Agentic commerce: how AI agents shop, what is live in 2026, and what to do first

By Simon Vreeman, Growth & Technical Marketing Advisor at Vreeman.com · Published 15 October 2026

On 15 October 2026 Google switched on checkout inside Search and the Gemini app for shoppers in the United States. The same week Shopify opened its catalog of a billion products to any AI agent, American Express published a framework for proving what a shopper meant, and Stripe quietly added Meta as a co-author of OpenAI's commerce protocol. Agentic commerce, where an AI agent finds, compares, buys and tracks products on a person's behalf, moved from announcements to shipping product this year. It also hit its first wall: OpenAI pulled its standalone checkout in March, agent-completed purchases are still measured in basis points, and nothing is live for European shoppers inside the big assistants.

This article collects what the companies have published, what the data from Adobe, Salesforce, Shopify, Contentsquare and Accenture show, what the courts and regulators have said, and what a webshop in the Netherlands should do first. Every claim links to its source. Where a number comes from a vendor's own panel or press release, the text says so. The author maintains a protocol tracker, checklist and free site audit at vreeman.ai/commerce, which this article draws on; he has no commercial ties to any vendor named here.

Key takeaways

• Checkout inside AI surfaces is live, in the US, for three of the four big assistants. Google's Universal Cart (15 October 2026), Microsoft's Copilot Checkout (April 2026) and Perplexity's Instant Buy with PayPal (November 2025) complete purchases without a page load. ChatGPT only shows products and hands off.
• OpenAI's retreat is the most important data point of the year. It ended standalone Instant Checkout on 24 March 2026 after very few merchants went live, and told merchants to send a product feed instead. Discovery in the assistant, purchase on the merchant's site, is the pattern that survived.
• Discovery is growing fast, agent checkout is tiny. Adobe: AI-referred traffic to US retail sites up 393% in the first quarter of 2026, converting 60% better than other traffic by July. Contentsquare: AI referrals are 0.2% of sessions. Accertify: agentic checkout is under 1% of transactions.
• The merchant stays merchant of record under every open protocol. UCP, ACP, Shop MCP and PayPal Agent Ready all leave the order, the customer, returns, fraud and support with the merchant. The platform keeps the first-touch comparison.
• The protocol stack has settled into layers. UCP and ACP for the shopping journey, AP2 and the card networks for payment and proof of intent, Visa's Trusted Agent Protocol and Web Bot Auth for agent identity, x402 and MPP for machine-to-machine payments, MCP and A2A underneath. Most are beta.
• Europe has the rails, not the surfaces. ING, Worldline and Mastercard ran the first European agentic payment with a Dutch merchant on 2 June 2026, and Visa has 31 issuers live. Google's UCP checkout is limited to the US, Canada and Australia; Copilot Checkout is US only; OpenAI has not launched ACP in Europe.
• Product data is the low-regret work. Every surface reads a Merchant Center or OpenAI feed. GTINs, consistent price and stock, Product markup and UTM-tagged feed links pay off in today's channels whether or not agents ever buy from you.

What agentic commerce is

Agentic commerce is shopping where software acts for the buyer. The agent takes a goal ("a black rain jacket under 150 euros that arrives before Friday"), searches across shops, compares, builds a cart, pays with a credential the person has delegated to it, and later checks the order status or starts a return. The agent can sit in a chat assistant (ChatGPT, Gemini, Copilot, Claude), in a browser (Perplexity's Comet), in a marketplace app (Amazon's Buy for Me) or in a personal agent the consumer runs themselves, which is what Sierra and Meta's Personal Agent Protocol (6 October 2026) is designed for.

Three things distinguish it from a chatbot that recommends products:

• The agent reads structured data, not pages. Google's checkout runs on the Merchant Center feed, ChatGPT on the OpenAI product feed, Copilot on Microsoft Merchant Center, Shopify's agents on Shopify Catalog. The product page is a fallback.
• The agent holds a payment credential. Under ACP that is a delegated token scoped to one checkout; under AP2 a signed mandate; under Mastercard Agent Pay a tokenised card. The merchant never sees a raw card number from an agent.
• Someone has to prove the agent is legitimate and the person consented. That is what Visa's Trusted Agent Protocol, Mastercard's Verifiable Intent, Amex's Intent Intelligence (13 October 2026) and the IETF's Web Bot Auth work are for.

