Your Customer Will Soon Stop Asking Google

Picture this scene: a potential customer is looking for a product, a service, a trades business in their area - and instead of typing the question into a Google search box, speaks it to their AI assistant on their smartphone. The assistant reads along on the screen, accesses appointments, books services, sends messages - and in the end delivers a single answer, not ten blue links your customer has to choose between themselves. Until now, this scenario on Android was largely reserved for Google's own assistant, Gemini. Since July 16, 2026, that is no longer legally tenable.

On that day, the European Commission issued two legally binding decisions against Google under Article 6(7) of the Digital Markets Act (DMA). They force the company to open up Android for competing AI assistants such as OpenAI's ChatGPT, Anthropic's Claude, Perplexity, and Microsoft Copilot - and to share anonymized Google search data with eligible third parties. For your business, this is not an abstract competition story out of Brussels. It's the starting shot for a shift that will help determine which channel your future customers use to find you at all - and whether your business shows up in the answer an AI gives them.

What the European Commission decided on July 16, 2026

The two decisions were issued under Executive Vice-Presidents Teresa Ribera (Clean, Just and Competitive Transition) and Henna Virkkunen (Tech Sovereignty, Security and Democracy). The first decision concerns AI interoperability on Android: Google must grant competing assistants equal access to eleven core system functions that had previously been effectively reserved for Gemini alone. The second decision obligates Google to share anonymized search data - queries, clicks, rankings, and views - with eligible third-party providers of search engines and search-capable AI chatbots.

Both decisions are not a proposal or a statement of intent, but legally binding requirements with fixed deadlines. Google has publicly voiced criticism: Kent Walker, President of Global Affairs and Chief Legal Officer, and Sameer Samat, head of Android, warn that the sweeping permissions for external apps could jeopardize device security, while the obligation to share search data touches on data protection, trade secrets, and, in part, national security interests. Whether and how Google will challenge the decisions is open at this point - the deadlines named are regulatory requirements, not yet facts already carried out. For business planning, though, that changes little: the direction has been set, and in 2026 the EU is showing itself to be in a particularly tough enforcement mood toward Google.

Anyone who optimizes only for Google today is optimizing for a gatekeeper that, from 2027, will no longer stand alone at the gate.

Eleven functions, one goal: AI assistants on par with Gemini

What's actually behind the eleven system functions is worth a closer look, because it shows how deep the intervention really goes. Until now, on Android, basically only Gemini was allowed to do certain things that competing assistants had to work around via individual app integrations. These include, among others: voice activation with a custom wake word instead of only "Hey Google," access to screen content and sensor data, cross-app actions in the so-called agent mode - meaning, for example, independently sending messages, booking appointments, or reserving services - background execution, and access to local hardware and on-device models.

In practice, that means: a user will in the future be able to set ChatGPT, Claude, or Perplexity as the primary assistant on their Android device, address it with a custom activation word, and the assistant can then - similar to what only Gemini could do before - act across app boundaries. It sees what's happening on the screen, understands the context, can independently trigger actions in other apps, and no longer has to be manually opened every time. That's the difference between a chat window that answers questions and an agent that acts on the user's behalf.

For your business, that means: the instance that mediates between your offering and your customer becomes more tangible. It's no longer just Google Search with its familiar ranking mechanism, but potentially one of several AI assistants, each with its own criteria for how it evaluates, weighs, and condenses information into an answer. When a customer asks their assistant, "Which trades business near me can take on an urgent heating repair?", it's no longer solely your Google ranking that decides whether you show up in the answer - but also how well your company data is prepared in a machine-readable way for that particular AI.

Here it's important to distinguish between the two EU decisions from July 16, 2026, because they concern two different layers. The Android decision opens up the system layer of the smartphone: it regulates which assistant can access which device resources and act on whose behalf. The search-data decision, on the other hand, opens up the data layer: it regulates what knowledge of past user behavior competing providers are allowed to use to train and improve their own systems. For your visibility as a business, both layers work together - better system access for AI assistants and better training data for their answer quality reinforce each other and jointly accelerate the shift away from pure link-list search toward the one, direct answer.

