AI agents, GEO & LLMS.txt

A new discipline for Shopify agencies

It's a Thursday in the summer of 2025: a customer from Lausanne opens her browser and types in "light rain jacket, breathable, under 100 francs". Seconds later, a digital shopping advisor suggests a selection of color-coded models in her size - in Swiss francs, with delivery in two days. She clicks on the first product - and without having clicked on a traditional search engine link, she has suddenly landed in your Shopify store.

What just happened? Not a funnel trick, not SEO magic from the Shopify agency - but the result of a quiet but profound change in e-commerce: AI agents, GEO Commerce and LLMS.txt. Three inconspicuous terms that together form the foundation for how people will shop in the near future - and how retailers will be found without having been searched for in the traditional way. The growth forecasts at the end of this article assume that fundamental changes could take place within the next two years.

The new search: When customers talk to AI, not Google

For many years, the goal was to appear as high up on Google as possible. H1 titles, meta descriptions, structured data, backlinks - the SEO basics were set. But with the growing spread of large language models (LLMs) such as ChatGPT, Claude or Gemini, the way people search for solutions online is shifting. More and more users are asking the AI directly. And this is not based on links, but on content.

This is where LLMS.txt comes into play - a simple text file in the root directory of your website that gives these models guidance: "Here you will find relevant content. Here are the pages you should know." LLMS stands for "Large Language Model Sitemap" - and is, put simply, the sitemap for AI systems.

The proposal comes from Jeremy Howard, a well-known data scientist, who put the format up for discussion in the fall of 2024. Since then, platforms such as LangChain,Fast.ai or Tinybird, and according to initial studies, over 2,000 domains had already been equipped with an LLMS.txt by mid-2025 - many of them through automatic generation via Yoast SEO.

There is still no official read access from OpenAI or Google. But as is so often the case in the digital world, it pays to get involved early. Because when the AI actually starts to evaluate content in a structured way in the near future, it will start where someone has prepared a proper LLMS.txt for it.

GEO Commerce - The store speaks your language

Parallel to the AI-supported search, expectations on the customer side are also changing. GEO Commerce means that the store recognizes where the user is coming from and reacts accordingly. Currency, language, taxes, shipping options - everything is adapted to the context. For the customer from Vienna, this means euros, Austrian shipping rates and, if necessary, localized content. For the customer from Zurich, it means CHF, French-language texts if necessary and a different product selection.

The aim: nobody should feel lost. Everyone should feel that they have been picked up directly - without selection menus, without language switches, without manual readjustment. This is exactly what Shopify Markets, IP-based redirects or, in the case of headless setups, free GEO logic in the middleware layer make possible.

AI agents as salespeople, consultants & helpers

And then there are the digital colleagues. AI agents are more than just chatbots - they can act, learn and prioritize independently. They suggest products, react to user behavior, answer questions, place orders and solve support cases. According to DemandSage, the market for AI agents will be worth around 7.6 billion US dollars by 2025 - and is set to grow to over 50 billion by 2030. According to Experro, more than 60% of global retailers are already using AI agents today - particularly in customer service, personalization and sales.

Customers who interact with agents buy faster, in a more targeted manner - and spend up to 25% more. First contact resolution is over 84% - significantly higher than with traditional support channels.

Headless + GEO - An ideal combination

You might wonder how headless commerce and GEO/LLM topics are related. In fact, they complement each other perfectly:

• Headless liefert die APIsthat AI agents and generative search engines can tap into. A company with headless architecture has already "liberated" its data and made it available - exactly what AI services need. In a scenario where, for example, an AI agent queries prices for a user or fulfills an order, an open API backend is worth its weight in gold. Shopify's headless storefront API, for examplealready enables AI agents to manage product catalogs or process orders today. Without headless, an AI would at worst have to try to "scrape" web interfaces or would not receive any live data at all.
• Headless facilitates the implementation of LLMs.txt. As content in headless setups is often stored in cleanly structured content repositories (e.g. headless CMS), it is easier to compile it in a structured overview. It is possible that headless CMSs will offer direct export functions for llms.txt in the future.
• Common goal: answer user questions better. Whether a human user visits the website or an AI assistant queries data - both benefit fromgood structure, clear preparation and fast delivery. TheConversational interfaces (chats, voice) are ultimately new front ends that can be operated thanks to headless architecture. You could say that headless commerce makes a company„AI-ready“by making the content available regardless of the channel.

