--- title: "Be the next best action with Autobasket" description: "Autobasket is an agentic commerce agency for Swiss and DACH shops: we prepare product data and checkout so ChatGPT, Gemini and Perplexity can buy from you." canonical: https://autobasket.store --- # Be the next best action with Autobasket A purchase used to start with a search and end on your checkout page. Now it often starts with someone describing a problem to ChatGPT, Perplexity, Claude or Gemini, and the assistant takes it from there: it explains the options, names a few brands, compares them and sometimes places the order and the refill. We make sure that at every step it finds you, trusts you and can buy from you. [Start selling smarter](https://autobasket.store/contact#session) [See how agents shop](#stages) Ask Explain Compare Buy Reorder Your Shop ## How do people and AI assistants now find, judge and buy a product? A purchase that used to run through a search engine and your product page now passes more often through an AI assistant first. The assistant searches, reads your offer as data and shortlists two or three products, and the person decides from that shortlist. Every offer is read twice, by the person and by the software working for them. How people bought 1. Search 2. Website 3. Checkout The shopper searched, opened a few sites and bought on the one that convinced them. Your page did the persuading. How people buy now 1. Ask 2. Explain 3. Compare 4. Buy 5. Reorder The shopper describes the problem and the assistant does the rest, reading your data before any person sees your page. | The moment | The person | The agent working for them | | --- | --- | --- | | Finding your product | In a search result, a social post, a shop they already know, or a name an assistant gave them | In the pages and data it retrieves when the person asks for something like a gravel bike under CHF 4,000 | | Judging your offer | From the photos, the reviews and what they already think of the brand | From the declared price, stock, GTIN, delivery date and return terms, and it drops an offer it cannot read | | Buying | Through your checkout, after opening one or two of the pages on the shortlist | Through your checkout, opened by UCP or ACP where the channel allows it, or by handing the person a link | | Coming back | When they remember you, or when your email reaches them | When a standing instruction tells it to reorder, provided it can confirm stock and price on your side | ## Where do shops lose the sale to an AI assistant? An assistant can drop your shop at four points in the same purchase, and each one looks different from your side. Your traffic may hold steady, because people still find you, while the orders the assistant places go to a shop whose data it could read and whose checkout it could use. - ### The assistant never names you It answers the shopper with the brands whose pages it could retrieve and quote. When your guides and product pages sit behind scripts or say little beyond marketing copy, another brand becomes the answer. [Fixed by Autobasket Discover](https://autobasket.store/products/autobasket-discover) - ### It misreads your offer Price, stock, delivery and returns that are not declared as data get guessed from the page or dropped. None of the 26 product pages in our July 2026 chocolatier audit declared a price an agent could read. [Fixed by Autobasket Discover](https://autobasket.store/products/autobasket-discover) - ### It cannot check out Where the shopper lets the assistant pay, it needs a checkout it can use. Without one it hands the shopper a link, and the order waits on a click that may never come. [Fixed by Autobasket Convert](https://autobasket.store/products/autobasket-convert) - ### It reorders somewhere else A refill months later goes to the shop where the assistant can confirm stock and price. When it cannot confirm yours, the standing order moves to a shop where it can. [Fixed by Autobasket Convert](https://autobasket.store/products/autobasket-convert) ## Why does Autobasket exist? Autobasket exists so that the shops people would pick for themselves stay in the answer when an assistant does the picking. We prepare your content, your catalogue and your checkout for software that reads them before any person does, and we measure each step so you can see what changed. - ### Found Assistants name your products when a shopper describes the problem they solve. - ### Trusted The price, stock and delivery date an agent reads match the page your customers see. - ### Bought An agent can place the order on your own checkout and come back