AI chatbots from Anthropic, OpenAI, and Google can now search the web on a user's behalf, and WIRED reports that these tools blend conventional search results with knowledge drawn from their training data. In an article published Monday, the outlet lays out which kinds of lookups suit AI search and which are better handled the old-fashioned way. The distinction, according to WIRED, comes down to how much the added AI layer helps versus how much it can obscure.
Why it matters
Search engines are increasingly folding generative AI into their results, and Google has been actively promoting an AI Mode for its search product, WIRED notes. That shift changes how people find information, but it also moves decisions about sourcing away from the user. WIRED's guidance matters because AI search can save time while making it harder to see where an answer came from, and because the tools can present a confident response even when the underlying source is wrong or missing.
Where AI search performs well
WIRED says simple, widely known questions are a safe use case. Asked for a fact like the height of the Empire State Building, a chatbot is very likely to return the correct figure, because the information is easy to find online and large language models tend to favor the most probable patterns.
Complex, multi-part requests are another strength. WIRED's example is a search for San Francisco restaurants suitable for a family of five with a budget of $200. A standard search would return a list of links ranked by fairly basic criteria, while an AI tool can query and combine several sources and produce a synthesized answer. The outlet notes that chatbots have never visited the city or eaten at those restaurants; they are drawing on public accounts written by people who have, and where no exact match exists they look for related terms such as "affordable" or "family."
Research from Stanford University, cited by WIRED, found that when an AI system locates an accurate source, it is very effective at extracting the right information from it. The failures tend to occur when it cannot find a good source — because none exists or because it misread the query — and answers confidently regardless.
Where the approach breaks down
The more obscure the question, the more scrutiny the answer deserves. WIRED gives the example of asking who the fifth-billed actor was in a little-known television show. With only one or two references available online, an AI may pick the wrong one.
Precision is another weak point. Because these systems are built to paraphrase, synthesize, and summarize, those same processes can introduce errors. When a search requires an exact quotation, a specific figure from a report, or a device specification, WIRED suggests a traditional search may serve better.
Timeliness is a further concern. WIRED reports there can be a lag in how current the data an AI pulls up is, which makes conventional search the safer choice for breaking news.
Control over sourcing is also reduced. Handing a search to a chatbot means trusting it to interpret the request, select sources, and extract facts correctly — convenient, but less transparent than reading a results page. WIRED points out that a user who knows which outlets they trust for a given subject can prioritize them in a standard search in a way an AI may not.
Researchers at UC San Diego found that AI web searches sometimes select sources because they match the query closely rather than because they carry authority, according to WIRED. A personal blog could therefore be chosen over a professional medical site if the system judges it a better fit. A person scanning a list of links would be far less likely to make that trade-off, WIRED notes.
High-stakes subjects are the clearest case for keeping manual control. WIRED advises doing more of the research yourself for searches touching on health, the law, or personal finances, where a mistake carries real consequences.
How to check an AI answer
WIRED's guidance is to verify sources whenever an AI search is used, double- and triple-checking the links the chatbot cites inside its response. If anything is unclear, users can ask the tool where a particular claim came from, which helps establish whether the model produced it rather than found it.
The outlet acknowledges that ordinary search results can also contain false information. The difference, it argues, is that a person working through a list of links is in a better position to judge what is accurate.
WIRED's overall assessment is that AI chatbots represent a meaningful improvement over conventional searching in how they process queries and assemble answers, provided users accept the caveat that verification remains necessary. The tools work best for common questions likely to be covered by plenty of accurate, current material online.
