Nearly half of American adults have used ChatGPT. One in five trusts what an AI chatbot tells them about the news. Both of those are 2026 numbers, and the gap between them is the whole story of AI search.
People switch because the assistant is faster, it talks back, and it does the reading for them. Belief has little to do with it, which is why a large share of them open Google afterwards to check.
This post pulls together the survey research published between mid 2025 and mid 2026 on why people use AI search, which tasks it has taken from Google, where it breaks, and what that means if you run a business that depends on being found.
Who is actually switching
The adoption numbers are no longer niche. Pew Research Center's February 2026 survey of 5,119 US adults found that 44 percent had used ChatGPT, up from 34 percent a year earlier and 18 percent in 2023. Among 18 to 29 year olds it was 61 percent. Among the over 65s, 19 percent (Pew Research Center).
Gemini came next at 24 percent, then Copilot at 17 and Meta AI at 14. So the picture is not one product. It is a cluster of assistants, with ChatGPT far ahead and the rest fighting for second.
Google's own figures are bigger because they are counting a default surface rather than an app people chose. In May 2026 Google said AI Mode had passed a billion monthly active users and that queries had more than doubled every quarter since launch (Google).
In the UK the shift arrived mostly through the back door. Ofcom's Online Nation 2025 report, as summarised by Mediaworks, put AI summaries on around 30 percent of Google searches, with 53 percent of UK adults saying they frequently see them, and counted 252 million UK visits to ChatGPT in August 2025 alone (Mediaworks). Most people did not decide to use AI search. Google decided for them.
Two caveats before anyone extrapolates. Pew's companion study found 51 percent of US adults still never use a chatbot at all, and two thirds of those non-users say they are unlikely to start within a year. The top reasons are not interest (60 percent call it a major reason), privacy (54 percent) and not trusting the accuracy (45 percent) (Pew Research Center). Half the market is still waiting to be convinced.
The tasks AI search has taken from Google
Usage is one thing. What matters for a business is which jobs people hand to the assistant.
Google's AI Mode data is the clearest window on this. The average AI Mode query is three times longer than a traditional search. Planning queries grew 80 percent faster than AI Mode overall across six months, brainstorming queries 30 percent faster since launch, and the fastest growing openers are "where to", "where should I" and "ideas for" (Google). Follow-up questions in the US rose more than 40 percent a month.
That is research and decision work. Comparing options, planning a trip, working out which of three things to buy. Not "opening hours" and not "plumber near me".
Local discovery is the surprise. Yext's March 2026 survey of 1,120 US adults found 47 percent had used AI to find a local business in the past month, and that among households earning between 150,000 and 175,000 dollars AI had edged past Google as the starting point, 53 to 49 (Yext). That was unthinkable two years ago and it is exactly the kind of query a Cyprus dental clinic or a UK law firm lives on.
News is smaller but growing. The Reuters Institute's 2026 Digital News Report puts weekly use of AI chatbots for news at 10 percent globally, up from 7 percent, and 16 percent among under 35s (Reuters Institute). Notably, growth stalled in the US, UK, France and Germany, and doubled in Greece, South Korea and Spain.
What people say they value
Strip the surveys down and three things keep coming up.
Speed is the obvious one, but it is not speed of loading a page. It is speed of not having to read six of them. Pew's browsing-panel study found that when an AI summary appears, users are far more likely to end their session on the results page: 26 percent of pages with a summary, against 16 percent without (Pew Research Center). The summary did the job, so they left.
Synthesis is the second. Reuters found the most common uses of chatbots for news were interpreting a story, summarising it, and making it easier to understand. People are paying for reading to be done on their behalf.
Follow-up is the third, and in our experience it is the one that changes behaviour permanently. When Reuters asked chatbot users what they valued most, the top answer at 42 percent was the ability to ask follow-up questions. Google's 40 percent a month growth in follow-ups says the same thing from the product side. Once someone has had a conversation about a decision, a list of ten blue links feels like being handed a phone book.
Where trust breaks
Here is the number that should temper every "search is dead" deck. Reuters puts trust in AI chatbot answers at 20 percent globally, against 37 percent for news in general. In the UK it is 6 percent (Reuters Institute).
People use the tool and do not believe it, and the behaviour that resolves that contradiction is verification. Yext found 74 percent of AI users rate their trust in AI recommendations at 4 or 5 out of 5, yet only 7 percent act on a recommendation without doing anything else. The rest check: 62 percent search Google straight afterwards, 58 percent go to the business's own website, 52 percent click the sources the AI cited, and 28 percent look for reviews (Yext). Trust level did not predict whether someone verified, only how.
The same survey found 57 percent still prefer a traditional search engine for personal, medical or financial topics. The higher the stakes, the faster people fall back to a source they can see.