for the refill. ## How does an AI agent buy, stage by stage? Every agent purchase passes through five stages: Discover, Understand, Choose, Buy and Retain. A shop can be strong at one stage and invisible at the next, which is how a brand gets recommended and still loses the sale. Autobasket measures each stage separately and fixes the one that loses the most orders first. Choose a route to follow A person searches, lands on your product page and orders on your site. A person asks ChatGPT, Perplexity, Claude or Gemini, which names products and sends them to the shop that answers best. The agent searches, compares and pays on its own, within the limits the person set, where the market and the platform allow it. 1. ### Discover Search result Named in the AI answer Asks your catalogue directly When a shopper describes the need, are you among the answers? 2. ### Understand Product page Product data the assistant reads Reads your product and offer data Can the product, price and seller be read without guessing? 3. ### Choose Comparison with other shops Shortlist inside the answer Compares your offer with other shops Does your offer win the comparison? 4. ### Buy Your checkout Link to your checkout Pays on your checkout within the approved limit Can the decision become an order? 5. ### Retain Account or order email Follow-up question to the assistant Reorders on a standing instruction Does the next order come back to you? ## How do Autobasket Discover and Convert divide the work? Autobasket works on both versions of your offer, in two products. Autobasket Discover gets your products named in more of the answers assistants give in your category, which is where a shopper's consideration now starts. Autobasket Convert lets an agent place the order on your own checkout, through the open protocols assistants use to buy. Discover, Understand, Choose ### [Autobasket Discover](https://autobasket.store/products/autobasket-discover) For marketing and SEO leads who need their products named in AI answers, and named correctly. - Pages written so assistants can quote your answers and name your products - A plain-text guide to your shop that assistants read first (`llms.txt`, `llms-full.txt` and Markdown copies of your pages), kept current - Prices, stock and product details that assistants read correctly, checked again after every catalogue change - A monthly figure for how often four assistants name you in your category, called Share of Model [See what Discover covers](https://autobasket.store/products/autobasket-discover) Buy, Retain ### [Autobasket Convert](https://autobasket.store/products/autobasket-convert) For CTOs and e-commerce architects who need agents to complete orders on the shop’s own terms. - Agents can search your catalogue, check stock and fill a cart directly, through an MCP server - Agents can pay on your checkout in the assistants that can buy in your market, through UCP or ACP - Each agent order is checked against what the shopper approved before you accept it, using AP2 mandates - Agent orders appear as their own channel in GA4 instead of disappearing into direct traffic [See what Convert covers](https://autobasket.store/products/autobasket-convert) ## What does an agentic commerce agency change in your shop? [Agentic commerce](https://autobasket.store/learn/what-is-agentic-commerce) is buying done by an AI agent on a person's behalf, from the first search to the order and the refill. An agentic commerce agency changes what that agent reads and uses in your shop: product data, policies, feeds, content and checkout. Autobasket does this on the platform you already run, layer by layer. | Layer | What changes for the agent | Where we do it | | --- | --- | --- | | Crawler access and HTML | Search and shopping crawlers are allowed in, and product content exists in the HTML without JavaScript | [Machine Readability Sprint](https://autobasket.store/services/machine-readability-sprint) | | Product and offer data | Price in CHF, stock, [GTIN](https://autobasket.store/learn/why-agents-need-a-gtin), [sizes and variants](https://autobasket.store/learn/sizes-and-variants-for-agents), shipping and returns declared as data that matches the visible page | [Machine Readability Sprint](https://autobasket.store/services/machine-readability-sprint) | | Content assistants quote | Answers, specifications and comparisons written as passages an assistant can lift, with their sources | [Autobasket Discover](https://autobasket.store/products/autobasket-discover) | | Product feeds | Google Merchant Center and the OpenAI product feed filled