A recent report found that ChatGPT uses Google Search results to provide live answers on fast-moving topics. The report says OpenAI used a paid scraping service called SerpApi to draw search results from Google. This setup lets ChatGPT answer questions about news markets and sports with up-to-date information.
- What the report says about how it works
- Evidence shown in public demonstrations
- What OpenAI asked for and what happened in court
- Why this matters for users and for the market
- How this fits into the wider rivalry among AI products
- Limits and technical challenges
- What company leaders have said
- Practical implications for journalists and businesses
What the report says about how it works
The report states that ChatGPT does not run its own live web crawler. Instead, it calls a customised tool that pulls snippets from Google Search. The tool uses SerpApi as the middle layer to fetch the search results. OpenAI is said to combine those results with its own models to generate an answer.
Evidence shown in public demonstrations
A demonstration by a former Google engineer showed that ChatGPT could surface content that only Google had indexed. That example was used to show the practical link between Google Search and ChatGPT answers. Reporting on the demonstration has been widely shared in technology news coverage.
What OpenAI asked for and what happened in court
OpenAI asked Google for direct access to the Google search index. According to trial testimony, the request was refused. That testimony came during the United States antitrust case involving Google. OpenAI witnesses told the court that the company still relies on external search sources while it builds its own index.
Why this matters for users and for the market
Search data is a core resource for any service that tries to answer questions about events that change by the hour. Owning a search index gives a firm deep control over freshness and coverage. Companies that lack their own index must find workarounds. For ChatGPT, those workarounds now include drawing on Google Search results via a third party.
How this fits into the wider rivalry among AI products
Other models have different arrangements. Some services use their parent company’s search data directly. Others tap public social platforms for fresh signals. The difference matters when users compare sources for speed and reliability. The reliance on Google highlights how closely search and generative AI have become linked.
Limits and technical challenges
Using scraped results can work well for many queries. It can also create legal and technical risks. Web scraping services must follow site rules and handle changes in index format. OpenAI has told courts and regulators that it plans to build its own index to reduce dependence on outside search data. Those plans remain a work in progress.
What company leaders have said
OpenAI executives have testified about search strategy in court. They said the company aims to process most search traffic using its own index over time. Company leaders also said that some alternative providers have quality problems. That testimony explains why OpenAI pursued access to Google and why it has used other search sources while it builds its own systems.
Practical implications for journalists and businesses
It is questionable whether reporters should use ChatGPT responses to matters that have just occurred. Firms that utilize ChatGPT in monitoring the market or the news need to cross-check important facts using independent sources. The new reporting conditions the need to know where the live signals behind a response are. The source of the search results may influence trust and accuracy.