Gemini App) constitute a decisive competitive advantage it does not wish to share. It is this logic that led to the refusal given not only to OpenAI, but also to Anthropic (Claude), whose similar request in October 2024 was also rejected.
The sworn testimony of Nick Turley
Another key revelation: during a hearing in the lawsuit brought by the U.S. Department of Justice against Google in April 2025, Nick Turley, Head of Product ChatGPT, acknowledged OpenAI's dependence on third-party sources. His statement resonates today as an admission:
"Our goal was to have 80% of user queries answered from OpenAI's own index by the end of 2025. We are, however, still years away from reaching that level."
In other words: OpenAI is far from having the means to dispense with external data, let alone compete with Google's sprawling index.
The study that changed everything: Alexis Rylko exposes OpenAI
For those closely interested in how ChatGPT Search works, the fact that OpenAI relies in part on Google's index is not news. As early as July 6, 2025, SEO consultant Alexis Rylko published an investigation in his SEO, Data & Growth newsletter. By analyzing the JSON files generated by SearchGPT, he demonstrates that:
Up to 90% of ChatGPT Search URLs correspond to Google.
Barely 30% match Bing.
The snippets and timestamps correspond to those displayed on Google's search results page.
Notably picked up by Search Engine Land, this study marked a turning point: the SEO community became aware that ChatGPT, for all its airs as an alternative, was directly relying on the index of the world's largest search engine.
Google tightens the scrape
It may be just coincidence, but let's remember that in January 2025, Google brutally strengthened its anti-scraping mechanisms: reinforced captchas, IP limitations, behavioral detection, and mandatory JavaScript rendering on certain SERPs. Result: several major SEO tools (Semrush, SimilarWeb, SE Ranking, etc.) went down for several days, unable to collect data.
A warning to Sam Altman's company: SerpAPI, today so precious to OpenAI, could one day be blocked.
Is the scraping of its search result pages by its competitors so problematic for Google?
We know it, OpenAI, and it is not the only one, aspires to build the search engine of the future. But facts are stubborn: without Google, the company cannot ensure the best quality for its answers, which require updating the chatbot's knowledge base.
Is this dependence of its competitors not good news for Google? SerpAPI provides access only to raw search results, not the ranking logic hidden behind the algorithm, nor the enriched data from the Knowledge Graph that Google has been building since 2012 and continuously refining.
In other words, OpenAI can draw from the surface, but Google's competitive advantage lies in its ability to organize, rank, and link information at scale to feed and train its large language models.
ChatGPT's arrival in November 2022 had caused panic and the implementation of a "code red" at Google. But two and a half years later, hasn't the wind already shifted?
Commentaires
Aucun commentaire pour le moment. Soyez le premier à commenter !
Ajouter un commentaire