News · 2026-09-23
Google puts AI Overviews between users and the web
@neuronium_ai
Google has moved AI summaries to the front of its search experience, making users scroll past a chatbot answer before reaching the familiar list of links. The change matters beyond interface design: research suggests that when an AI Overview answers well enough, people are less likely to visit the sites it cites. For publishers whose advertising depends on those visits, Google’s answer-first search is becoming an economic dispute as well as a change in habit.
What changed
Google still shows links, but AI Overviews now occupy the first layer of the search experience. In May, the company said it planned to add more AI features.
Liz Reid, Google’s vice-president of search, wrote in August last year that “AI is providing the biggest improvement in the history of Google Search.” Google says it has not seen an overall decline in traffic and describes contrary claims as based on “flawed methodologies.” Reid also wrote that year-on-year organic traffic from Google Search to websites had remained relatively stable.
The evidence from academic research points to a more complicated outcome: a sufficiently good summary can keep users on Google even when it names the sites behind the answer.
That is why several large American media companies have sued Google. They argue that AI summaries take information from websites while diverting or reducing the traffic those sites rely on for revenue.
The practical change is simple:
Users see a synthesized answer first.
Links remain available, but require more scrolling.
Follow-up questions can replace visits to multiple sites.
Publishers risk losing the visits that support their advertising businesses.
The Guardian asked readers whether Google had improved in the AI era and how their search habits had changed. Their answers split into three distinct groups.
The users who prefer the new Google
For some readers, the link list was never the product. It was an obstacle between them and the answer.
Sionn Ahog, a 76-year-old retired teacher from Dublin, prefers Google because it guides users toward an answer instead of making them search independently. He trusts AI to look for information, while warning that Google AI and similar systems are not reliable. He said one bot once admitted that it had lied to him and apologized.
Peter Warb, 75, from the UK, finds AI answers more informative than Google’s earlier approach of directing users to other sites.
José Garcia Jr, a 40-year-old recovery coach from Dallas, now uses AI mode for every search and says it has improved Google.
Alan Caruana, a 62-year-old consultant from Malta, prefers analysis and chat features to a ranked list of sites that he would otherwise have to work through.
Eleazar Eke, a 67-year-old electrical engineer from Onitsha, Nigeria, says search is faster, more realistic and more accurate.
Mark Johnson, a 52-year-old software engineer from London, rarely opens links now. He says follow-up questions help him understand a subject more deeply.
Samuel Lieberstein, a 90-year-old retired teacher from Cherry Hill, New Jersey, says AI is better at finding rare information, comparing material and making complex knowledge accessible.
This group is not necessarily asking Google to be a better directory. It is asking Google to do more of the research.
The users who want the links back
The opposing complaint is not that AI answers are always useless. It is that they stand between users and the sources they still need to inspect.
Sean Peters, a 41-year-old records manager from Portland, Oregon, says Google AI hides sources, takes credit for information it did not create and diverts traffic from the sites it uses.
Ewen Herbert, a 33-year-old English teacher in Japan, now has to scroll through every search to reach the actual site, read the full material or check whether the summary invented something. When especially frustrated, he adds -ai to his query.
Chris Gibson from Dundee considers search worse and disabled AI Overviews with an extension.
John Wilkinson, a 43-year-old personal trainer from Manchester, calls the change unequivocally bad because he wants to decide which results are useful and relevant himself. He says Google no longer lets him disable the summaries.
The disagreement is really about who gets to perform the first act of judgment. Google’s preferred version asks its system to select, compress and connect the material. Its critics want the user to see the underlying choices before accepting a conclusion.
When users stop starting with Google
The strongest response is not a complaint about an extra scroll. It is changing the front door.
Sofia, a 30-year-old physics researcher from Lausanne, misses the old Google Search and now uses another search engine. It works well in many cases but is worse at finding information in languages other than English, and she considers its image search weaker than Google’s earlier version.
Others have built their own search sequences:
Craig Garrett, a 52-year-old book editor from New York, starts with DuckDuckGo, though he often finds Grokipedia pages and other “junk” among the first results.
Guy Arthur Simpson, a 66-year-old writer from Spain, no longer uses Google and chooses a search engine without AI Overviews.
Dan Jennings, a 54-year-old IT specialist from Southampton, deliberately chose a browser that does not put AI results first and now uses Ecosia.
Catherine O’Leary, a 17-year-old student of law, politics and sociology in Newcastle upon Tyne, considers the AI features a nightmare. She tries to avoid AI search because she does not want to accidentally use an AI data center that consumes water if she forgets to add the right suffix to her query. For research, she turned to her local library’s case-law database because she trusts it more.
Pedro Reyes, 61, from Galicia, Spain, no longer begins with Google. He starts with Wikipedia, moves to DuckDuckGo with AI disabled, then Reddit, and uses Google only last.
My read is that Google is optimizing for answer completion while many of its users still need source discovery. Those goals overlap for quick factual searches, but they diverge sharply when the user needs a complete article, local reporting, historical context or a source they can trust.
What I’d want to know is how often Google’s AI answers send meaningful traffic back to the sites they cite, rather than merely displaying their work inside Google. The company’s claim that overall organic traffic remains stable does not resolve that narrower question—and the users leaving Google suggest that trust may be harder to preserve than traffic totals reveal.
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
