AI search behavior statistics: how customers find businesses now

Five years ago, if you needed a heat pump installer, an accountant or a physical therapist, you typed a few words into Google and clicked through a handful of websites. People still do that. But there's now a second path: ask an AI assistant and get a ready-made answer, often with business names in it. This article lays out the available data. Throughout, we separate what a study actually measured from what we take away from it.

How many people use AI assistants?

The best-known yardstick is ChatGPT's weekly user count. At OpenAI's developer conference in October 2025, Sam Altman said ChatGPT had passed 800 million weekly users².

Independent research confirms the scale. A National Bureau of Economic Research working paper, written by economists together with OpenAI researchers, estimates that by July 2025 ChatGPT had been adopted by about 10% of the world's adult population¹. What matters more for businesses is what people use it for. Practical guidance, seeking information and writing together account for nearly 80% of all conversations, and non-work use grew to more than 70%¹. In other words, a large share of these conversations are questions people used to type into a search engine.

Google is building AI straight into search results. According to Alphabet CEO Sundar Pichai, AI Overviews had more than 2 billion monthly users by mid-2025, across more than 200 countries and territories and 40 languages³. The standalone Gemini app grew to more than 750 million monthly active users, according to Alphabet's results for the end of 2025⁴. You can't add these numbers up, since the same person can use every service. But they make clear that AI answers aren't a niche anymore.

What about the Netherlands?

For the Netherlands, Eurostat is the most reliable source. In 2025, the EU statistics office asked about generative AI use for the first time. Across the EU, 32.7% of people aged 16 to 74 had used such tools in the previous three months⁵. The Netherlands is well above that at 44.7%⁶. Only Denmark (48.4%) and a few other countries score higher⁵.

One caveat: Eurostat measures generative AI use in general, not searching for businesses specifically. So the figure tells you about the reachable audience, not how many Dutch people ask an AI model to recommend a supplier. As far as we know, no Dutch study on that specific question has been published yet.

AI as the starting point for search and shopping

Bain & Company surveyed more than 1,100 US consumers at the end of 2024. About 80% rely on AI-written results for at least 40% of their searches, and Bain estimates this cuts organic web traffic by 15% to 25%⁷. The same research found that about 60% of searches now end without the user clicking through to another website⁷.

For businesses, the key question is what people use these tools for. Bain found that 68% of LLM users use them to research and summarize information, and 42% ask for shopping recommendations⁷. In its own survey of more than 5,000 Americans, Adobe found 39% had used AI for online shopping, and 85% of them said it improved the experience⁹.

Local businesses and service providers

For small businesses, the question is whether this holds for local services too, not just online stores. BrightLocal's annual consumer survey gives a first answer: in 2026, 45% of US consumers said they use ChatGPT or other generative AI tools for local business recommendations⁸. The same set of statistics also shows a risk. Only 68% of business contact details in ChatGPT and Perplexity matched the business's Google Business Profile⁸. An AI model can name you, but with an outdated address or phone number.

These figures come from the United States. We don't know exactly how far Dutch or other European consumers have moved. Given that AI use in the Netherlands is above the EU average, it's reasonable to expect the same direction; the exact percentage may differ.

How an AI answer shapes the choice

Usage is one thing. Influence on the final choice is another. There's less hard data on that, but the research points in a clear direction. BrightLocal's statistics page includes findings from a user study by the agency Sagapixel on how people find local service providers through ChatGPT. In that study, only 27% looked at just one business⁸. According to the same set of statistics, 39% of ChatGPT users click through to a business website when looking for local businesses, and 21% switch to Google afterward to verify the information⁸. The AI answer is often the start of a comparison, not the end.

That fits what Adobe sees for online stores. Visitors arriving from an AI assistant stay 48% longer on the site and view 13% more pages per visit than other visitors⁹. They arrive with a specific question and a short list in mind. If you're on that list, you get a serious prospect. If you're not, you're not even considered at this stage.

Trust is growing too. In Adobe's survey, 66% of respondents said AI tools give accurate results when shopping⁹. That isn't blind trust, since many people still double-check. But it does mean the first impression increasingly comes from an AI answer.

What analysts expect

In February 2024, Gartner predicted that traditional search engine volume would drop 25% by 2026 as AI chatbots and other virtual agents take over queries¹⁰. That's a prediction, not a measurement, and whether it plays out exactly that way hasn't been established. Google, for its part, says AI Overviews drive more than 10% more queries for the types of searches that show them³. Both can be true at once: people search a lot, but a growing share of answers is delivered without anyone visiting a website.

What this means for your business

The figures above measure different things in different countries, but they point the same way. Our reading:

• There's a second storefront. Next to Google's search results, there's now the answer from an AI model. If you're not in it, you miss part of the customers who are researching, even if your website ranks well on Google.
• The shortlist gets made earlier. When an AI model names three or four businesses, customers often start with those names. For a local business, that can be the difference between getting the call or not.
• Accuracy matters. If one in three AI mentions has the wrong contact details, checking what AI models say about you isn't optional.
• You won't see it in your analytics. A customer who gets your name from ChatGPT and then calls you directly doesn't show up as an AI visitor. The influence is bigger than the measurable traffic.

Here's a concrete example. Someone in Arlington asks which installer is good with hybrid heat pumps. The AI model names three businesses, each with a line about reviews or specialties. The customer looks up those three names, compares prices and calls two of them. A fourth business, with a solid website and a good Google ranking, never comes up. That business notices nothing: no visit, no bounce, no missed conversion in its analytics. Just fewer calls.

None of this means SEO no longer matters. Many AI answers draw on search results and web pages. It does mean you need to answer one more question: do AI models name my business, and if so, how? And because every AI model has its own sources and preferences, the answer differs by AI model.

How GeoReady measures this

GeoReady asks ChatGPT, Perplexity, Gemini and Claude the questions your customers ask, through the official APIs of those AI models. It runs 3× per week, with repeated runs per prompt, because a single answer tells you nothing about the average. For each prompt you see your win rate (how often you're named), your share of voice against competitors, sentiment with the exact sentence it's based on, and which sources and URLs the AI models cite. That turns into concrete actions, and GeoReady shows the measured effect after 30 days. Together, this makes up your GeoReady score. Try it free for 7 days, then Pro unless you cancel.