When OpenAI launched SearchGPT, demonstrations suggested that everything about how people search for things online would change instantly and forever. But “wow” became “wow, that’s embarrassing” when examples of the AI search engine in action were shown to be somewhat flawed. The challenge to Google’s reign as search engine king is still being rethought.
According to a new part in The Washington PostSearchGPT is still shaky on the facts. Google may not have to worry about losing its digital search throne anytime soon, even if it rapidly implements its own AI search tools.
The problems aren’t hard to understand. SearchGPT is supposed to merge OpenAI’s AI models with real-time web data for faster, more accurate answers. Questions and keywords provide a summary of the requested information instead of the standard Google links. It can be fast and informative. Unfortunately for OpenAI, that initial factual error is starting to look more like the rule than the exception. As the After pointed out, early testers saw SearchGPT claim that OpenAI CEO Sam Altman would be speaking at a tech conference in the near future that he wasn’t actually attending. It’s a hallucination that’s just as bad as anything ChatGPT makes up.
And even if SearchGPT were guaranteed to tell only the truth, that’s no good salve if it has no way of answering your questions. Tests that are done with the After SearchGPT’s ability to help with local information in particular has been discredited. That information has to come from somewhere. Google’s decades of refined data on a vast number of companies and the products and services they provide makes it a breeze to find most of the information people want about the places around them. And if there’s something it doesn’t already have in its databases, its partners and subsidiaries can probably fill it in for them. SearchGPT and OpenAI don’t have that kind of database access, so the answers are either nonsense or nothing.
AI Search Parties
OpenAI’s prominence in the AI scene makes SearchGPT stand out, but the idea of mixing AI and search isn’t exclusive. Google’s aforementioned ambitions include AI Overviews and answers relayed by its Gemini AI assistant. Of course, Google’s AI search tools have had their own issues, with the occasional bizarre and downright dangerous piece of advice. Not that that stopped Google from going global with AI Overviews after some revision.
OpenAI is still cautious, and has only released SearchGPT to a very limited number of users. Those eager to experience how generative AI models can transform online search have other options, from well-funded startups trying different strategies. Perplexity AI is one of the most prominent. Rather than creating an AI model and then grafting it onto a search engine engine, Perplexity leverages existing models built by OpenAI, Anthropic and other developers. Perplexity sidesteps hallucinations with stricter guardrails, insisting on including links to support what the AI writes, an innovation it adopted very early. Perplexity built its own web index while still using Google and Bing to fill in the gaps, while SearchGPT apparently relies more on outside data. The result is a more structured approach that sidesteps some of the bugs that reportedly plague SearchGPT.
Google’s long-term approach to search doesn’t seem likely to face any formidable opponents, not even from itself. Whoever creates an AI search engine will have to at least come close to matching Google’s speed, precision, and accuracy to entice the Google-loving world. Given the years and billions of dollars it’s taken Google to reach its current heights, OpenAI may have to be a little more patient or a lot more wasteful in its spending to achieve the same. And those resources will likely take more than just ChatGPT’s subscriber base to reach
The stumbling blocks for SearchGPT underscore the broader challenges that generative AI tools face when trying to compete with established search engines. While AI tools like ChatGPT and SearchGPT can offer impressive conversational capabilities, they lack the deep, structured, real-time data needed to handle everyday searches at scale. SearchGPT may be on the right track, but finding the ideal AI search engine will take some more searching.