What happened
Google’s Generative UI used to run only in AI Mode. It is now reaching AI Overviews — the block that sits at the top of ordinary search results, with far wider coverage.
What Generative UI does is build a whole visual layout with interactive components on the spot, matched to the question, rather than returning a block of text. Google Research describes it as creating a custom interactive experience for any prompt; the back-to-school rollout added practice quizzes and study-notebook tools alongside it.
The line that matters comes from Search Engine Journal: what it generates is not limited to study aids — mortgage calculators are in there too, and sites built around calculators, converters and comparison tools will feel the overlap first.
Until now, a user finished reading a generated paragraph and then decided whether to click through to somewhere that would do the calculation. Now they can do it inside the answer.
Why it matters to you
For the past two years, plenty of sites have treated “build a small tool” as the answer to zero-click: add a calculator, a converter, a comparison table — something that makes visiting unavoidable. The logic was intuitive. A summary can copy text; it can’t copy something that needs an input before it produces an output.
Generative UI moves that line back a step.
What it takes is generic calculation: monthly loan repayments, unit conversion, BMI, exchange rates. The formulas are public and need none of your data, so the AI generates them itself, inside the answer, where the user never leaves.
Three things it doesn’t take, and the difference is where the inputs come from:
- Anything needing your proprietary data: your actual pricing bands, your stock and bookable slots, the specs only you have measured. Google can build the shell of a calculator. It can’t supply your numbers.
- Anything that depends on who’s asking: account history, order tracking, customer-specific terms.
- The transaction itself: ordering, booking, paying. Generating an interface isn’t the same as being able to take someone’s money.
So the test changes to one sentence: the tool isn’t the moat, the data feeding it is. A calculator anyone could build is now training material for Google. A calculator that only runs on your pricing logic still isn’t portable.
What to do about it
Don’t go and delete the tool pages you already have. They still catch people arriving from elsewhere, and those pages are usually among the few on your site with real structure and concrete numbers in them.
What needs rethinking is which tool you build next. If what you had planned was the kind of calculator whose formula is a search away, the return on it is not what it was six months ago, and it’s worth reordering. If you build it anyway, build it in the direction that needs your own data.
One thing you can check today: open Google, run the handful of questions your customers ask most, and see whether AI Overviews has already started producing interactive components. Rollout speed differs by region and language, so what happens in Traditional Chinese is something you have to look at yourself. Nobody can hand you a schedule, and anyone promising you one is selling something.
Working out which tools are still worth building and which have already been absorbed is harder than building one. It requires knowing how completely the AI already answers those questions, and what the user still lacks once it has. That list isn’t in your analytics.
If you want to know which layer of your site still holds, we can take a look first.
Further reading
- Summaries can take your answers. They can’t take your tools. — the answer, evidence and conversion layers, and what belongs in the third one
- Ranking in the top 10 no longer buys you a citation — the other half of the intercepted click
- GEO priorities for four site types — ecommerce, local services, B2B and media