Under every open protocol published so far, the merchant remains the seller of record. Google says so for Universal Cart, Stripe for ACP, Microsoft for Copilot Checkout, PayPal for Perplexity Instant Buy. The order lands in your system, the chargeback is yours, the return is yours. What you give up is the first-touch comparison, which happens inside the assistant.

The customer journey and which protocol covers which step

MCP (the Model Context Protocol, created by Anthropic and now hosted by the Linux Foundation) and A2A (agent to agent, originally from Google, also at the Linux Foundation) sit under all six stages as transport. Neither is a commerce protocol; both are how the commerce protocols move messages.

The numbers: discovery is real, agent checkout is not yet

The studies below use different panels, periods and definitions of "AI traffic". Read the trend, not the decimals, and note who published each one.

Our reading: three different things are being measured. Adobe and Shopify measure visits that arrive from an assistant and then convert on the merchant's site; that is growing fast from a small base, and the shoppers are good. Salesforce measures influence on a season's sales with a method it has not published. Accertify measures purchases an agent finished by itself; that is where the basis points are. The growth that matters today is in discovery. The checkout layer is infrastructure built ahead of demand, which is why the business case section below is cautious.

Where customers can buy with AI today

Status as of 15 October 2026. "Yes" means a purchase completes inside the assistant. "Discovery only" means the assistant shows products and sends the shopper to the merchant to pay.

What happened at OpenAI: the pivot to discovery

On 29 September 2025 Stripe and OpenAI published the Agentic Commerce Protocol, Apache 2.0 licensed, and launched Instant Checkout in ChatGPT with US Etsy sellers, with "over a million Shopify merchants coming soon". Shopify, Walmart, Target and Sephora were announced as partners. PayPal joined as a payment provider on 28 October 2025.

Six months later OpenAI stopped. In "Powering product discovery in ChatGPT" (24 March 2026) it wrote that the first version of Instant Checkout "did not offer the level of flexibility" it wanted, and that merchants can now use their own checkout while ChatGPT focuses on discovery. CNBC reported that very few merchants had gone live; a Forrester analyst quoted by Exploding Topics put it at about 30 Shopify merchants in February 2026. Target, Sephora and Nordstrom were named as discovery partners. Purchases now complete on the merchant's site or in the merchant's app.

What survived is the feed. OpenAI's product feed spec asks for nine required fields (item id, title, description, URL, brand, seller name, image, availability, price), treats GTIN as optional, wants a full snapshot over SFTP at least daily, preferably as parquet, and requires a stable product URL with the variant selected. ACP itself is still maintained: the latest stable snapshot is dated 17 April 2026, and since 13 October 2026 Stripe's documentation lists Meta as a third co-creator. On the same day OpenAI documented OAI-AdsBot, a crawler that checks landing pages submitted for ChatGPT ads. Ads, not checkout, is where OpenAI is monetising shopping.

Our reading: the pivot settled the pattern for 2026. Discover in the assistant, buy on the merchant's own checkout. Google's Universal Cart offers both and lets the shopper choose. Do not build a ChatGPT checkout unless OpenAI enables it for you; do send the feed.

Google: UCP, Universal Cart and Merchant Center as the hub

Google has moved in three steps:

• Agentic checkout (13 November 2025): a price-tracking feature that, when a tracked product hits the target price, buys it on the merchant's own site with Google Pay after the shopper confirms. Early merchants: Wayfair, Chewy, Quince and some Shopify stores. Merchants need guest checkout and Google Pay.
• The Universal Commerce Protocol (11 January 2026, at NRF): an open standard co-developed with Shopify, Etsy, Wayfair, Target and Walmart and endorsed by more than twenty others including Adyen, American Express, Mastercard, Stripe, Visa and Zalando. It covers discovery through post-purchase, is compatible with A2A, AP2 and MCP, and is published at ucp.dev under Apache 2.0. The same post launched Business Agent, a branded chat agent on Search, and dozens of new Merchant Center attributes for product questions, accessories and substitutes.
• Universal Cart (previewed at I/O on 19 May 2026, launched 15 October 2026): one cart across Search and the Gemini app, with YouTube and Gmail to follow. It tracks price drops and restocks, uses Google Wallet to suggest the card with the best perks, and checks out with Google Pay through UCP at Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair and "many Shopify merchants", or moves the items to the merchant's cart. It draws on the Shopping Graph's 60 billion listings. US only at launch; the May post promised Canada and Australia "in the coming months" and the UK later.