The timeline through 2028 – why you shouldn't wait until 2027 to start

The EU didn't mandate the overhaul overnight, but staggered it in clear, dated phases. Search-data sharing is set to start in January 2027; by February 1, 2027, Google must present the access criteria for third parties, so it's clear who gets access to the anonymized Google search data under what conditions. Most of the eleven Android functions must be available at the latest with the release of Android 18, by August 1, 2027. The simultaneous, parallel recognition of multiple wake words from different assistants - meaning, for example, that a device responds both to "Hey Google" and to a custom activation word from ChatGPT - doesn't follow until Android 19, by August 1, 2028 at the latest. The full overhaul thus stretches over a good two years.

For you as a business owner, this staggered timeline above all means one thing: a window of opportunity. Between today and summer 2027, you realistically have one to two years to build up your digital presence so that it's readable and citable for AI assistants, before the big rush of competitors chasing exactly this visibility sets in. Anyone who only reacts once Android 18 has rolled out and the first customers are actually asking their AI assistant instead of Google has already squandered their lead. Anyone who starts now can build, test, and refine structured data, clean content, and solid trust signals at a calm pace - without the pressure of competitors already being visible while you aren't. Especially for small and medium-sized businesses that rarely have their own large marketing departments, this long lead time is a rare gift: unlike many short-lived tech trends, you already know today the exact dates when things will change, and can plan your limited resources accordingly instead of reacting in a rush.

That the opening of Android is not an isolated event is shown by a look at the broader regulatory environment around Google in the summer of 2026. Just one week after the Android decision, on July 23, 2026, the European Commission imposed a separate DMA fine of 890 million euros on Google - 460 million euros for self-preferencing in Google Search, 430 million euros for violations in Play Store steering. Google has 60 days to stop the practices in question; otherwise, daily penalty payments of up to five percent of worldwide daily turnover are threatened. This fine legally concerns a different matter than the Android opening - it shouldn't be confused with the AI interoperability decision. In substance, though, it shows the same pattern: in 2026 the EU is in a phase of particularly active, tough DMA enforcement against Google, on several levels at once - and Google has little room to escape this development overall.

For marketing practice at your company, the search-data component is especially relevant. When Google has to share anonymized query, click, and ranking data with eligible third-party providers of search engines and AI chatbots starting in January 2027, Google loses its previous exclusivity as the only source of user search behavior. Competing search services and AI assistants could, as a result, learn over the medium term which content is actually relevant, which rankings work, and which click patterns emerge - their answer quality is likely to improve, and with it their usage. To put the previous starting position in perspective: Google's own assistant Gemini had, as of March 2026, around 750 million monthly active users - double the 350 million from March 2025. It's precisely this dominance that the two EU decisions are meant to break open. The comfortable assumption "I optimize for Google, that's enough" is thus increasingly becoming a risk rather than a solid strategy.

What this means for your marketing: Generative Engine Optimization

The technical term for the response to this shift is Generative Engine Optimization, or GEO for short, sometimes also called AI Answer Engine Optimization. GEO is, for the age of AI assistants, what classic SEO was for Google Search - except the goal is no longer a spot on page one, but a correct, concise mention in the one answer an AI assistant gives your customer. The most important building blocks are pleasantly concrete and achievable with a reasonable amount of effort.

First and foremost is structured data markup according to the Schema.org standard: Organization markup with correct company data, opening hours, and locations, Product markup for your services and offerings, FAQ markup for frequently asked questions, and Breadcrumb markup for a clear page structure. This markup makes what's actually on your website machine-readable for search engines and AI systems alike, instead of an AI having to guess from unstructured running text. Second, clear FAQ pages in question-and-answer format help: phrase the questions the way a customer would actually ask their assistant, and answer them directly and without detours in the first sentence. Third, a clean heading hierarchy counts - H1, H2, H3 in logical order - so AI systems correctly grasp the structure of your content and can cite individual sections in a targeted way. Fourth, and often underestimated, E-E-A-T signals are crucial: experience, expertise, authoritativeness, and trustworthiness, made visible through named authors, traceable references, genuine customer reviews, and solid source citations.

A word of caution belongs here too: in the GEO discussion, the so-called llms.txt keeps coming up - a Markdown overview file in a website's root directory, modeled on the principle of robots.txt. A study by SE-Ranking covering around 300,000 domains found, after a year and a half of discussion around this format, an adoption rate of only about 10.13 percent. Google itself already officially stated in July 2025 that it does not support llms.txt for AI Overviews and AI Mode. Current evidence from 2026 clearly suggests llms.txt is not a direct lever for AI visibility. Invest your time and budget instead in robust structured data and content quality, rather than relying on a single, contested file format.

