OpenAI is building SearchGPT search engine, as new AI battlefronts emerge
A policy change de-links any publisher or content creator’s contribution to training OpenAI’s gen AI models, with their links appearing in search results.
This is perhaps not entirely unexpected, and though artificially intelligent (AI) chatbots do have a significant role to play as search tools for users, OpenAI isn’t letting it be as is. The company has announced prototype version SearchGPT, an AI search engine that’s designed to rival Google Search, and pretty much every other search engine that’s available. This battle would be interesting to watch as it unfolds, because Google too has been giving its Search engine a serious AI upgrade, in recent months. As has Perplexity.
In case you are wondering what the core difference is between SearchGPT and asking the ChatGPT chatbot the same query, the reasons are two-pronged – first, the information served in response to a query will be as time relevant as possible with a relevant link if you would like to read more, and secondly, there will be context for follow-up queries in the same search. From the demo that HT has seen, there is a significant visual element to search results, complete with mentioning the source of those results. HT has not been able to test SearchGPT yet.
OpenAI doesn’t say if there will be any integration with existing search engines, such as Microsoft Bing. “Getting answers on the web can take a lot of effort, often requiring multiple attempts to get relevant results. We believe that by enhancing the conversational capabilities of our models with real-time information from the web, finding what you’re looking for can be faster and easier,” says OpenAI, in a statement shared with HT.
A few weeks ago, AI company Perplexity announced updates for its Pro Search engine, with conversational search as well as integration with the Wolfram|Alpha engine for academic research. OpenAI’s directly competing with Google Search too, though they’ve obviously not said as much, comes after months of Google layering its Search engine with AI.
This summer, at the Google I/O keynote, we saw the first glimpses at expanded AI Overviews, AI-organised search results using Gemini’s multi-step reasoning capabilities and getting troubleshooting help via a video upload.
Access to SearchGPT is limited for now, and there’s a waitlist you have to join at present. This is still very early in the search engine testing stage, but OpenAI is insistent that this process will bear fruit, with a “plan to integrate the best of these features directly into ChatGPT in the future.” For now, a limit number of users and a small group of publishers are on board the trials.
“SearchGPT is designed to help users connect with publishers by prominently citing and linking to them in searches. Responses have clear, in-line, named attribution and links so users know where information is coming from and can quickly engage with even more results in a sidebar with source links,” the company further details the product. OpenAI is insisting the plan is to build a knowledgebase for SearchGPT to be able to deliver better localised information.
In recent months, AI companies including OpenAI have faced increasing criticism for using data from publishers, content creators and artists, to train their models without explicit consent. For SearchGPT, OpenAI says they’ve de-linked any publisher or content creator’s approval or opting out of contributing to training OpenAI’s generative AI models, and the same publishers’ links appearing in search results.
In use for the search engine will be the OAI-SearchBot crawler, which will find the links and the content residing in these links, to surface in search results. This has, they say, no link to any training data sets. This is separate from GPTBot, which crawls web pages for content to put into the data sets for training the company’s AI models.