OpenAI tests ‘SearchGPT’ prototype as direct challenge to Google search engine

Sam Altman’s OpenAI unveiled a test version of its long-awaited AI search engine on Thursday in a direct challenge to Google’s dominance of the Internet search market.

Called “SearchGPT,” the tool will cite information obtained from websites and news publishers, including OpenAI content partners such as The Post’s parent company News Corp and The Atlantic, to provide “quick and timely answers with clear and relevant sources”.

The search engine is designed to be more conversational and interactive than traditional search engines.

Users enter their initial query, such as a search for “music festivals in Boone, North Carolina in August.”

From there, they can ask follow-up questions, such as whether a particular concert venue is family-friendly — with SearchGPT preserving the context of the conversation to find the best answer.

The tool is currently being tested by a “small group of users and publishers to get feedback,” OpenAI said in a blog post.

Eventually, it will be integrated into ChatGPT, which passed more than 100 million weekly users last fall. At the moment, users who want to try the tool are instructed to join a waiting list.

“Sam and the truly talented team at OpenAI naturally understand that for AI-powered research to be effective, it must be based on the highest quality and most reliable information provided by trusted sources,” said in a statement News Corp CEO Robert Thomson.

“For the heavens to be in balance, the relationship between technology and content must be symbiotic and provenance must be protected,” added Thomson.

The launch of OpenAI represents another headache for Google, which is awaiting a federal judge’s decision this fall on a landmark Justice Department antitrust lawsuit targeting its alleged monopoly over the Internet search market.

Shares of Google parent Alphabet fell nearly 2% after the OpenAI announcement.

OpenAI said that information obtained from specific news media and publishers will be prominently cited with “clear, inline, named attribution” and hyperlinks “so that users know where the information is coming from and can quickly engage with the more results”.

“We are committed to a thriving ecosystem of publishers and creators,” added OpenAI. “We hope to help users discover publisher sites and experiences, bringing more choices to search.”

OpenAI did not specify a timeline for a full rollout of SearchGPT. Publishers who decide to opt out of allowing their work to be used to train OpenAI models can still appear in SearchGPT results, the firm added.

Earlier this month, News Corp’s Thomson described the OpenAI content licensing deal as a “significant moment to recalibrate the search world” long dominated by Google.

In addition to the deal with News Corp and The Atlantic, OpenAI has signed partnerships with the Associated Press, Politico parent Axel Springer, Dotdash Meredith and other news publishers.

At the same time, OpenAI was sued by the New York Times, which accused the AI ​​leader of copyright infringement for using its articles without permission.

OpenAI has denied wrongdoing.

Google has roughly a 90% market share in internet search – edging out current rivals such as Microsoft and DuckDuckGo.

The federal DOJ lawsuit against Google alleges that the company relies on billions of dollars in annual payments to Apple, AT&T and other firms to ensure that its search engine is the default setting on most smartphones.

Google has faced harsh criticism for taking copyrighted content from publishers without proper attribution or permission — and then using that information to train its own AI models.

In June, Google unveiled a controversial search feature called “AI Summaries,” which put automatically generated summaries at the top of search results while dismantling traditional links to publishers and media outlets.

“AI Summaries” had a disastrous performance. Google was widely mocked for the strange results generated by this tool, such as telling readers to put glue on their pizza and eat rocks.

News Media Alliance CEO Danielle Coffey — who heads a nonprofit that represents more than 2,200 publishers, including The Post — said the feature was a “perverse turn to innovation” that will be “catastrophic for our traffic.” .

Elsewhere, the $1 billion Amazon-backed startup Perplexity, which has built an AI-powered search engine, was recently criticized by Forbes for regurgitating its exclusive reporting without proper credit or permission.

Forbes has since threatened legal action against Perplexity.