OpenAI just launched its AI search update for ChatGPT. Three months after the company’s initial announcement of a SearchGPT prototype, OpenAI’s vision for the future of AI search is now available to the public.
“We’re focused on making ChatGPT the best place to answer any question, including live information from the web,” says Adam Fry, the product lead for search on ChatGPT. Referred to by Fry as “ChatGPT search” rather than “SearchGPT,” the feature enters an increasingly crowded and contentious field of AI search options for users—with competition from smaller startups, like Perplexity, as well as tech giants, like Google with its AI Overview search results. So far in 2024, journalists have criticized both Google and Perplexity’s implementations of AI search for improperly copying aspects of original work and hallucinating fake information.
In the lead-up ChatGPT’s search upgrade, OpenAI reached content licensing deals with multiple online publishers, such as The Atlantic, Vox Media, and Condé Nast, WIRED’s parent company. These deals allow the AI startup to use publishers’ work to train its systems in exchange for a fee. (Similar to the firewall between advertising and editorial teams, business deals have no influence on WIRED’s coverage.)
I was quite anxious about the ramifications when I first saw my writing for WIRED referenced by ChatGPT back in 2023. After a few hours of testing a prelaunch version of ChatGPT’s new AI search, it’s clear to me that OpenAI has made significant progress since its initial messy foray into web browsing, with more interactive elements and clearer attribution of its sources. I could see a subset of early adopters really latching on to the new ChatGPT search.
Taking that into account, the product needs improvements before it’s able to truly compete with the dominance of Google for essential search experiences, such as online shopping. ChatGPT also makes some of the same mistakes as other AI search tools, such as hallucinating and citing incorrect information. Curious about trying out the update for yourself? Here’s how to use it and some examples from my initial experiences.
How to Use ChatGPT’s Search Tool
To immediately try out this update, you need to pay for one of OpenAI’s subscriptions. If you have the ChatGPT Plus plan for $20 a month or ChatGPT Teams through work, the new search experience should be available. OpenAI will likely roll this out to users who have Enterprise and Edu plans sometime later in November. Free users will have to wait the longest, likely until early next year.
ChatGPT search is powered by a custom version of GPT-4o, OpenAI’s most recent generative model. It’s available to users through the ChatGPT website, mobile app (Android, Apple), and web app (Mac, Windows). In any country where ChatGPT is available, the AI search feature will likely be an option.
Screenshot: Reece Rogers via OpenAI
Fry encourages users to lean into the tool’s natural language abilities and ask multipronged questions. “This isn’t your traditional search engine, where you just do keywords and have to game what you’re asking,” he says. Also, users should interact with the source links and see what ChatGPT cites to glean a more holistic understanding of how the AI arrives at its outputs.
First Impressions of OpenAI’s New AI Search
The updated search experience might seem familiar to anyone who’s used Perplexity before. You type in a prompt, and then the AI tool gathers links from around the web to generate an output highlighting key information related to your search. So, is it really revolutionary that ChatGPT can now display up-to-date stock charts? Not on its own, sure, but within the larger context of OpenAI’s decisions, this update is emblematic of how the company sees ChatGPT as its all-purpose software.
For example, you can now have audio conversations with the bot, invite it to help you edit a document, and thoroughly search the web. The research preview version of ChatGPT running on the GPT-3.5 model is a distant memory compared to this increasingly productized version available to millions of users today.
During a demo, Fry offered an example of ChatGPT search as a potential starting point for when you’re on the hunt to find new products, and demonstrated by prompting the tool to look for WIRED’s pick for the best electric bike. The search result prominently linked to the WIRED website and WIRED commerce editor Adrienne So’s well-researched articles. “We’re really looking forward to even deeper integrations with commerce and product partners to be able to help complete that user journey,” Fry says. How affiliate revenue is shared between OpenAI and a content publisher when a reader uses ChatGPT’s search engine to shop for a publication’s recommended product picks could be a future point of contention.
As a journalist, a core task I see myself experimenting with ChatGPT for is the initial research phase for nonsensitive articles, and only as a small part of the overall research process. It’s a lower-stakes task usually completed using Google. Potentially incorporating AI search methods early in my writing leaves plenty of opportunities to catch any hallucinations that might pop up.
The internet isn’t just full of research articles and stock prices, though; explicit content drives search interest and proliferates online. But not for AI search tools—erotic content goes against OpenAI’s policies, and nudity is unlikely to appear in any of your image results. When I asked for recommendations as to which OnlyFans creators are worth subscribing to, ChatGPT’s first pick was “Jane Doe,” and her supposedly wholesome content includes workout tips and nutrition plans. A photo of a real woman, who’s casually dressed and does not appear to be an OnlyFans creator, surfaced with the result.
In an effort to test the limits of ChatGPT’s search further, I followed up with a more specific request for creators who are “male bottoms.” The software started to generate a foul-mouthed bulleted list, with real creators aggregated from a website: “Elijah is a very attractive bottom who keeps it tight, oiled up, and very hot.” But almost as soon as the words generated, OpenAI’s software struck the output as violating guidelines and deleted it. OpenAI claims it is working to improve how ChatGPT responds to violations of safeguards.
I was most disappointed to see ChatGPT surface racist and debunked information suggesting that people from specific countries are lower in intelligence. In October, a WIRED investigation by reporter David Gilbert uncovered a pattern of AI search tools citing racist and debunked IQ scores for African countries such as Liberia and Sierra Leone. ChatGPT’s search highlighted the debunked 45.07 IQ number as potentially relevant; at the same time, it also linked to David’s reporting as a counterpoint within the result.
In response, Niko Felix, a spokesperson for OpenAI, says, “Although ChatGPT acknowledges criticisms of these particular studies from sources like WIRED, there is still room for improvement in its responses.”
Despite some of the initial flaws in ChatGPT’s search update, I expect OpenAI to continue improving the user experience throughout 2025 and build upon this wave of web results. A few days before this announcement, the news leaked that Meta also has its own AI team working on search tools. While still nascent, AI search is no longer some niche part of the software market, and more companies will try their hand at it. And if user habits really do shift in the few years, then controlling the next hot info-gathering tool, with shopping and sports scores galore, is a billion-dollar business.
