For years search engine optimisation ran on a simple rule of thumb: put something useful on your website — a calculator, a converter, a comparison table — and you earn clicks for it. Google could answer the question, but it could not replace the tool. That is exactly the boundary now moving.
On 19 August 2026 Google announced that generative UI would extend beyond AI Mode into AI Overviews, the considerably larger of the two AI surfaces in search. Generative UI builds a bespoke layout for each question in real time and, where Google's model considers it useful, an interactive tool or simulation as well. Not as a link. As part of the answer.
1. What generative UI is — and what is new about it
Generative UI means search no longer only writes text, it builds an interface. From components such as charts, tables, timelines, comparison views and simulations, Google assembles a layout that fits the specific question at the moment it is asked. The answer is then not a block you read and leave behind, but something you can operate.
Google's own examples make this tangible. A search for "pH scale" returns an interactive scale in the AI Overview. A follow-up question — about citrus fruits, say — produces a more specific variant with those fruits placed on the same scale. And in AI Mode Google has already shown a mortgage calculator that lets you compare terms directly inside the answer.
The decisive difference: in a classic AI Overview the summary was the product and the click was the deeper look. With generative UI the interaction itself can happen in the answer. What used to be the reason for the click becomes part of the search result.
2. The rollout: from AI Mode into AI Overviews
The chronology matters, because it shows how predictable this development is. Generative UI first appeared in AI Mode in November 2025, alongside Gemini 3. At the I/O developer conference in May 2026 Google announced it would make the feature available free to all search users over the course of the summer. The announcement of 19 August is the next step: confirmed in AI Mode, now beginning in AI Overviews.
What Google has actually committed to — and what it has not
Committed: in AI Mode the feature is available globally in English, and in AI Overviews the rollout has begun and is meant to be complete "in the coming weeks". Not committed: which queries trigger generative UI at all, when German-language results follow, and how often search really serves an interactive element instead of text. Anyone promising you firm numbers on that today is guessing.
The occasion was school — the use case is not
Google wrapped the rollout in a back-to-school announcement, together with practice quizzes, Lens coaching and study notebooks. That framing is easy to misread. The mortgage calculator on show is not a learning tool, it is a classic commercial query. The educational packaging says little about where the feature bites economically.
3. Who it hits: calculator, comparison and definition pages
The overlap appears wherever the entire value of a page sits in a function Google can generate itself. A page that does nothing but multiply three numbers and display a result has little to hold against a widget built inside the answer. A page backed by your own data, real prices, concrete availability or a named contact has a great deal.
The assessment below is a schematic one from Grünberg.Digital., based on the examples Google has shown (the interactive scale and the mortgage calculator). It is not Google data — Google has not published which queries trigger generative UI. Status: August 2026.
High overlap
At risk are pages built on pure calculation logic without a data basis of their own: interest and loan calculators using standard formulas, unit converters, calorie or consumption calculators, price comparison tables with no prices of their own. The same goes for short definition pages that answer a question in three sentences — the classic prey of AI Overviews, now presented visually as well.
Low overlap
What stays robust is everything Google cannot know. A configurator that reflects your real product variants and delivery times. A calculator that works with your terms and ends in an enquiry. A service page with references, prices and a person who calls back. Search can build an interface — it cannot build your offer.
4. Why this matters to smaller companies now
The most common reflex runs: English only, still rolling out, it will reach us next year. That is comfortable, and risky, for three reasons.
First, the sequence is by now well known: AI Mode first, AI Overviews later, English first, other languages later. Google has already run through that chain once with AI Overviews themselves. The lead time is months, not years.
Second, AI Overviews are the larger surface. AI Mode is a separate mode you have to choose; AI Overviews appear in ordinary search. Whatever moves in there reaches vastly more people.
Third, AI visibility moves faster than annual planning cycles can capture. One example from the same month: Reddit's share of citations in ChatGPT search fell, according to analysed data, from 3.83 per cent in the period from 18 July to 7 August to 0.52 per cent between 14 and 17 August — a drop of 86.4 per cent within a few days, with no confirmed cause. If your visibility rests on a single surface, you have no safety net.
Our assessment: generative UI is not a reason to panic, but it is a clean prompt for taking stock. The question is not "when does it reach Germany?" but: which of our pages live purely on a function Google could offer itself within twelve months?
5. Four levers against the generated answer
Your own data instead of general logic
A calculator that applies a publicly known formula is replaceable. A calculator that draws on your terms, your delivery times, your industry benchmarks or your inventory data is not. Test every tool on your website with one question: could a model produce the same result without access to our data? If yes, extend the data basis — or accept that this page will carry less traffic in future.
From result to next step
A tool that ends in a number has done its job and releases the user. A tool that ends in a next step — request a quote, book an appointment, receive an individual analysis — gives someone a reason to stay on your site. That step is precisely what a generated interface in search cannot offer.