from the same source as the page | [Machine Readability Sprint](https://autobasket.store/services/machine-readability-sprint) | | Checkout | Cart and payment through [UCP or ACP](https://autobasket.store/learn/agent-commerce-protocols-explained) where the channel is open, with AP2 mandates when the person is not present | [Autobasket Convert](https://autobasket.store/products/autobasket-convert) | | Measurement | How often each assistant names you against competitors, and agent orders counted as their own channel in GA4 | [Discover](https://autobasket.store/products/autobasket-discover) and [Convert](https://autobasket.store/products/autobasket-convert) | ## Which shops does Autobasket prepare for AI agents? Agents shop differently in each category. A bike is chosen on frame size and stock per size, a sofa on dimensions and lead time, a serum on ingredients and reviews, a B2B reorder on a signed mandate. Each solution page shows how agents buy in that category and which stage usually breaks first. ### By industry - [**Sports and outdoor**Bikes, ski, running and camping gear sold by size and fit](https://autobasket.store/solutions/sports-outdoor) - [**Fashion and apparel**Clothing, shoes and accessories bought on size, material and returns](https://autobasket.store/solutions/fashion-apparel) - [**Food and drink**Chocolate, coffee, wine and speciality food sold on origin and gifting](https://autobasket.store/solutions/food-drink) - [**Consumer electronics**Devices and accessories compared on specs, compatibility and price](https://autobasket.store/solutions/consumer-electronics) - [**Home and furniture**Furniture, lighting and homeware bought on dimensions and delivery](https://autobasket.store/solutions/home-furniture) - [**Beauty and personal care**Skincare, fragrance and cosmetics chosen on ingredients and reviews](https://autobasket.store/solutions/beauty-personal-care) ### By business model - [**Brands selling direct**D2C brands that need to be the source AI answers quote](https://autobasket.store/solutions/brands-selling-direct) - [**Multi-brand retailers**Retailers who win the comparison on price, stock and delivery](https://autobasket.store/solutions/multi-brand-retailers) - [**B2B and wholesale**Suppliers whose buyers delegate reorders to procurement agents](https://autobasket.store/solutions/b2b-wholesale) ## What did our audit of a Swiss chocolatier find? In July 2026 we audited a bean-to-bar chocolatier near Geneva that sells 205 products. AI answer engines named the brand in 58% of the Google AI Overviews and 33% of the Perplexity answers we tested, yet none of the 26 product pages we rendered declared a machine-readable price, so agents could recommend the shop but could not buy from it. We ran 24 prompts per engine, so the 95% confidence intervals run from 36 to 77% for Google and 18 to 53% for Perplexity, and rendered the product pages in a full browser on 13 July 2026. The merchant stays unnamed because the audit used public data only, and the [full audit is published as a use case](https://autobasket.store/use-cases/chocolatier-findable-not-buyable). Named in Google AI Overviews 58% Named in Perplexity answers 33% Product pages with a structured price 0 of 26 ## What does the data say about AI shopping in 2026? AI assistants already send shoppers who buy. Adobe measured traffic from AI tools to US retail sites up 693% over the 2025 holidays, and Shopify reports AI-referred orders up nearly 13 times in a year. In Switzerland, Germany and Austria only 5% of shoppers have used an agent so far, which leaves Swiss shops time to prepare. | What was measured | Source, date and scope | | --- | --- | | Traffic from generative AI tools to US retail sites rose 693.4% year on year in the 2025 holiday season. | [Adobe Analytics](https://news.adobe.com/news/2026/01/adobe-holiday-shopping-season), 7 January 2026. United States, 1 November to 31 December 2025 | | On Shopify, sessions referred by AI chats convert at nearly 50% higher rates than organic search, and AI-referred orders grew nearly 13 times year on year in the first quarter of 2026. | [Shopify](https://www.shopify.com/enterprise/blog/ai-search-insights), 11 May 2026. Shopify merchants, product page sessions | | In March 2026, visits referred by AI tools to US retail sites converted 42% better than other traffic, a year after converting 38% worse. | [Adobe Analytics](https://techcrunch.com/2026/04/16/ai-traffic-to-us-retailers-rose-393-in-q1-and-its-boosting-their-revenue-too/), reported by TechCrunch, 16 April 2026. United States retail | | 5% of online shoppers