| Quick answer
What is agentic commerce, and what should WooCommerce stores do about it? What it is: Agentic commerce is when AI agents in ecommerce, like ChatGPT, Gemini, or Copilot, find, compare, and sometimes buy products for a shopper using a store’s product data: its feed, attributes, prices, and stock. |
What the evidence shows:
- AI discovery is growing. AI traffic to US retail sites grew 393% year over year in Q1 2026, and in March it converted 42% better than non-AI traffic (Adobe).
- Product data is the weak link. Retail product pages scored 66% for machine readability, the lowest of any page type Adobe measured (Adobe).
- Shoppers are cautious about agents paying. Half of Americans worry about the security of their payment details when AI shops for them (Global Payments).
- WooCommerce is partway there. Native MCP is still a developer preview (WooCommerce docs), and neither checkout protocol, ACP or UCP, is built into WooCommerce core.
TL;DR for stores doing $300K to $4M
- Build now: clean product data and an accurate feed into ChatGPT and Google.
- Build second: read-only MCP access for your own team.
- Wait: native agent checkout, until native support and conversion data catch up.
- Ignore: headless rebuilds, custom protocol adapters, and bets on a protocol winner.
Walmart put about 200,000 products into ChatGPT’s in-chat checkout. Those purchases converted at one-third the rate of shoppers who clicked through to Walmart’s own site. By March 2026, OpenAI had stepped back from standalone in-chat checkout too.
Agentic commerce is when an AI agent finds, compares, and sometimes buys products for a shopper, working from your store’s data: your feed, your product attributes, your stock, and your prices.
This is our point of view at WisdmLabs, written for owners of stores doing $300K to $4M who are deciding what to build this year. We’ll sort the work into three piles: build now, wait, and ignore.
Where agentic commerce hurts at $300K to $4M
If you run a store this size, you’re probably not short on AI ideas. However, you’re probably short on developer hours, clean data, and a straight answer about which of these pitches will pay off.
Here’s what we hear from store owners and decision-makers at this stage:
- Every vendor is selling “AI-ready.” Apps, agencies, and platforms all have an agentic commerce pitch. Few of them tell you what to skip.
- Your catalog has years of shortcuts in it. Products added by different people, attributes typed into descriptions, variations set up three different ways. It works for human shoppers. Agents read it differently.
- Your developer time is already spoken for. Whether it’s a freelancer, an agency retainer, or one in-house developer, every AI project competes with checkout fixes, speed work, and the next sale.
- You can’t see AI traffic clearly. Some visits arrive tagged, many don’t. So it’s hard to know if any of this is worth money yet.
- Your store isn’t standard. Subscriptions, wholesale pricing, bundles, and custom checkout steps are often why you chose WooCommerce. They’re also the parts agents handle least well.
The rest of this article answers those five problems in order of what to do first.
| If you’re scaling fast ($2M to $4M+ and growing): your problems look different. Your catalog is growing faster than anyone can keep tidy, your ops team spends hours on order lookups, and you may be selling in more than one country or to wholesale buyers. We call out where the advice changes for you throughout. |
What agentic commerce means for a WooCommerce store today
Agentic commerce touches your store at three layers. Each one has its own tools and moves at its own speed, so it’s easier to decide on them one at a time.
Discovery: product feeds decide whether agents can recommend you
Before an agent can suggest your product, it has to know the product exists and understand what it is. That happens through product feeds and structured data. OpenAI takes merchant feeds through an application on its merchant page, and Google pulls from the Merchant Center.
If your data is thin, agents skip you. They can’t recommend a size, fit, or compatibility detail you never published.
Access: MCP lets agents read and act on your store
MCP (Model Context Protocol) is an open standard that lets AI assistants like Claude connect to software and use it. WooCommerce has shipped native MCP support since version 10.3, and its documentation still labels it a developer preview.
In practice, MCP is how an assistant could look up an order, check stock, or update a product listing in your store. Right now, your own team gets the most use out of it.
Checkout: ACP and UCP decide where the sale closes
Two competing standards cover what happens when an agent tries to complete a purchase. ACP (Agentic Commerce Protocol) came from OpenAI and Stripe in September 2025. UCP (Universal Commerce Protocol) came from Google, built with Shopify, Etsy, Wayfair, Target, and Walmart, in January 2026.