The Merchant Center help page for UCP-powered checkout is precise about what a merchant needs: products eligible in the US, Canada or Australia; an interest form, because it is "available to select merchants"; the native_commerce checkout eligibility attribute on each listing; a Google Pay and Wallet Console account; and a payment service provider integrated with the Google Pay API, or your own encryption and PCI documentation. Payments are limited to standard card numbers stored in Google Wallet. Merchant Center "remains the central hub", and Google's feed documentation says that accurate product matching cannot be assured without a GTIN.

Google's own documentation for AI search says the same thing it says about agentic shopping: there is nothing special to do beyond the fundamentals. See our GEO report for how AI Mode retrieves and cites content, and note that Google's generative AI optimisation guide explicitly ignores llms.txt.

Shopify, Microsoft, Perplexity and Amazon

• Shopify is the merchant platform that appears in every announcement. It co-developed UCP, is a design partner of the Personal Agent Protocol, sells into Copilot through Shopify Catalog, and on 13 October 2026 opened Shop MCP at mcp.shop.com: one MCP connection, 17 tools, search across more than a billion products with no API key, carts and checkout on the selling store, Shop Pay for a signed-in shopper after they confirm the total, and order lookup. Orders appear in the merchant's admin like any other. There are no webhooks at launch, so an agent polls for checkout status. Merchants manage Google, Gemini, Copilot and ChatGPT from Agentic Storefronts in the admin.
• Microsoft made UCP-ready feeds in Microsoft Merchant Center generally available in the US on 21 April 2026, said Copilot Checkout now spans desktop and mobile with catalog data from more than 500,000 merchants, and linked Target Circle loyalty to Copilot. Copilot Checkout is for US buyers in English and US dollars. Microsoft's current page says a Merchant Center account is required; a January 2026 post said it was not. Follow the current page.
• Perplexity launched Instant Buy with PayPal for all US users on 25 November 2025, with more than 5,000 merchants whose catalogs sync through PayPal. PayPal handles payment, buyer protection and identity; the merchant stays merchant of record.
• Amazon runs Buy for Me (April 2025), which buys from other brands' websites inside the Amazon app, with Amazon inserting encrypted payment details and the brand handling delivery and returns. By November 2025 it covered more than 500,000 products, and CNBC reported on 6 January 2026 that several retailers objected to being included without consent. Amazon publishes no open protocol and takes no commission on Buy for Me.
• Anthropic published a commerce agents blueprint on 2 September 2026: open-source reference code for a shopping agent and a merchant agent, deployable on the Claude API, Bedrock, Foundry or Vertex, with Shopify, Visa and Mastercard as partners. It leaves payment to the merchant's existing checkout and says Claude is "an intelligence layer, not a storefront or checkout". There is no consumer checkout in Claude.

The protocol stack in one table

Twelve protocols and integrations, grouped by layer. Status is what the backer says, except where marked. The technical reference has endpoints, discovery files and a per-platform implementation guide.

Two frameworks sit beside the table. American Express launched its ACE Developer Kit and Agent Purchase Protection on 14 April 2026, a merchant playbook on 6 October 2026, and Intent Intelligence on 13 October 2026, which captures and authenticates what the shopper asked for before an agent pays and scopes the credential to it. The protection covers purchases by a registered agent that deviate from the authenticated intent, after the cardholder has tried a return first; the terms are not public yet. And the IETF's Web Bot Auth working group is standardising the HTTP message signatures that Visa's protocol, Mastercard's agent registration and Cloudflare's verified bots all build on.

The payment layer: proof of intent is the product

The card networks have converged on one idea: an agent payment needs a verifiable record of what the person authorised. Google's AP2 calls it a mandate. On 28 April 2026 Google donated AP2 and Mastercard's Verifiable Intent to the FIDO Alliance and released v0.2 with "Human Not Present" payments, where the agent buys later on instructions given in advance, such as tickets the moment they go on sale. AP2 plugs into UCP as an optional extension that only activates when both sides advertise it.

Visa's approach is agent identity first. The 2 July 2026 announcement from the Visa Payments Forum in Paris described live agent purchases at lastminute.com, Frasers, Cleverbridge and BrickDepot, authenticated with Visa Payment Passkeys to satisfy European Strong Customer Authentication, with 31 issuers including ING, Klarna, Revolut, Barclays, BBVA and Nordea. Mathieu Altwegg of Visa Europe: "We're now seeing AI agents buy on behalf of people directly with independent merchants." Mastercard's Agent Pay for Machines (10 June 2026) goes the other way, toward fractions of a cent paid by software to software, settled on cards, accounts or stablecoins, with Adyen, Checkout.com, Cloudflare, Coinbase, Stripe and Tempo among the first adopters.