in Switzerland, Germany and Austria have already used an AI agent to help them shop, and about 18% plan to. | [University of St.Gallen (HSG), survey of just under 3,000 people](https://www.unisg.ch/en/newsdetail/news/ki-agenten-im-onlinehandel-warum-konsumentinnen-und-konsumenten-beim-einkauf-noch-zoegern/), 24 June 2026. Switzerland, Germany and Austria | | Over 80% of people in Switzerland reject a fully automated AI payment without confirmation, while nearly two thirds can imagine an AI assistant recommending how to pay. | [Swiss Payment Monitor 2/2026, University of St.Gallen and ZHAW](https://alexandria.unisg.ch/entities/publication/c1fa42b2-d39d-4420-9ee4-dfe284cbfc14), August 2026. Switzerland, consumer survey, May 2026 | | None of 240 online shops in French-speaking Switzerland scored an A for agent readiness. The median score was 57 out of 100, and 104 shops scored D or F. | [Autobasket audit of 240 Romandie online shops](https://autobasket.store/use-cases/romandie-online-shops-agent-readiness), July 2026. Switzerland, Romandie, 240 shops | | On 117 of 211 shops (55%), the only price an agent could find was in the page text, with no Offer price in the structured data to confirm it. | [Autobasket audit of 240 Romandie online shops](https://autobasket.store/use-cases/romandie-online-shops-agent-readiness), July 2026. Switzerland, Romandie, 211 shops read in full | ## Where would you like to start? Some teams book a conversation after one look at the problem, others want the numbers, the method or the people first. All four routes below lead to the same place, a shop that agents can read and buy from, so pick the one that matches how your team decides. - ### Talk to us this week You already see assistants answering questions in your category. We look at one of your product pages with you and say what we would fix first. [Start selling smarter](https://autobasket.store/contact#session) - ### See the evidence first Read what a real audit measured, with prompts, confidence intervals and the stage where the purchase broke. [Read the chocolatier audit](https://autobasket.store/use-cases/chocolatier-findable-not-buyable) - ### Read the full method The five-stage framework, the twelve checks and the protocols, laid out in tables with their sources. [Go to the framework](#framework) - ### See how we work with you What happens after the first call, who on your side is involved, and what you have in hand at the end. [See the four steps](#working) ## How does the five-stage agentic commerce framework work? At each of the five stages, the person who approves the budget and the agent doing the shopping need different things from your shop. The table shows both, and what Autobasket puts in place so that stage stops losing sales. Read it row by row, or go straight to the stage where you suspect the loss. | Stage | What the decision-maker needs | What the AI agent does | What Autobasket puts in place | | --- | --- | --- | --- | | Discover | To know whether the brand shows up when customers ask an assistant for a recommendation | Searches the web, retrieves passages and ranks sources across ChatGPT, Perplexity, Claude and Gemini | Pages assistants can quote, open access for AI crawlers, and AI referrals counted in your analytics (`llms.txt`, crawler rules) | | Understand | Evidence that the integration is safe, compliant and cheap to run | Extracts the product, its attributes and the merchant's identity from structured data | Product facts that agents read the way your customers do, with every figure sourced (JSON-LD schemas, Markdown mirrors) | | Choose | A comparison with the alternatives on proof, cost and effort | Compares offers on price, availability, delivery time, returns and ratings | Price, delivery and return terms an agent can compare without guessing (Offer, shipping and return markup, HTML comparison tables) | | Buy | A scoped project that can be approved in one meeting | Calls tools to build a cart, then presents a payment mandate the user signed | A checkout agents can complete, which accepts only the orders the shopper approved (MCP tools, UCP or ACP checkout, AP2 mandates) | | Retain | Monthly figures on AI visibility and on orders placed by agents | Reorders under a standing mandate and reads order status through the API | Monthly figures on how often assistants name you and what agents ordered, and reorders that go through on their own (Share of Model, GA4 attribution, reorder endpoints) | ## How can you check whether AI agents can read your shop? The Autobasket readiness check reads one product page the way an AI shopping agent does, in under a minute. It runs 12 checks in three stages: whether agents may reach the page, whether they can read the product and its price, and whether they have enough to decide. Every failure comes with the evidence found and the fix. 1. ### Reach Agents can open the page without a browser and see its content before any JavaScript runs, and your `robots.txt` lets search and shopping agents in. 2. ### Read The agent can tell exactly which product it is, by its GTIN barcode number or SKU, read the price, currency and stock, and find the same price the page shows, with a complete title and preview (Schema.org Product data). 