Both define how an agent passes cart, payment, and order details to a merchant. And neither is built into WooCommerce core today.
| Layer | Protocol or standard | Who’s behind it | What it does | WooCommerce status today |
|---|---|---|---|---|
| Discovery | Product feeds (OpenAI feed format, Google Merchant Center) | OpenAI, Google | Tells agents what you sell, at what price, and whether it’s in stock | Works now through feed plugins |
| Access | MCP | Open standard introduced by Anthropic, built on the WordPress Abilities API in Woo | Lets AI assistants read and change store data | Native, still a developer preview |
| Checkout | ACP | OpenAI and Stripe | Agent-to-merchant checkout | Announced for Woo through the Stripe extension (waitlist, Dec 2025) |
| Checkout | UCP | Google and Shopify, with a tech council that now includes Amazon, Meta, Microsoft, and Stripe | Agent-to-merchant checkout on Google surfaces | No native support; third-party plugins only |
| What this means for you: only one of these three layers is ready for a WooCommerce store to use with confidence today, and it’s the cheapest one. That’s good news for your budget. |
Why AI agents in ecommerce stalled at checkout but kept growing at discovery
In-chat checkout had a rough 2026. OpenAI’s merchant page now says it plainly: “We’re moving away from a standalone Instant Checkout experience in ChatGPT and prioritizing better shopping discovery and merchant-owned checkout experiences.”
Walmart’s numbers show why. Daniel Danker, Walmart’s EVP of Product and Design, called the in-chat experience “unsatisfying,” according to Search Engine Land. Walmart is moving to its own Sparky assistant inside ChatGPT, with purchases completing in Walmart’s system.
Shoppers finding products through AI is a different picture. Adobe found that AI traffic to US retail sites grew 393% year over year in Q1 2026. In March, that traffic converted 42% better than non-AI traffic.
Agents are sending buyers. Those buyers still want to check out with you.
Shoppers are split on letting an agent pay. In Global Payments research covering 8,027 consumers across seven countries, 69% were comfortable with AI buying groceries or clothing up to $100. That fell to 58% for electronics. Half of Americans said they worry about the security of their payment details.
The bigger surveys show where shoppers draw the line. Menlo Ventures’ 2026 State of Consumer AI surveyed 5,067 US adults with Morning Consult. Among shoppers who use AI, 50% used it to compare prices and 43% used it to narrow a shortlist. Only 26% let it complete a purchase.
Accenture’s Consumer Pulse Research 2026, covering 25,590 consumers in 16 countries, shows the same drop-off. Most (74%) would have an agent handle tasks like reordering or renewing a subscription. 32% would let it choose a product within a set budget. Just 9% would let it start and finish a purchase on its own.
Shoppers hand agents the research and the routine reorders. They keep the final click for themselves. Repeat, low-stakes purchases suit agents well, and considered purchases still bring shoppers back to your product page.
| The honest read for your store: this is good news if your product pages and checkout already convert well. AI sends you a shopper who’s done their research, and your site closes the sale. If your product pages are weak, AI traffic won’t fix that. It’ll just show it to more people. |
Build now: fix product data and get your catalog into AI shopping
Product data is the one investment that pays off no matter which protocol wins. Feeds, Google Shopping, on-site search, and AI agents all read the same attributes.
Adobe measured how readable retail websites are for AI, and product pages scored just 66%. That’s the lowest of any page type it checked. Adobe says this means “much of the content is currently invisible to LLMs.” And product pages are where your sales happen.
Often it comes down to small formatting slips. In a WordPress.org support thread from September 2026, a store owner’s ChatGPT feed got rejected for misformatted columns. The feed said “in stock” where OpenAI wanted in_stock, and a price like “1,099.90 RON” failed because of the comma.
Most agent-readiness work is careful catalog cleanup.
What agent-readable WooCommerce product data looks like
Agents compare products on facts. If a shopper asks for “a waterproof hiking boot under $150 in wide fit,” your listing has to state waterproofing, price, and width in structured fields an agent can read.
Run through this before you touch any protocol.