Below the card networks sit two HTTP 402 protocols for paying per request: x402, whose public counter showed about 24 million dollars and 75 million transactions in the thirty days to 12 October 2026, and Stripe and Tempo's Machine Payments Protocol (18 March 2026), which adds sessions and subscriptions and can serve x402 clients. These matter for paid APIs and content more than for a webshop, but they are how an agent will one day pay for your product data.

Nobody has solved liability. The AP2 specification lists who pays for a wrong agent purchase as an open problem, Amex's protection is announced but not live, and the Amex merchant survey found 93% of merchants want the AI provider to pay. Until that is settled, an agent order is a card-not-present order with an extra audit trail.

Agent trust: bots, terms of service and the Amazon v. Perplexity ruling

A merchant that opens checkout to agents needs to tell a trusted agent from a scraper. Visa's Trusted Agent Protocol and Web Bot Auth answer that with signed requests the merchant verifies against a key directory, typically at the CDN. The protocol notes on the checklist recommend not blocking an agent only because it lacks a card scheme registration, since registration is not self-service yet.

The legal boundary was tested this year. Amazon sued Perplexity in November 2025 over its Comet browser agent logging into customers' Amazon accounts; a San Francisco judge granted an injunction on 9 March 2026. On 4 August 2026 the Ninth Circuit vacated it, holding that when a user tasks the agent, it is the user who "accesses" Amazon's computers under the Computer Fraud and Abuse Act, not Perplexity. The court called the agent a software tool under the user's direction, said the ruling does not limit Amazon's ability to enforce its terms of service, and left open whether a more autonomous agent would be judged differently. The case continues in the district court. For a merchant the lesson is practical: a blanket block on agents is a terms-of-service decision, not something the anti-hacking law does for you, and it makes you invisible on the surfaces where customers now compare.

Separately, decide what your robots.txt says. Agents that fetch a page because a user asked (Google's user-triggered fetchers, ChatGPT-User, Claude-User, Perplexity-User) are different tokens from training crawlers (GPTBot, ClaudeBot, Google-Extended). Blocking the first group blocks the shopper. The bot token list and the site audit show which ones your shop currently allows. The same split applies to publishers; see our news SEO article.

Europe and the Netherlands: rails yes, surfaces not yet

Every checkout surface in the table above is US-first. For a Dutch or European webshop the picture on 15 October 2026 is:

• The first European agentic payment ran through a Dutch merchant. On 2 June 2026, at Money20/20 in Amsterdam, Worldline, ING and Mastercard completed a live end-to-end agentic payment in production: an ING cardholder's agent bought concert tickets within a set budget from a merchant in the Netherlands, with explicit approval before payment. The same setup runs in Belgium.
• Visa's European programme includes ING, Klarna and Revolut among 31 issuers, with Payment Passkeys for Strong Customer Authentication, since 2 July 2026.
• Mollie in Amsterdam announced ACP compatibility on 24 November 2025 through the delegated payment API, and noted that OpenAI had not launched ACP in Europe. That is still the case. Adyen Agentic supports both ACP and UCP.
• Zalando endorses UCP and notes in its 2026 business update that the UCP checkout in AI Mode and Gemini is limited to the US.
• Google's UCP checkout is eligible in the US, Canada and Australia, with the UK announced. The Netherlands is not on any published list. Copilot Checkout is US only. Perplexity Instant Buy is US only.
• Thuiswinkel.org published the ShoppingTomorrow bluepaper "Agentic Commerce: the future of B2C" on 8 October 2026, with Sia, AWS, Sunweb, TUI and ANVR. Its advice: prepare now, optimise to be referenced by the agent rather than to attract visitors, and make two moves at once, be chosen by the agent (data, protocols, authority) and be trusted by the consumer (positioning, loyalty, own touchpoints, privacy, human recovery). Its sharpest number: 64% of the British retailers it studied got at least one wrong answer about themselves from an agent.