3. ### Decide The agent can read shipping costs, delivery times, return terms and customer ratings, and finds an `llms.txt` file that describes your shop. ## Which open protocols and standards does Autobasket work with? Agents place orders through open protocols that connect to your systems directly. Autobasket implements and maintains the Model Context Protocol, the Universal Commerce Protocol, the Agentic Commerce Protocol, the Agent Payments Protocol and x402, and reads the card networks' agent programmes, so the agents your customers use can query, order and pay. | Protocol | Who stewards it | What it lets an agent do | Where it stands on 12 October 2026 | | --- | --- | --- | --- | | [Model Context Protocol (MCP)](https://modelcontextprotocol.io) | Agentic AI Foundation, Linux Foundation, since December 2025 | Call your tools, such as catalogue search, stock lookup or cart creation, over JSON-RPC 2.0 | Widely supported | | [Universal Commerce Protocol (UCP)](https://shopify.engineering/UCP) | Co-developed by Google and Shopify, January 2026 | Complete checkout with any merchant, over REST, MCP, AP2 or A2A | Live in Google AI Mode | | [Agentic Commerce Protocol (ACP)](https://openai.com/index/buy-it-in-chatgpt/) | OpenAI and Stripe, September 2025 | Send an order and a delegated payment token while you stay merchant of record | Narrowed in ChatGPT, 2026 | | [Agent Payments Protocol (AP2)](https://blog.google/products-and-platforms/platforms/google-pay/agent-payments-protocol-fido-alliance/) | Created by Google, donated to the FIDO Alliance on 28 April 2026 | Prove what the user authorised with signed mandates, including Human Not Present payments in v0.2 | v0.2 published | | Visa Trusted Agent Protocol and Mastercard Agent Pay | The card networks, with Mastercard's Verifiable Intent also going to FIDO | Let a merchant tell a verified shopping agent from a scraping bot, and log what the user approved | Market by market | | x402 | Open specification started by Coinbase | Pay per request in stablecoins such as USDC, using the HTTP 402 status code | Early adoption | [Read how MCP, UCP, ACP and AP2 fit together](https://autobasket.store/learn/agent-commerce-protocols-explained) ## What happens after you get in touch? You talk to the people who will do the work, and you see the problem on your own product pages before you commit to anything. Every engagement starts from what agents actually do on your shop today, ends with fixes your developers can ship, and is measured again afterwards with the same missions. 1. ### A first conversation We open one of your product pages together and read it the way an agent does. You leave with the stage that loses you the most. 2. ### A written proposal Scope, timing and a fixed fee for the audit or the Sprint, based on your catalogue, your platform and your markets. 3. ### The work, with your team The Buyer Journey Check maps where agents drop your shop. The Sprint makes your product pages readable to agents in seven days, on your current platform. 4. ### Measured again We rerun the same missions and the same checks, so you see what changed in figures rather than in a slide. ## What do merchants ask before working with us? These are the questions e-commerce leads, marketing leads and CTOs ask in the first conversation, answered in full. Each answer stands on its own, so you can forward one to a colleague, and the same text is published as structured data for the assistants that read this page. ### What is agentic commerce? Buying done by software on a person’s behalf. An AI assistant such as ChatGPT, Perplexity, Claude or Gemini searches, compares and, where the merchant allows it, places the order. The shop is read by the agent before any person sees it, so what the agent can read decides whether the shop is considered. ### How can AI assistants increase revenue for an online shop? They send shoppers who are close to buying. Shopify reports