Agent-ready product data checklist
- Confirm every product has a GTIN, MPN, or brand-plus-SKU identifier
- Store size, color, material, and compatibility as WooCommerce attributes, separate from the description
- Check that each variation has its own price, stock status, SKU, and image
- Write descriptions that state specs in plain sentences before the marketing copy
- Format feed prices without thousands separators and with a currency code
- Map stock status to the exact values your feed spec expects
- Publish shipping costs, delivery times, and return windows where crawlers can read them
- Add Product schema markup and test it
Our SEO for WooCommerce guide covers the structured data side in more depth.
| Do this next (under $1M, owner-led): don’t try to fix the whole catalog. Export your top 20 to 50 products by revenue and run them through the checklist above. Those are the products an agent is most likely to recommend, and you or a VA can usually fix them without a developer. |
| If you’re scaling ($2M to $4M+): one-off cleanup won’t hold. Once several people add products, the data drifts back within months. Set required attributes per category, give one person ownership of product data, and have your developer add checks that block a product from publishing with missing fields. That’s the step that keeps your feed accurate as the catalog grows. |
| If you sell B2B or use customer-specific pricing: a public feed shows your public price. Agents won’t see wholesale tiers or member discounts that sit behind a login. Decide which price you want AI shoppers to compare against before your feed goes live |
Getting your WooCommerce catalog into ChatGPT and Google
For ChatGPT, OpenAI takes catalog submissions through an application form and says there are “no fees on purchases that start in ChatGPT.” Shopping there is US-only for now. Shopify and Etsy catalogs are connected automatically, so WooCommerce stores need to send a feed.
That’s why we at WisdmLabs built the free Product Feed for ChatGPT plugin. It turns a WooCommerce catalog into JSON or CSV for ChatGPT, maps variable products, and regenerates the feed on a schedule you set.
For Google, the Merchant Center is the entry point. It powers Google Shopping today, and Google lists it as a requirement for agentic checkout in AI Mode and Gemini.
| Do this next: if you already run Google Shopping, your Merchant Center feed is your starting point. Fix the errors it already flags before you build anything new. If you sell mainly outside the US, put Google first, since ChatGPT shopping is US-only for now. |
Build second: open your store to MCP, read-only first
Once your data is clean, MCP is the next sensible step. Let your own team use it first, then decide about outside agents.
Picture your store manager asking Claude, “Which orders from last week are still unfulfilled?” or “Which products are low on stock?” and getting an answer without opening five admin screens. That’s where MCP starts paying off.
What WooCommerce’s native MCP can and can’t do today
WooCommerce 10.9 added canonical abilities for products and orders. Agents can query, create, update, and delete products. They can query orders, change order status, and add order notes. It all runs on the same store logic as the WooCommerce REST API.
The data those abilities return is still thin. An open GitHub issue filed in October 2026 notes that orders come back with “no customer name, address, phone, country, shipping method or customer note.” Products come back with “no categories, tags, attributes, weight, dimensions or shipping class.”
The same issue flags a gap for stores that run extensions: “An extension that adds data to orders… can show it everywhere in WooCommerce except to agents.” If your store depends on custom fields, plan for custom abilities.
Native MCP gives you a good base to build on, but it’s still a preview. Test it on staging, and keep it away from customer data until you’ve decided who can see what.
Why read-only access comes before agents that take action
Store owners tend to underestimate permissions. The WooCommerce MCP docs warn that order and customer operations can expose personal data.
Plugin developers are being careful here too. When a store owner on the WordPress.org forums asked whether an MCP plugin covered WooCommerce endpoints, its author replied: “I don’t want to just expose every WooCommerce REST endpoint blindly through MCP.”
We agree with him. Our WordCamp US recap lays out the order that works: make data readable, set permission boundaries, and keep an audit trail before you grant any write access.
| Business Owner takeaway: Decide what an assistant may read and what it may change before you connect one. That list is a business decision, so make it yourself. |
We at WisdmLabs built AI Conduit for this stage. It’s a free plugin that gives MCP-compatible assistants read-only access to your WordPress data, with role-based permissions and a log of every request.
If your developer already works with MCP, testing native write abilities on staging is reasonable. For anything that needs store-specific logic, our AI automation team builds custom abilities scoped to exactly what an agent should touch.
| Do this next (under $1M): MCP is optional for you this year. If your team is two or three people and you can answer order questions in a minute, the payoff is small. Come back to it when order lookups start eating real hours. |
| If you’re scaling ($2M to $4M+): this is where MCP starts to pay off. List the questions your ops and support team answer most often, like order status, stock levels, and which products sell together. If they come up dozens of times a week, read-only MCP access can take a lot of that load off. |
Selling to EU customers? Check how order data reaching an AI tool fits your privacy obligations before you connect anything.
Wait: native agent checkout through ACP or UCP
We’d hold your checkout budget for now. WooCommerce stores have paths to agent checkout on paper, but none is mature enough to justify a custom build at your size.