A practical example: a trades business before and after the AI opening

To make tangible what this shift means for an SMB's day-to-day, a concrete scenario is helpful. Take a fictional but typical case: a regional plumbing and heating business with twelve employees, the kind found in nearly every German town. Today, this business's customer acquisition works essentially through classic Google search: a website with a services overview, a Google Business Profile, a few reviews, occasionally an ad. A customer with a broken heater types "emergency heating repair [city]" into Google, gets a list of results, and calls the business that appeals to them most - usually based on position in the search results, star rating, and website impression.

Now picture the same situation starting summer 2027, once the eleven Android functions are available for competing AI assistants. The customer no longer opens Google, but speaks to their AI assistant - whether that's Gemini, ChatGPT, Claude, or Copilot: "My heater has failed, I need someone nearby quickly." The assistant processes this request, matches it against available information - and ideally delivers not ten links, but a direct recommendation, perhaps even with the option to request an appointment or send a message to the business right away via agent mode. Whether the trades business shows up in this one answer at all now depends on whether its company data, services, availability, and reviews are structured in a way the AI can reliably match and cite.

The decisive difference between the two scenarios lies not in the trade itself, but in the preparation. A business that already fully maintains its Google Business Profile today, equips its website with Organization and Service markup per Schema.org, sets up an FAQ page with realistic customer questions like "Who handles a weekend heating emergency in [region]?", and visibly integrates customer reviews gains a structural advantage that reaches beyond the mere switch of search interface. It's also conceivable that such a business will eventually run its own small AI agent that automatically screens, prioritizes, and matches incoming appointment requests from various channels against the technicians' calendars - so that an AI-mediated inquiry actually turns into a booked appointment, instead of disappearing into an unwatched inbox. A business that does neither one nor the other continues to rely solely on a good Google ranking - while the customer may no longer even be using that channel. That's the core of the change: it's not that Google disappears, but that an additional, increasingly used intermediary layer emerges that operates by its own, partly still-unknown criteria - and one that structured, trustworthy data can already prepare for today.

What you should concretely do now

The good news in this whole regulatory story is: you don't have to wait until August 1, 2027 to act, and you don't have to panic either. The fundamentals for GEO are largely the same solid fundamentals that also apply to classic SEO and to trustworthy customer communication - just with a stronger emphasis on machine-readability and clear structure. Start with an honest inventory: is your company data - name, address, opening hours, services, ways to get in touch - consistent everywhere and backed by Schema.org markup? Do you have FAQ content that actually answers what your customers ask, rather than what your company would like to say about itself? Is your website built, both in content and structure, so an AI can grasp what you stand for without guesswork?

In doing so, deliberately don't bet on individual tactical fads like llms.txt, but on the robust fundamentals: structured data, clean content, genuine trust signals. This investment pays off regardless of whether your customer ultimately finds you via Google, Gemini, ChatGPT, Claude, or one of the new competitors - because good, clearly structured information is an advantage in every one of these systems.

It also makes sense not to treat this topic solely as an IT or website task, but as a task for your entire customer communication. Anyone who already documents today which questions customers actually ask by phone, email, or chat has, at the same time, gathered the raw material for future-proof FAQ content. Anyone who has so far communicated opening hours, scope of services, and price ranges only verbally or in a PDF should anchor this information on their own website in a machine-readable, consistent form. And anyone already working with AI agents, or thinking about it, should check where in the company recurring customer inquiries arise that could sensibly be handled by their own agent - from the first inquiry through scheduling to confirmation. That closes the loop: your customer's AI assistant finds you more easily, and an agent on your side ensures that this found inquiry actually turns into an appointment, a quote, or an order, instead of getting lost somewhere in day-to-day business.

With its decision from July 16, 2026, the EU has created a concrete, verifiable timeline that clearly differs from vague AI-hype warnings: fixed dates, named parties responsible, traceable deadlines. Use the window of opportunity through 2027 to prepare your digital presence for multiple channels instead of just one single gatekeeper, instead of only reacting once the overhaul is already complete and your competitors have made better use of the preparation time.