A practical example of interaction: Let's take a B2B dealer for technical spare parts. It has opened its product catalog via API with Headless. Now the company is developing aAI-based digital assistants for its major customers, who can answer technical queries ("Which component fits machine X?") and place orders straight away. ThisAgent accesses the APIs to obtain real-time data (availability, compatibility, etc.). At the same time, the company has allms.txt with all important product categories and specifications. If an external AI assistant (e.g. an industry-wide AI service) searches for information, it can quickly find the relevant data sheets via llms.txt. This means that the retailer is optimally represented both with its own agent and in the general AI search.

The future: It can be predicted that companies that rely on headless early on will also be ahead in the AI era. Companies are already reportingnoticeable successes through AI integration: Companies that use AI agents see higher conversion rates compared to traditional web experiences. Companies with headless commerce structures also report higher conversion rates compared to those with monolithic setups. These effects are likely to add up - in other words, headless + AI could take e-commerce to a new level of performance in the coming years.

Integrate vector databases

In this context, a new technology is currently conquering ecommerce: - so-calledVector databases. What sounds abstract is based on a simple principle: content - such as product descriptions - is converted intomathematical vectorsconverted. These not only represent words, but alsoMeaning.

A little experiment shows just how revolutionary this is. Let's take four products: elegant leather shoes, comfortable sneakers, business shoes and running shoes. Instead of simply searching for keywords such as "dinner" or "comfortable", an AI model likeMiniLM converts the texts into vectors - a kind of fingerprint of their meaning. Then a search system (such as the open source databaseFAISS) these vectors with the user's request.

The effect? On the search query"Shoes for an elegant dinner, but comfortable" the system responds promptly with :

• "Elegant black leather shoe for men"
• "Business shoe in brown with soft footbed"

None of the product descriptions contained the exact words "dinner" or "comfortable". And yet the system found exactly what the customer was looking for -because it has understood the content.

Modern AI technology is behind such functions:Embedding models (such as from OpenAI or Hugging Face), databases such asWeaviate, Pinecone or Redisand a new type of search that does not search for strings, but forsemantic proximity asks.

ForHeadless commerce provider this opens new doors: you can not only use classic search bars, but alsoIntelligent chatbots and AI agents with access to these vector databases. The result? Personalized product recommendations, precise answers, natural dialogue.

What Amazon and Walmart do differently

Major retailers are setting a good example. According to Business Insider, Amazon, for example, expects over 700 million US dollars in additional profit in May 2025 from its AI shopping agent "Rufus" alone. The agent is gradually replacing the traditional search function - and combines user behavior, season, location and trends to provide context-based product suggestions.

Walmart is even going one step further. According to the Times of India, the US retailer wants to generate more than 50% of its online sales via autonomous agents by 2030. The systems there are called "Sparky" and not only provide advice, but also handle stock planning, marketing automation and real-time pricing.

The message: AI is not a bonus, it is becoming the operating system of digital commerce.

How to implement it yourself

The good news: you don't need a billion-dollar budget to take the first step. With Shopify Plus, a headless setup and smart planning, you can get started now. Here is a small to-do list:

To-Do 1: Create an LLMS.txt file with a clear structure. Link your most important pages, add short descriptions and make sure they are up to date.

To-Do 2: Build in GEO personalization - via Shopify Markets, headless routing or IP detection. Show customers content that matches their context.

To-Do 3: Test at least one AI agent use case: shopping cart reminder, FAQ dialog, product advisor. Choose a manageable task and measure the effects.