On the ACP side, WooCommerce announced in December 2025 that Stripe’s Agentic Commerce Suite would come to Woo stores through the official Stripe extension, starting with a waitlist. Since then, ChatGPT, the most visible ACP surface, has moved purchases back to merchant sites.
On the UCP side, the protocol’s tech council added Amazon, Meta, Microsoft, Salesforce, and Stripe in April 2026, per PPC Land, while ACP stays active alongside it. WooCommerce hadn’t documented native UCP support as of late summer 2026, so today’s options are third-party plugins or a custom adapter.
If you build a checkout for a protocol now, you may be rebuilding it next quarter.
| The other side of this: waiting has a cost too. Shopify stores can already sell inside some AI channels, so a competitor on Shopify may show up with a buy button where you don’t. For most Woo stores your size, that gap is still small next to the cost of a custom build. Keep an eye on it anyway. |
The signals that would make agent checkout worth building
We’d revisit this call when one or more of these shows up.
| Signal | What to watch | Why it changes the call |
|---|---|---|
| Native support lands | WooCommerce ships ACP or UCP support, or Stripe’s suite becomes generally available for Woo | You’d configure checkout instead of building it, which cuts cost sharply |
| The protocols settle | ACP and UCP converge, or a major AI platform drops one | Less risk of building for a standard that fades |
| Agent checkout converts | Public data shows in-chat purchases matching click-out conversion | The gap Walmart saw has closed |
| Your own numbers move | AI-referred orders become a meaningful share of your revenue | The build can pay for itself |
Start tracking AI referrals in GA4 now. ChatGPT tags many outbound links with utm_source=chatgpt.com, and referrals from Perplexity and Gemini show up by domain, so you’ll see the last signal when it arrives.
| Do this next: set up a GA4 report for AI referrals this week and check it monthly. Not every AI visit arrives tagged, so treat what you see as a floor. If you run subscriptions or custom checkout steps, ask your developer to list which of those an agent checkout would need to support. That list becomes your requirements doc when the time comes. |
What to ignore for now (and the stores where that advice flips)
Some agentic commerce advice is written for enterprise retailers, and some is written to sell you a rebuild. At $300K to $4M, here’s what we’d skip.
- A headless rebuild because “the frontend is dead.” Shoppers who arrive from AI tools still land on your product pages and check out there.
- Custom protocol adapters. Writing your own ACP or UCP layer means maintaining it through every spec change.
- Betting on a protocol winner. The governance is still moving, even among the companies backing each one.
- A custom shopping agent for your own site before your catalog data can support it.
Our AI readiness checklist applies the same principle more broadly: prove the business problem before you fund the build.
Some stores should do the opposite. If you sell low-consideration, repeat purchases like consumables, pet food, or replenishment items, shoppers already feel comfortable letting AI buy in your category. Watching agent checkout closely and joining waitlists early makes sense for you.
What you sell matters more than how loud the hype gets.
| A quick test before you say yes to any AI pitch: ask the vendor three things. Which of my current products will this help sell? How will I see the result in my own analytics? What happens to it if the protocol changes? If they can’t answer all three, it can wait |
Should WooCommerce AI readiness push you toward Shopify?
On agent distribution today, Shopify is ahead. It co-developed UCP, launched Agentic Storefronts to manage ChatGPT, Copilot, and Google AI channels from one admin, and Shopify catalogs feed into ChatGPT automatically.
It has also opened that reach to brands on other platforms. The Shopify Agentic plan syndicates products to ChatGPT, Google AI Mode and Gemini, Copilot, and Perplexity with “no Shopify online store required,” no monthly fee, and card rates from 2.9% + 30¢.
For a WooCommerce store, that’s worth evaluating as a sales channel. You’d keep your store and test agent sales through Shopify’s checkout, paying per transaction.
WooCommerce’s case rests on different strengths. You own your data and hosting, MCP is native, and custom pricing, B2B, or subscription logic stays fully in your control. When agents read your store, they read your real business rules.