Regulation

• The UK Competition and Markets Authority published guidance on using AI agents on 9 March 2026. Its one-line summary: "if an AI agent you use does something illegal, you are responsible."
• The EU has no agent-specific consumer rules yet. CERRE's July 2026 issue paper by Christoph Busch says it is unclear whether contracts concluded through AI agents are valid under current consumer law, that the AI Act's ban on manipulative AI is untested for agentic commerce, and that the coming Digital Fairness Act may not cover agents. The AI Act's transparency duties for AI that interacts with people apply from August 2026. PSD2 and Strong Customer Authentication apply to every agent payment, which is why Visa's European programme runs on passkeys.
• The Dutch data protection authority warned on 12 February 2026 about autonomous agent platforms with broad access to a user's accounts, and stressed that the deployer stays accountable under the GDPR.

Our reading: a Dutch shop cannot sell through Gemini, Copilot or Perplexity checkout today, and should not build for it. It can be found by all of them, and by ChatGPT, through feeds and structured data, and it can already receive agent-initiated card payments through ING, Mastercard and Visa rails. Do the data work now; watch Google's country list and OpenAI's European launch; ask your PSP what it has switched on.

Product data is the base layer

Every surface in this article reads a feed. The feeds disagree on details, and the differences are where shops fail.

Three rules follow. First, add the manufacturer's GTIN everywhere one exists; Google ties matching to it and Microsoft rejects items without it. Second, price, availability, titles and specs must be identical on the product page, in the JSON-LD, in every feed and at the agent endpoint; Google Merchant Center requires price and availability to match the landing page and checkout. Third, keep the product URL stable with the variant selected, because OpenAI's feed uses it as the landing page and Google uses it for matching. Schema.org 30.1 (16 September 2026) added retail properties worth a look: isOftenBoughtWith, specification, consumerNotice, itemPopularity, and the EU Digital Product Passport fields. Jarno van Driel's profile covers structured data in depth.

Adobe's AI Content Visibility Checker scored US retailers' homepages at 75% and product pages at 66% machine-readable on average in April 2026. Product pages are the weak spot, and they are the pages agents need.

Measuring it: UTMs, referrers and agent orders

Agent purchases can happen over an API with no page view, and assistants strip referrers inconsistently. Expect undercounting, and build the baseline before the experiment:

• Tag your own feed URLs. OpenAI's spec only asks for a stable product URL and does not add UTMs. Use utm_source=chatgpt.com and either utm_medium=ai-assistant, which lands in GA4's built-in AI Assistant channel with no setup, or your own medium plus a custom channel group placed above Referral. Without that, the traffic lands in Unassigned. Google and Microsoft Merchant Center let you tag the ads redirect URL separately from the organic link. The UTM convention the author uses has per-surface templates.
• Segment on referrer hostnames chatgpt.com, gemini.google.com, perplexity.ai, copilot.microsoft.com and claude.ai. These are observed, not documented by the vendors.
• Log agent user agents at the server or CDN, and API and MCP calls if you expose any.
• Mark agent orders in the order system so finance, support and fraud can see them. Shopify shows Shop MCP orders in the admin; UCP and ACP sessions carry an agent identifier.
• Run a hold-out if you want to know whether the revenue is incremental or moved. Nobody has published that number yet.

The business case: a sizeable bet, not a sure thing

The honest version of the case, drawn from the business page of the tracker:

• Facts: customers can already buy through Google, Copilot and Perplexity in the US. One UCP integration reaches Google and Microsoft. Better product data helps every channel. You keep the order and the customer.
• Assumptions: that purchase without a page load converts better, that early presence compounds, that agents will prefer shops with protocol endpoints over shops without. No public data supports or refutes these yet.
• Risks: specs change (UCP, ACP and AP2 are beta or pre-release and the author expects breaking changes); platform strategy changes (OpenAI, March 2026); fraud and bot traffic on a young trust layer; unclear liability for wrong purchases; revenue that is shifted rather than added; weaker attribution.

Size the bet before you place it: what it costs to build and run, what it is worth if it only moves existing sales, and what breaks if the spec or the platform changes. Prefer whatever your platform or payment provider maintains for you. Choose between a small experiment (one surface, part of the catalog, a fixed period, a stop rule) and a moonshot (your own agent, a ChatGPT app, a custom UCP implementation) with a budget you can lose. Waiting on checkout is cheap. Waiting on product data and measurement is not.

On the moat: you keep the customer relationship, but the assistant does the comparison. If price or catalog breadth is your only advantage, agents make comparison easier for the shopper and harder for you. The Thuiswinkel bluepaper's framing is the right one: the agent chooses what can be measured, the consumer chooses what has meaning, and you need both.