Citations are the weak link. In Pew's data, Google users clicked a link inside an AI summary on just 1 percent of visits, and clicked a traditional result on 8 percent of visits when a summary was present, versus 15 percent when it was not (Pew Research Center). Reuters found only 42 percent of chatbot news users always or often click through to the original source, ranging from 56 percent in South Korea to 26 percent in Denmark.
So the citation is doing reputational work rather than traffic work. Being cited tells the user you exist and are considered credible enough to quote. It does not, on its own, bring them to your site. We wrote about the practical side of that verification step in AI search trust is falling while use keeps rising.
What stays on Google
The AI search share of total search is still small once you measure it properly. SparkToro's analysis of Similarweb panel data for January to April 2026 found AI Mode accounted for 0.34 percent of US Google searches, while AI Overviews appeared on more than 20 percent of them and cut click-through by roughly 60 percent when present (SparkToro). Zero-click searches hit 68 percent.
Read those together. The damage to clicks is coming from AI inside Google's own results page, not from people leaving for ChatGPT. Gartner's February 2024 forecast that traditional search volume would fall 25 percent by 2026 as people moved to chatbots (Search Engine Land) has aged as a directional guess about clicks, not as a prediction about where queries go.
Pew's data also shows which queries never trigger an AI answer in the first place. Only 8 percent of one or two word searches produced a summary, against 53 percent of searches of ten words or more and 60 percent of searches that began with a question word. Short navigational and transactional queries, brand names, "near me" searches and anything with a map, a price or a booking button behind it still resolve the old way. We went deeper on that split in the research versus buy query divide.
Our reading is that Google is holding the bottom of the funnel and losing the top of it. The exploring, comparing and planning is drifting into conversations. The moment of choosing a specific business, checking it, and paying is still a search, a website visit and a review page.
The key figures in one place
| Finding | Figure | Source and date |
|---|---|---|
| US adults who have used ChatGPT | 44% (61% of 18 to 29s) | Pew Research Center, survey Feb 2026 |
| US adults who never use a chatbot | 51%; 45% cite accuracy as a major reason | Pew Research Center, June 2026 |
| Google AI Mode monthly users | 1 billion; queries 3x longer; follow-ups up 40% a month (US) | Google, May 2026 |
| Google searches showing an AI summary | 18%; clicks on results 8% vs 15% without; 1% click a summary link | Pew Research Center, data March 2025 |
| US Google searches with no click | 68%; AI Mode 0.34% of searches | SparkToro, Jan to Apr 2026 |
| Weekly use of AI chatbots for news | 10% globally, 16% of under 35s; trust 20% (UK 6%); 42% click through to sources | Reuters Institute, June 2026 |
| US adults who used AI to find a local business in the past month | 47%; 62% then search Google, 52% click cited sources, 7% act without checking | Yext, March 2026 |
| Prefer traditional search for medical or financial topics | 57% | Yext, 2025 survey |
| UK adults who frequently see AI summaries | 53%; ChatGPT 252m UK visits in Aug 2025 | Ofcom Online Nation 2025 via Mediaworks, Dec 2025 |
What a business should change
Given all of that, here is what we tell clients to do differently, in rough order of how much it moves.
Write for the verification visit, not just the discovery one. More than half of AI users will land on your site to check what the assistant told them. That page needs to confirm the claim fast: the price, the location, the credential, the thing the AI said you do. If they have to hunt, you have lost the one click AI search was ever going to send.
Treat citations as brand exposure and measure them that way. A citation that gets clicked 1 percent of the time is not a traffic channel. It is a mention, and the follow-on behaviour is a branded Google search or a direct visit. Watch branded queries and direct traffic alongside AI referrals or you will conclude nothing is happening while your name is being said more often. This is the same pattern behind impressions up, clicks flat.
Cover the planning and comparison questions, because that is the traffic that left. "Which", "where should I", "is X worth it for Y" queries are growing fastest in AI Mode and get absorbed into an answer. Being the source that answer is built from requires content that actually compares and recommends, which most service pages do not. Our content strategy work now starts from those questions rather than from head terms.
Keep the transactional layer boring and correct. Google Business Profile, reviews, structured data, opening hours, a fast page with a phone number. That is where the 62 percent who go back to Google end up, and it is still local SEO doing the work. Semrush's own study found the average LLM visitor converted at 4.4 times the rate of an organic visitor (Semrush), which makes sense once you realise those visitors already did their research elsewhere.
Do not chase the whole assistant market at once. ChatGPT at 44 percent and Gemini at 24 percent are the ones your customers are in. Be visible in those two, check Perplexity and Copilot if your sector is technical, and stop there. Our AI search optimisation programme is built on that ordering, and the AI Mode playbook covers what changes on the Google side specifically.
Put the surveys together and you get a fairly unromantic picture: people use AI search because it is quick and conversational, distrust it enough to check, and go back to Google and to your website to do the checking. If you want to know which of those visits your site is currently winning and losing, a free SEO review is the quickest way to find out.
