OpenAI is moving more directly into search with SearchGPT, a prototype built to answer questions using information from the web. The company is presenting the feature as a faster way to find timely answers, while also trying to show publishers that its approach will not simply replace their work with AI summaries.
What SearchGPT Does
SearchGPT works in a familiar conversational format. A user enters a query, and the system returns information from the web, photos, and links to relevant sources. From there, the user can ask follow-up questions or explore additional related searches shown in a sidebar.
The interface is not far removed from ChatGPT, but the emphasis is different. Instead of relying only on a chatbot-style response, SearchGPT is designed around web results and source links. OpenAI describes the feature as a way to provide “timely answers” to questions.
The prototype is powered by OpenAI models, specifically GPT-3.5, GPT-4 and GPT-4o. OpenAI says SearchGPT is launching today for “a small group” of users and publishers, with a waitlist available. The company also says it plans to bring some SearchGPT features into ChatGPT in the future.
Why Location Matters
Some SearchGPT results can use location. According to a support doc described in the source article, OpenAI says SearchGPT “collects and shares” general location information with third-party search providers to improve result accuracy.
The examples given are practical: nearby restaurants or a weather forecast. SearchGPT also includes a settings toggle that lets users share more precise location information.
That location layer matters because search is often local. A generic answer may be useful for broad questions, but searches about restaurants, weather, or nearby services depend on where the user is. OpenAI’s approach suggests SearchGPT is not only aimed at general research queries, but also at everyday searches where current and local information shape the answer.
The Search Market OpenAI Is Entering
SearchGPT arrives in a field that already includes Google, Bing, Perplexity and other AI-powered search tools. OpenAI’s interest in a search product had been rumored before the announcement. The Information reported in February that a product, or at least a pilot, was in development.
The timing is sensitive. AI search tools have been criticized for plagiarism, inaccuracies and content cannibalism. Those concerns are not abstract. The source article points to several examples that have shaped the debate around AI-generated search answers:
- Google’s AI Overviews suggested putting glue on a pizza.
- The Browser Company’s Arc Search told one reporter that cut-off toes will grow back.
- Genspark at one time recommended weapons that could be used to kill someone.
- Perplexity was accused of ripping off news articles from outlets including CNBC, Bloomberg, and Forbes without credit or attribution.
These examples show the challenge OpenAI faces. A search product must do more than produce fluent answers. It has to handle facts, sources and user trust in a context where mistakes can spread quickly and where publishers are watching closely.
Publisher Concerns Are Central
OpenAI is trying to position SearchGPT as a more responsible version of AI search. The company says SearchGPT “prominently cites and links” to publishers with “clear, in-line, named attribution.” It also says it is working with publishers on the experience and giving website owners a way to manage how their content appears in search results.
That publisher focus is important because AI-generated overviews can reduce the need to click through to original articles. The source article cites one study finding that AI Overviews could negatively affect about 25% of publisher traffic because article links are de-emphasized.
For publishers, the core issue is not only whether a tool names the source. It is whether the search experience still sends readers to the sites that produced the underlying information. A search answer that is useful enough may satisfy the user before they ever open the original article.
OpenAI is drawing a line between search results and model training. In its blog post, the company says, “Importantly, SearchGPT is about search and is separate from training OpenAI’s generative AI foundation models. Sites can be surfaced in search results even if they opt out of generative AI training.” It also says, “We are committed to a thriving ecosystem of publishers and creators.”
What Comes Next
SearchGPT is still a prototype, and its first audience is limited to a small group of users and publishers. That gives OpenAI room to test the product before adding parts of it to ChatGPT.
The larger question is whether OpenAI can build an AI search experience that is useful without weakening the web sources it depends on. The company is promising source links, named attribution and publisher controls. At the same time, the broader AI search category is already under pressure for factual errors and the risk of diverting traffic away from original reporting.
SearchGPT puts OpenAI more directly in competition with established search players and newer AI search companies. Its success will depend not only on the quality of its answers, but on whether users and publishers believe the product handles web content responsibly.