It has not displaced classic search yet, at least not in visitor numbers. What has shifted is the expectation: in a YouGov survey of 509 business decision-makers in Germany, 55 percent expect that in three years, searching via AI tools will matter more for winning customers than classic Google search, and 54 percent already see the change today. Measured in traffic, the share is still tiny. It only gets interesting once you stop counting visitors and start counting enquiries.
In visitor numbers, AI search has not displaced classic search: around half a percent of visitors come from AI answers, around 22 percent from Google search. But those few visitors turn into an enquiry more than six times as often. What has mostly shifted is the expectation, and you only see the shift if you break enquiries down by source instead of counting sessions.
- The survey measures expectation, not displacement: 55 percent see the switch happening within three years, 28 percent disagree.
- In traffic, AI currently accounts for around half a percent of visitors, measured across an agency's B2B client projects.
- Those visitors produce a disproportionate number of enquiries, more than six times as often as Google search does.
- Large companies count on the change more firmly than small ones; owners are the most sceptical group.
- On the user side the shift has gone further: half of internet users in Germany at least sometimes reach for an AI chat instead of a search engine.
What the survey measures and what it does not
The figures come from online interviews with 509 business decision-makers between 16 and 24 July 2026, collected by quota and weighted, run by YouGov for the Munich online marketing agency Seokratie. What was asked for was an assessment, not revenue and not a channel share. Anyone who reads “55 percent” as “AI search brings in half of all new customers today” is reading the study wrong.
Something else in it does hold up, though: the respondents are describing their own behaviour. 55 percent use AI tools at work at least several times a week to research products or service providers, 23 percent daily or almost daily, and only 14 percent never do. In B2B, then, the person on the other side of the buying process is someone who has long since been asking rather than searching.
Where the shift is measurable today
The most telling numbers in the release are not the survey results but the traffic data from Seokratie's B2B client projects. They are not representative, but they show a pattern you can check on your own website.
| Channel | Share of website visitors | Of those, submit an enquiry |
|---|---|---|
| Classic Google search | around 22 percent | around 1.5 percent |
| AI answers (ChatGPT & co.) | around 0.5 percent | around 10 percent, one in ten |
Averages across B2B client projects, published on 17 September 2026. Enquiry means a form, an email or a phone call. A visit from an AI answer is therefore worth more than six times a visit from Google search.
That explains why two people can judge the same situation completely differently. Look only at the visitor share and you wave it off. Break enquiries down by source and you are looking at one of the most efficient channels in the house.
On the demand side the shift has gone further than the traffic suggests. According to a telephone survey by Bitkom of 1,156 people aged 16 and over (Berlin, November 2025), half of internet users at least sometimes use an AI chat instead of classic search, 7 percent mostly and 5 percent exclusively. Among 16- to 29-year-olds, only 30 percent still rely on search engines alone. Research is moving faster than the click, because many AI answers no longer produce a click at all.
Why large companies count on it more than SMEs
Expectation depends heavily on company size. In companies with more than 250 employees, 64 percent assume AI search will matter more than Google in three years; in companies with fewer than 50 employees, 43 percent do. The most sceptical group is owners and partners: only 36 percent agree and 41 percent disagree, while among salaried executives roughly 60 percent agree. And 27 percent of owners never research via AI tools, twice the average.
That fits an earlier read-out of the same survey, published on 29 July 2026: 58 percent expect companies without a presence in AI answers to lose customers, but only 20 percent know what generative engine optimisation means, that is, the deliberate work of getting cited in AI answers. Among small companies, 33 percent consider themselves well prepared; among large ones, 52 percent.
For small providers that is an opportunity rather than a threat. A results page has positions two through ten; an AI answer, as Seokratie founder Julian Dziki puts it, often has no second place. Answer a question properly and you stand there next to far bigger brands. What a page needs in order to do that, we have summed up in What is GEO?
What you can do in the next four weeks
- Report AI traffic separately. Set up a segment in your web analytics for referrers such as chatgpt.com, perplexity.ai or copilot.microsoft.com. How to combine that with measuring mentions directly is covered in referral traffic from AI sources.
- Judge enquiries, not visitors. Work out the enquiry rate per channel, not the session share. Only that number tells you whether a small traffic share is noise or your best channel.
- Collect the twenty real questions. Not keywords, but whole questions the way customers type them: “What does custom software cost for a 50-person company?” Sources are sales emails, support tickets, forums. Those are exactly the questions our Scouting collects from eight sources, among them Google and Brave queries, Reddit, gutefrage, LinkedIn, X and YouTube. The free plan includes a one-shot question research run; weekly scouting is part of the paid plans.
- Close the gaps. For every question, check whether your website answers it fully, and write the missing answer as its own section: the direct answer first, then use cases, technical data and prices where relevant. Our Answer Check does this automatically and names the passage down to the section; for partial answers it says what is missing.
If you want to try it first: the free plan costs €0 and includes 1 project, 50 URLs across all projects, 1 free article and a one-shot question research run, no credit card.
Does this mean classic SEO is finished?
+
No. In the same survey, 71 percent agree that SEO remains important for their visibility in the age of AI, the highest level of agreement in the whole study. Simple queries still run through Google. GEO complements SEO, it does not replace it.
Does this apply equally to every industry?
+
No. The more complex a product and the more explanation it needs, the more likely prospects are to research it with AI systems, for example with custom software, industrial machinery or IT service providers. The question is not either-or, it is where your customers will be searching next.
How reliable is a survey of 509 respondents?
+
For a decision-maker survey that is a normal order of magnitude. The sample was collected by quota and weighted, representative by company size, gender, age, six NACE industry groups and Nielsen regions. It measures the opinions of decision-makers in Germany, not market shares.
Why do I see so little AI traffic in Google Analytics?
+
Because many AI answers end without a click. Your brand being named in an answer shows up in no web analytics tool, and it still has an effect. That is why you need a second measurement alongside referral traffic: how often are you named or cited in the answers to your core questions?
How quickly do I have to react?
+
The survey names a three-year horizon, but your content needs lead time before AI systems read and cite it. Starting with the twenty most important questions makes more sense than waiting for a channel shift to show up in your own statistics.