Future outlook and forecasts

Experts expect the use of AI search systems to continue to increase in the coming years and profound changes in the search market. Market analyses by Gartner and Semrush predict that AI-based search could overtake traditional search traffic by 2028. Specifically, a recent Semrush study assumes thatby the beginning of 2028 at the latest, more visitors will reach websites through AI search assistants than through traditional search engines. In addition to the rapid increase in the use of chatbots, this is also being driven by the integration of AI into established platforms: If Google makes its AI mode (SGE) the standard, this turnaround could happen even sooner. Gartner predicts in a survey that theorganic search terms could collapse by over 50 % by 2028 -the rapid adoption of generative AI in search services would fundamentally shake up the existing SEO model. Marketing experts are therefore already advising companies to prepare for a future in which classic Google searches will deliver far less traffic and AI platforms will gain in importance instead.

Ambitious goals have also been set in terms of user numbers: OpenAI itself is aiming to reach 1 billion active ChatGPT users by the end of 2025 - which does not seem unrealistic given the curve so far. Some analysts even speculate that ChatGPT could overtake Google in terms of visits in a few years. Growth expert Kevin Indig, for example, calculated that withContinuation of current growth trends ChatGPT could overtake Google in terms of visitor numbers around 2027In an extreme case (extrapolated from recent growth), this could possibly be as early as 2026. Although such aggressive scenarios are uncertain, they illustrate the dynamics of the market. Much more likely is a scenario in which Google defends its dominance with hybrid models - in other words, the boundaries between search engine and AI assistant become blurred. Google is increasingly integrating AI features in order to remain competitive and Microsoft is expanding Bing with GPT technology. In the future, the standard search experience could be conversational (keyword: chat in search), with Google, Microsoft and new players vying for supremacy in these AI-supported answers.

User behavior and expectations will continue to shift in parallel: Younger generations are already used to asking questions to AI systems, and voice and chat searches are likely to increase further. Forecasts by McKinsey & Co. suggest that the proportion of natural language search queries will continue to grow andCompanies must increasingly optimize their content for conversational queries (Conversational SEO) - similar to what is already the case with voice assistants today. The topic of trust and quality will also come into focus: AI providers are working on minimizing hallucinations and incorrect answers, as this is crucial for broad acceptance.

To summarize, AI-based search and response systems are a rapidly growing sector that is already beginning to have an impact on Google's dominance. Users are increasingly turning to ChatGPT & Co. directly - whether out of convenience, better contextual answers or curiosity. So far, Google is still the top dog and is even benefiting in part from increased search volumes thanks to new features. But the rules of the online search game are changing: away from ten blue links and towards curated AI answers that help users immediately. Studies and market data underpin this change - from 10% of users already starting their search with AI to predictions that in a few years' time, more search queries will be made via AI agents than via traditional search engines. The coming years will show how quickly this trend will manifest itself. However, one thing is clear: user behavior is shifting measurably towards AI, and both the major search engine operators and companies, marketers and Shopify agencies worldwide must adapt to a fundamentally changed search landscape.

Sources: The information is based on current studies and market analyses, including from Semrush (in cooperation with Statista), SparkToro, Gartner and other market researchers. All statistics and forecasts quoted are taken from the linked sourcessearchengineland.com,contentmanager.de,firstmotion.com,9rooftops.com etc. These provide detailed evidence of the aforementioned usage figures, trends in search behavior and future projections.

What does this mean for my store

Of course, this depends on the competitive environment in which your business operates. As a small niche provider, you can certainly continue to rely on the Shopify Standard Shop or that Shopify LLMS will be integrated natively or that apps will take over this function.

The situation is different in the competitive e-commerce environment and B2B. Here it can be assumed that headless commerce with special solutions brings decisive advantages. Be it in the provision of content and vector databases via API, special LLMS.txt or simply the use of different AI models in the content management system (CMS). With ChatGPT-5 expected to be launched in August 2025, it's time for a Shopify agency to take a closer look at AI agents. The future of ecommerce remains colorful - and AI will not change that.