The 2026 agentic commerce checklist

1. Fix the product data: GTINs wherever assigned, valid check digits, the seven Google and nine OpenAI required fields, consistent price and stock across page, JSON-LD, feeds and checkout, stable variant URLs.
2. Feed every surface you can: Google Merchant Center, Microsoft Merchant Center (import your Google feed), and apply at chatgpt.com/merchants if you sell in the US. Add Product and Offer JSON-LD to every product page.
3. Check what you block: allow user-triggered agents and AI search crawlers on product pages; decide on training crawlers separately. Run the site audit to see your discovery files, bot access and markup.
4. Set the baseline: UTM-tagged feed links, an AI assistant channel in GA4, referrer segments, agent user agents in logs, agent orders flagged in the order system.
5. Ask your platform and PSP: Shopify Agentic Storefronts and Shop MCP; Adyen, Stripe, Mollie and PayPal all have ACP or UCP routes. Turn on what they maintain before building anything.
6. Pick one experiment: one surface, part of the catalog, a period and a stop rule. Agree with finance, risk and the PSP on agent orders, SCA and disputes. For Dutch shops the surface today is discovery, not checkout.
7. Do not build ChatGPT checkout unless OpenAI enables it for you. Send the feed.
8. Prepare the trust layer: verify Web Bot Auth signatures at the CDN, retrieve Visa's agent keys if you accept Visa in Europe, and do not block an agent only for lacking a scheme registration.
9. Watch, do not build: Personal Agent Protocol, AP2 at FIDO, Know-Your-Agent, Amex's protection terms, Google's country list, OpenAI's European launch, and EU guidance on agent contracts.
10. Keep the human path: the surveys agree that people will let an agent research and compare but want to approve the payment. Make confirmation, returns and human support easy for an order that arrived through an agent.

Frequently asked questions

Sources and credits

Research for this article was carried out on 15 October 2026. The protocol, feed and bot facts come from the author's tracker at vreeman.ai/commerce, which links each claim to the most official source and marks items that rest on press or trackers rather than the backers. Some Google Merchant Center and OpenAI pages block automated requests; those were checked from saved copies on 12 October 2026, and the links here point to the official pages. Primary sources by group:

• Google: UCP announcement · Universal Cart at I/O · Universal Cart launch · Agentic checkout · UCP guide · UCP-powered checkout help · Product data specification · AP2 to FIDO · Optimizing for generative AI features · GA4 channel groups
• Protocols: ucp.dev · agenticcommerce.dev · Stripe ACP docs · Stripe ACP announcement · Personal Agent Protocol · Sierra announcement · AP2 · Visa Trusted Agent Protocol · Visa Europe, 2 July 2026 · Mastercard Agent Pay · Agent Pay for Machines · Know-Your-Agent · Amex ACE · Amex Intent Intelligence · x402 · MPP · PayPal Agent Ready · MCP · A2A · Web Bot Auth
• Platforms: OpenAI, product discovery · OpenAI product feed spec · OpenAI crawlers · Shopify Shop MCP · Shopify Agentic Storefronts · Shopify AI search insights · Microsoft Advertising, April 2026 · Microsoft Merchant Center catalog · Retail Dive on Buy for Me · CNBC on Buy for Me · Claude commerce agents · Adyen Agentic · Mollie
• Data: Adobe, first quarter 2026 · Adobe, July 2026 · Salesforce holiday 2025 · Contentsquare 2026 benchmark · Accertify via Digital Transactions · Accenture Consumer Pulse 2026 · Amex playbook and Trendex · Rezolve study
• Europe and law: Worldline, ING and Mastercard · Thuiswinkel.org bluepaper · Zalando business update 2026 · CMA guidance · CERRE issue paper · Autoriteit Persoonsgegevens warning · Ninth Circuit, Amazon v. Perplexity
• Author's tracker: Overview and audit · Business · Tech · Signals · Checklist · MCP server

About the author

Simon Vreeman. Freelance Growth & Technical Marketing Advisor and Growth Advisor at Eidra, Simon Vreeman excels in Growth & Technical Marketing, blending CRO, SEO, Web Analytics, and AI. He builds AI tools and agent-ready websites, champions becoming a "semantic architect for the AI age," and is a community speaker at Google Search Central Live Deep Dive Europe 2026.

Website: vreeman.com · LinkedIn · Full profile and Q&A with Simon Vreeman