Write to be quotable
As long as the answer is generated, the quality of the source decides whether it gets named. Clear statements, figures with a date and a provenance, clean structure, visible authorship and structured data are the currency. None of that is new, but under generative UI it matters more, because less room is left for many sources. The groundwork is the same as for generative engine optimisation generally.
Build demand beyond search
The most durable protection against changes in the search result is demand that does not originate there: a newsletter with real open rates, a recognisable brand, customers who recommend you, a presence in your industry's professional channels. Channels you own do not get re-rolled-out.
6. Preferred Sources: the button that bypasses the algorithm
Almost at the same time Google rolled out a second change that gets lost in the discussion but is the more practical lever for companies: an embeddable Preferred Sources button. Anyone who clicks it on a page sets that website as a preferred source and lands back at the same point in the text.
According to Google's documentation, sites selected this way appear more readily in Top Stories and can be labelled accordingly in AI Overviews and AI Mode. Google says more than 600,000 sources have already been selected — in May the figure was over 345,000. The embed code is available in the Search Central documentation.
Why this is interesting: the lever runs through your audience, not through the algorithm. Once set, the preference follows that person across Top Stories, AI Overviews and AI Mode. For companies with a returning readership — a specialist blog, a newsletter, a customer magazine — it is the cheapest measure of this summer.
7. Measuring: how the effect shows up in your data
Generative UI produces no segment of its own in Search Console. The effect shows up indirectly, and only if you know in advance where to look.
Impressions stable, clicks falling
That is the typical pattern of displacement by the answer surface: your page is still served, but clicked less often. Track the click-through rate of your tool and guide pages separately from the rest of the site — in the overall average this effect disappears. The generative AI report in Search Console is the closest thing to a starting point.
The right comparison period
One confounding factor: on 18 August 2026 Google additionally rolled out a spam update. If you see movement from mid-August onwards, separate the two events rather than attributing every change to the AI surface. The sensible comparison is individual page groups against the previous year and the previous month, not the overall curve.
Page-type groups instead of single URLs
Set up three groups in your reporting: tool and calculator pages, guide and definition pages, offer and service pages. Only that separation makes it visible whether you have an AI problem or a general ranking problem.
8. Checklist: your next seven days
The stocktake is small enough to fit into one week. List the pages whose entire value is a function Google could generate itself. Note for each whether it uses data only you have, and whether it ends in a next step or in a number. Split your reporting into the three page-type groups above so that the click-through rates of tool pages stop hiding inside the site average. Then decide, per page, whether to extend it or let it go.
If you want to know which of your pages are exposed to AI answer surfaces and which are worth expanding, arrange a free initial consultation with our team.
9. Tools become a commodity, access stays yours
Generative UI marks a transition that has been forming since AI Overviews launched: first the summary became a commodity, now the function follows. What a model can build from publicly available knowledge, it will build sooner or later — free, inside the answer, with no detour.
What remains is everything a model does not have: your prices, your availability, your experience with this one industry, your offer, and the person who calls back. Companies whose visibility rests on generic utilities will lose ground over the coming months. Companies that couple their tools to their own data and to a next step will actually gain — because the competition made up of interchangeable calculator pages disappears from the result.
Put briefly: the tool is not the value, what happens afterwards is. Shape this transition now rather than enduring it, and you come out of the rollout stronger.
10. FAQ: the key questions about generative UI
What is generative UI at Google?
Generative UI is a Google search feature that assembles a bespoke layout for each question in real time, including interactive elements such as charts, comparison views, simulations or calculators. Instead of a plain block of text, the answer itself can be operable. Google has demonstrated this with an interactive pH scale and with a mortgage calculator in AI Mode.
Since when has generative UI been in AI Overviews?
Generative UI launched in November 2025, initially in AI Mode alongside Gemini 3. On 19 August 2026 Google announced that the feature would also roll out into AI Overviews. In AI Mode it is available globally in English; for full availability in AI Overviews Google named only the coming weeks — no fixed date and no list of languages.
Which pages are most affected by generative UI?
The greatest overlap is with pages whose only benefit is simple calculation logic or a standard table: converters, price and loan calculators, generic comparison tables and short definition pages. Search can generate those functions itself. Considerably more robust are tools built on your own data, offers or configurations, along with local service and offer pages with a real company behind them.
What are Preferred Sources and how do they help?
Preferred Sources are favourite sources chosen by users themselves. Websites can embed a button that sets their own site as a preferred source with one click. According to Google's documentation, selected sites appear more readily in Top Stories and can be labelled accordingly in AI Overviews and AI Mode. Google says more than 600,000 sources have already been selected — in May the figure was over 345,000.
