Search used to be a map. Now it’s a vending machine.
You type a question, and Google hands you a finished answer before you see a single link. For some people, that’s the upgrade they always wanted. For others, it’s the moment Google stopped feeling like Google.
A new collection of reader stories from The Guardian makes this divide hard to ignore. It isn’t a verdict on whether AI search works. It’s a look at two very different ways people now use the same results page.
Camp One: The Answer-Takers
For these users, the AI summary is the search.
A London software engineer, Mark Johnson, said he finds information faster and can ask follow-up questions to explore a topic, and he now rarely clicks website links. Consultant Alan Caruana likewise prefers the analysis and chat functions to working through a ranked list of sites, which he finds slow and frustrating.
That’s a real shift in behavior. The ten blue links were never the goal. They were homework. Answer-takers are happy to skip it.
But even inside this camp, trust has limits. A Dublin reader welcomed the way AI guides a search but warned that it is not trustworthy. Convenience and doubt now live side by side.
Camp Two: The Source-Checkers
The second group treats the AI answer as noise to scroll past.
Ewan Herbert, an English teacher in Japan, now has to scroll below the AI response to reach a full article or discussion, and he wants to check whether the summary has made something up. A Manchester reader objected to being unable to turn off AI summaries, saying he wants to judge which results are useful for himself.
That complaint cuts deeper than design. It’s about control. These users don’t want a machine deciding which parts of the web matter.
Some aren’t waiting for Google to fix it. One reader disabled AI Overviews with an extension, while another stopped using Google and switched to a search engine without them. If you’re in this camp, for more on how to turn off Google AI Overview walks through the methods that still work in 2026.
The Hidden Cost of Leaving Google
Here’s the detail most coverage skips.
A researcher in Switzerland said her replacement search engine performs worse for searches in languages other than English and for image search.
That matters far beyond Switzerland. For users searching in Urdu, Arabic, Hindi, or dozens of other languages, Google’s index is still hard to beat. Opting out of AI Overviews often means opting into weaker results. The escape hatch exists, but it isn’t free.
Why Trust Keeps Breaking
The skeptics aren’t being paranoid. AI Overviews have a track record.
Earlier this year, a Guardian investigation found that some AI summaries contained inaccurate health information, putting people at risk of harm. Google later removed some health-related queries from the feature. Google, for its part, argued that AI Overviews’ accuracy is on a par with other Search features, including featured snippets.
The core problem is structural. A list of links shows you where information comes from. A summary hides that behind a confident paragraph. When the paragraph is wrong, you have no easy way to know.
The Click Was a Receipt
Every time you opened a website, you created a small receipt. It told the publisher someone valued their work. It also let you inspect the source yourself.
AI Overviews quietly remove that receipt.
The Guardian points to academic studies showing people navigate less to outside sites when an AI Overview gives a satisfactory answer, even when it cites those sites. Large US media companies are now suing Google, alleging its summaries divert visits and cut revenue.
Google disagrees. It says it has seen no decline in aggregate traffic, and Search vice-president Liz Reid wrote last year that total organic clicks had stayed relatively stable year over year.
Both sides can be partly right. Total clicks can hold steady while clicks on AI-answered queries collapse. The averages hide the losers. Digiday’s analysis found something similar: visibility in AI answers doesn’t guarantee clicks.
What This Means for Publishers
If answer-takers keep growing, publishers can’t win by being summarized. They win by being checked.
That means content the source-checkers actively seek out: original reporting, lived experience, depth a summary can’t compress. It also means being the passage AI systems choose to cite. Research covered in our piece on why B2B brands may be invisible in AI answers shows a narrow, specific answer from a small company beat a broad, generic page from a much bigger competitor across a large dataset. Google seems aware of the tension. In May, it announced updates to AI Mode and AI Overviews promising relevant article suggestions, direct links within responses, and previews of websites and personal perspectives. Whether that brings clicks back is still an open question.
Our Take: Search Now Has Two Modes
The Guardian’s readers aren’t arguing about AI. They’re revealing that search has split into two jobs.
One is comfort mode: give me the answer and let me move on. The other is trust mode: show me the evidence. The Reuters Institute predicted exactly this split, citing a forecast that audience behaviour will split into two modes: comfort mode and trust mode.
Google’s real mistake isn’t building AI Overviews. It’s forcing one mode on everyone. A simple, permanent toggle would respect both camps.
Until then, the most important design decision on the internet is also the simplest one: what appears first, the answer or the source.