that sessions referred by AI chats convert at nearly 50% higher rates than organic search, and Adobe measured AI-referred visits to US retailers converting 42% better in March 2026. That revenue goes to the shops an assistant can read and compare, so the first lever is product data it can quote. ### Which AI assistants can complete a purchase today? It depends on the market. In the United States, Google’s AI Mode and the Gemini app complete checkout through the Universal Commerce Protocol for eligible retailers since 11 January 2026, and ChatGPT handles purchases inside apps through the Agentic Commerce Protocol since OpenAI moved Instant Checkout there in March 2026. In Switzerland most assistants still recommend and link, so we check which channels can buy in your market before building anything. ### Is agentic commerce relevant for a Swiss online shop in 2026? For being found, yes, and for being bought, soon. Agent checkout launched in the US first, and only 5% of shoppers in Switzerland, Germany and Austria have used an agent (University of St.Gallen, June 2026). Assistants already recommend Swiss products today, and in our July 2026 audit of 240 Romandie shops none scored an A for agent readiness, so a shop that fixes its data now competes against shops that have not. ### How is this different from SEO? SEO ranks pages for people who click. Agentic commerce prepares pages and data for software that reads, compares and buys without clicking. The two overlap on crawler access and structured data, but agents also need complete offers, machine-readable policies and, for checkout, protocols that SEO never touched. ### Do we need to change our e-commerce platform? No. The Machine Readability Sprint works on top of Shopify, WooCommerce, Salesforce Commerce Cloud, SAP Commerce and custom stacks. We add what agents need to read and order, structured data, Markdown copies of your pages, an llms.txt file and agent endpoints, to the platform you run today. ### How long does it take to become agent-ready? The Machine Readability Sprint makes your product pages readable to agents in seven days. A Buyer Journey Check tells you beforehand which stage loses the most sales, so the first week goes to the fix that matters. Checkout protocols take longer and depend on the channels open in your market. ### What does it cost? The Buyer Journey Check and the Sprint are quoted as fixed fees per shop, after a first conversation about your catalogue and markets. Discover and Convert are monthly retainers. The free readiness check on this page costs nothing and shows where one product page stands today. ## Where can you learn how AI agents choose products? Our guides explain the mechanics in plain language, with the sources named: how assistants retrieve and cite pages, what `llms.txt` does and does not do, and how the payment protocols divide the work. Start with whichever question your team is asking this week. [See all agentic commerce guides](https://autobasket.store/learn), the newest articles on the [agentic commerce blog](https://autobasket.store/blog) or the [glossary](https://autobasket.store/glossary). - [GEO vs SEO, what changes for an online shop? GEO vs SEO for online shops: both start with crawling and indexing, 76.1% of AI Overview citations rank in the top 10, and what AI answers add.](https://autobasket.store/learn/geo-vs-seo-for-online-shops) - [How does agentic commerce work for an online shop in Switzerland? Agentic commerce in Switzerland: CHF prices with VAT in Offer data, TWINT at checkout, Swiss delivery as shippingDetails, and our audit of 240 shops.](https://autobasket.store/learn/agentic-commerce-switzerland) - [UCP vs ACP vs AP2 vs MCP, which protocol does what in a purchase? UCP vs ACP, AP2 and MCP compared for merchants, with owner, launch date, status and who needs each, plus where x402 and A2A fit for a Swiss shop.](https://autobasket.store/learn/agent-commerce-protocols-explained) - [How do you get your products recommended by ChatGPT? How to get your products recommended by ChatGPT: what OpenAI says it ranks on, how product feeds reach it, and what the GEO research measured.](https://autobasket.store/learn/how-ai-assistants-choose-products) ## Make your products the ones agents can buy One conversation on one of your product pages is enough to see which stage loses you sales, so the next agent that asks about your category leaves with your product in its basket. [Start selling smarter](https://autobasket.store/contact#session) [See the Buyer Journey Check](https://autobasket.store/services/buyer-journey-check)