Google search once felt like the quickest route to a useful answer. Now, many people describe the experience in harsher terms: “Google search has gone from being very handy to useless.” The change comes as Google pushes AI summaries into search, while chatbots such as Claude, ChatGPT, and Gemini try to answer questions in a more direct way.
That shift creates a strange problem for Google. Its AI summary can answer a question without sending someone to another webpage, which may leave fewer opportunities for the company to show ads. The criticism has already been summed up this way: “Google’s AI summary is punching a huge hole in their own profits – is this good?” Google admits AI search makes less money, but it is addressing that gap with more ads. Another claim says Google’s negative cash flow has ballooned to -$9.06 billion, with its stock price tumbling after Q3 earnings.
When AI search helps — and when it fails
Chatbots are often good at retrieving information from the web. They can search for you, combine material from several places, and present an answer that feels easier to use than a page of links. For a complex question that a standard Google search would struggle to parse, that can be a real improvement.
Imagine asking for San Francisco restaurants using several requirements at once. A traditional search would probably produce a list of blue links that rank restaurants by basic criteria. An AI system can query and collate multiple sources, then present a synthesized answer. The chatbot has never been to San Francisco or eaten there, though. It relies on public accounts written by people who have.
AI search works best when a question is common and a lot of accurate, up-to-date information exists online. If you need the height of the Empire State Building or another simple, widely known fact, AI is a safe choice for a quick look-up. It can also feel incredibly useful when the web contains plenty of material about the subject.
The risk grows when the answer depends on a rare detail. Ask for the fifth-billed actor in an obscure television show, and there might be only one or two references online. The AI may select the wrong one and present it with confidence. Research from Stanford University points to the same weakness: when AI finds an accurate source, it is very good at pulling the right information from it, but it can fail to find accurate sources and still produce a confident answer.
That is why precise searches still favor traditional Google results. A celebrity quotation, a report statistic, or a device specification may require exact details rather than a smooth summary. Heading to Google also gives you more control over the sources you use, even though classic searches can still lead to false information on the web.
Why the source matters more than the smooth answer
Researchers from UC San Diego found that AI web searches sometimes choose sources because they match the query, rather than because they rank as the most authoritative. Relevance and authority are not always the same thing. A page can contain words that fit a question while still offering a weak basis for an answer.
Traditional searching remains the safer path for breaking news and high-stakes questions about health, the law, or finances. If an answer could land you in jail, more research is necessary. When an AI approach is useful, check the sources linked inside the answer twice, and ask the chatbot where a claim came from if there is any doubt that it might be an invention.
The choice is not as simple as old search versus new AI. In both the way they process queries and the way they package information, chatbots are a major upgrade on googling. But the upgrade works best when users understand its limits and do not confuse a well-written answer with a verified one.
A web with more machines than people
The search debate sits inside a larger change: the web is filling with automated systems. For years, concerns about accounts not run by real people hovered between conspiracy theory and truth. Dead internet theory began as a conspiracy theory, but reality caught up with the idea as automated systems became active participants in online life.
In June, Cloudflare announced that bots had overtaken human traffic online for the first time in the internet’s history. Not all bots are created equal. The change is not just the number of automated systems, but what they can do. Bots once acted as passive consumers of the web; now agentic AI tools can take actions on a user’s behalf.
That change first drew public attention through bots that automatically posted on social media. It is now growing in scale with agentic tools. Matthew Prince, Cloudflare’s chief executive, said the threshold arrived sooner than he expected: “That happened faster than I predicted,” he said, adding that he thought it would be the end of 2027 and then early 2027, “but agentic traffic [is] growing so fast”.
Prince also said “crawler”, “bot”, and “agent” are all “synonyms depending on whether you want them to be a good or bad thing”. The distinction matters because automated traffic can help people find information, but it can also make it harder to know who is reading, posting, buying, or making a decision online.
Platforms want to know who is behind the screen
Reddit’s move toward “a primarily logged-in experience” shows how platforms are responding. Chief executive Steve Huffman wrote, “As someone whose life has been defined by the open internet, it’s a big change for me”. Google’s focus on AI overviews has led Reddit’s traffic to plunge, and Reddit wants to control and monetize the users who remain.
A logged-in audience gives Reddit more information about each user, which it can use to sell ads. The same pressure appears in other parts of the internet. Meta’s latest results, following the unveiling of its Muse AI assistant, said buying things was one of the top activities performed with the system. More than a quarter of those users had an average purchase size above $200.
Automated buying is already familiar in live events. Touts use bots to buy tickets and resell them for a profit, leaving people in long queues and forcing them to prove they are real. OpenAI chief executive Sam Altman’s company World takes that concern to another level by scanning people’s eyeballs as “proof-of-personhood”.
The result is a web where search tools can save time, but can also blur the line between a source and a generated answer. People may not know whether they are dealing with a human, a bot, or an agent acting for someone else. As one warning puts it, “It is easy to imagine a world in which sellers don’t know if they are giving their products to a real person, and consumers don’t know exactly why they are buying the same product”.
For now, the practical rule is simple: use AI search for speed and synthesis, but return to traditional search when precision, authority, or personal responsibility matters. The more automated the web becomes, the more valuable it is to know where an answer came from.
Based on
- Google search has gone from being very handy to useless | First Dog on the Moon — theguardian.com
- Google admits AI search makes less money, but it’s fixing that with more ads | Page 4 | Ars OpenForum — arstechnica.com
- Opinion | The Guardian — theguardian.com
- When You Should—and Shouldn’t—Use AI to Search the Web | WIRED — wired.com
- How the zombie internet will change the web – and the world | The Independent — independent.co.uk