Agent readiness alone isn’t a reason to migrate. Weigh it alongside cost, customization, and control, which our WooCommerce vs Shopify comparison breaks down.
| How to think about this at your stage: if you’re under $1M with a simple catalog, a Shopify Agentic plan test is a low-risk way to see whether AI channels sell for you at all. If you’re scaling with subscriptions, wholesale, or custom logic, a second checkout means a second place to manage orders, stock, and customer data. Test with a small product set, and let the numbers decide. |
Your next 90 days: a practical sequence
If you only take one thing from this article, take this order of work.
| When | Under $1M (owner-led) | $1M to $4M+ (scaling team) |
|---|---|---|
| Weeks 1 to 2 | Set up AI referral tracking in GA4. Run your top 20 to 50 products through the data checklist. | Set up AI referral tracking. Audit your Merchant Center feed errors. Name one owner for product data. |
| Weeks 3 to 6 | Get your catalog into ChatGPT (US) and Google Merchant Center. | Fix attributes category by category. Add required-field checks so new products can’t go live incomplete. Submit to ChatGPT. |
| Weeks 7 to 12 | Check your AI referral numbers. Decide whether a Shopify Agentic plan test is worth it. | List your team’s most common order and stock questions. Pilot read-only MCP access on staging with clear permissions. |
| Ongoing | Review AI traffic monthly. Revisit checkout when the signals table above changes. | Same, plus a quarterly review of which agent checkout requirements your store would need. |
None of this needs a big budget. Most of it is cleanup and tracking you’d benefit from even if AI shopping disappeared tomorrow.
Is your WooCommerce store ready for agentic commerce? A 6-question self-check
Answer yes or no to each.
- Does every product have a GTIN, MPN, or brand-plus-SKU identifier?
- Are size, color, material, and compatibility stored as WooCommerce attributes?
- Is your catalog live in a feed for ChatGPT or Google Merchant Center?
- Can you see AI-referred sessions and orders in your analytics?
- Have you decided which store data an AI assistant may read, and which it may change?
- Is one person responsible for product data quality?
5–6 yes: You’re in good shape. Read-only MCP and channel tests like Shopify’s Agentic plan are sensible next steps.
3–4 yes: The foundation is partly there. Close the data and tracking gaps before adding agent access.
0–2 yes: Start with the product data checklist above. It’s the cheapest item on this list, and it helps your Google Shopping results too.
So, your score shows you where to start. The answers below cover what high-scaling store owners usually ask next, from protocols and fees to whether any of this pays off at your size.
FAQs
Do I need to choose between ACP and UCP?
Not yet. Both are active, the UCP tech council now includes Stripe, and WooCommerce has native support for neither. Clean product data and a working feed serve you under either standard.
Is WooCommerce MCP safe to turn on for a live store?
WooCommerce labels its MCP integration a developer preview and warns that order and customer operations can expose personal data. Test it on staging first, use read-only access where you can, and decide which roles can reach customer information before connecting any assistant.
Does ChatGPT charge fees when shoppers find my products there?
OpenAI’s merchant page says there are no fees on purchases that start in ChatGPT, and checkout completes on your own site. Shopping discovery is US-only for now, and stores outside Shopify and Etsy apply through a form.
Can a WooCommerce store sell through Google AI Mode today?
Your products can appear through Google Merchant Center feeds. Agentic checkout in AI Mode runs on UCP, which WooCommerce doesn’t support natively yet. In-chat buying needs a third-party plugin, a custom build, or a channel like Shopify’s Agentic plan.
Is agentic commerce worth the investment for a store my size?
The data and feed work is worth it now, because it also improves Google Shopping and on-site search. Agent checkout can wait until your own numbers show demand. When merchants in a Shopify Community thread asked for real results, agencies reported only “a growing trickle” of ChatGPT sessions, with no sales figures shared.
I’m growing fast. Should I hire for this or bring in a partner?
It depends on what’s in the way. If the work is mostly product data cleanup, an in-house catalog owner usually pays off over time. If it’s custom MCP abilities, feed logic for complex products, or privacy checks, a partner gets you there faster without a permanent hire.
If you’ve read this far, you probably have a catalog that needs work and a list of AI ideas competing for the same developer hours.
That’s a normal place to start. If you want someone to look at your product data, your stack, and your AI traffic before you spend on any of this, here’s how we work at WisdmLabs:
1. A quick call (30 minutes). We look at your store, your catalog, and what AI traffic you’re getting so far. No sales deck, no upsell.
2. A clear scope. You get a plain-language plan covering what to fix now, what to wait on, and what it costs. Before anything starts.
3. We build it. Feed fixes, read-only MCP access, or custom abilities. Our team handles the technical side, and you’re involved where your input matters.
4. You review, we launch. Nothing goes live until you’re happy with it.
5. You own it. Everything is documented and handed over. You don’t need us to keep it running unless you want to.
