Skip to content

Agentic SEO for Magento and Adobe Commerce: what’s shipped, what isn’t

Picture of Rishabh Gupta

As of 4 October 2026, Adobe has shipped tools that measure and improve how AI assistants read an Adobe Commerce catalog, but not a documented way for a Magento store to sell inside ChatGPT, Gemini or Copilot. Adobe has committed to Google’s Universal Commerce Protocol (UCP) and OpenAI’s Agentic Commerce Protocol (ACP), yet I couldn’t find setup docs or release notes for either. Adobe’s Commerce MCP server, which would give AI agents live store data and tools from product discovery to checkout, is still marked coming soon. None of the commerce-specific features are documented for Magento Open Source.

So making your products readable to assistants is still your job. On Luma or Hyvä, start with robots.txt and your CDN’s bot rules, then fill the gaps in your Product markup and feed. On PWA Studio or Adobe’s Edge Delivery storefront, first check that product content is in the HTML before JavaScript runs.

In short
  • As of 4 October 2026, Adobe has no documented way for a Magento store to sell inside ChatGPT, Gemini or Copilot. Commerce MCP is still coming soon
  • On Luma or Hyvä, start with robots.txt and your CDN’s bot rules, then fill the gaps in Product markup and the feed
  • On PWA Studio or the Edge Delivery storefront, check that product content is in the HTML before JavaScript runs. Prerender is opt-in
  • None of the commerce AI features are documented for Magento Open Source. Making products readable to assistants is still your job
0setup docs found for Adobe’s committed UCP and ACP support, eight months after the announcementAdobe blog, February 2026; checked 4 October 2026
3places where Adobe’s own pages disagree: UCP and ACP, MCP, and Catalog AgentAdobe product page against Experience League
18 Aug 2026Content API for Shopping sunset. Feed tools that push to Merchant Center need the Merchant APIGoogle
On this page
  1. What’s shipped vs what isn’t
  2. What does a Magento page show an AI crawler?
  3. Feeds after the Google Shopping ads Channel
  4. Checklist for Magento Open Source, by impact
  5. Checklist for Adobe Commerce, by impact
  6. What can an AI agent see in a B2B catalog?
  7. What to leave alone for now

What’s shipped vs what isn’t

Statuses as of 4 October 2026. In Adobe’s feature comparison, PaaS is Adobe Commerce on cloud, which you patch and upgrade yourself. Adobe Commerce as a Cloud Service is SaaS that Adobe upgrades for you.

Item Status Applies to Source
Brand Visibility (was LLM Optimizer) GA in August 2026 Standalone Adobe app Release notes, 2025 launch
Catalog Agent and catalog enrichment Blog says available, docs say beta PaaS with Brand Visibility Experience League
Product page enrichment for AI agents Available, served to AI agents at the CDN Commerce with Brand Visibility Experience League
Brand Concierge GA since November 2025. Discovery only, checkout planned Connects to Commerce PaaS via MCP FAQ, launch post
Commerce MCP server Coming soon Adobe Commerce Experience League
UCP and ACP support Committed in February 2026, no setup docs found Edition not stated Adobe blog
Agent Payments Protocol (AP2) Adobe was a launch partner, nothing to configure None yet Google Cloud
Commerce Prerender Shipped, opt-in Edge Delivery storefronts Experience League
Adobe Commerce as a Cloud Service Shipped as SaaS, no Luma New builds Experience League
Adobe Commerce Optimizer Shipped as SaaS. Improves on-site search Commerce and other backends Release notes
Google Shopping ads Channel extension End of life, deprecated in 2.3.5 All editions Adobe KB
ChatGPT shopping Live in the US, feeds for approved partners Stores that apply OpenAI
UCP checkout in AI Mode and Gemini US pilot merchants only US products in Merchant Center Google
Copilot Checkout Rolling out in the US since January 2026 PayPal, Stripe and Shopify merchants Microsoft

Adobe’s own pages disagree in three places, so ask your account team for the documentation page, not the slide.

Adobe’s April Summit post also says its PayPal partnership is bringing feed-based checkout to Perplexity and Copilot, without a date.

What an AI crawler gets from each storefront
Luma and HyväServer-rendered. Name, price and description are in the HTML. Markup gaps: availability, brand, GTIN, shipping, returns
PWA StudioHTML shell, React builds the page. Needs server-side rendering or prerendering
Edge Delivery storefrontClient-side product rendering. Commerce Prerender fixes it, but it is opt-in
Vercel’s December 2024 analysis found none of the major AI crawlers it measured ran JavaScript, except Gemini on Googlebot’s infrastructure.

What does a Magento page show an AI crawler?

Rendering comes first

Luma and Hyvä render product pages on the server, so the name, price and description are in the HTML a crawler downloads. Hyvä’s theme became free and open source in November 2025.

PWA Studio’s UPWARD server sends an HTML shell and React builds the page in the browser. Adobe’s Edge Delivery storefront, the default for Adobe Commerce as a Cloud Service, renders product information with client-side JavaScript, which Adobe says “limits discoverability for search engines and AI systems that don’t run JavaScript.” Its fix, AEM Commerce Prerender, publishes product and category HTML ahead of time, but it’s opt-in. Vercel’s December 2024 analysis found none of the major AI crawlers it measured executed JavaScript, except Gemini, which uses Googlebot’s infrastructure.

Structured data in Luma and Hyvä

Luma’s product page uses schema.org microdata: name, SKU, description and image, an Offer with price and currency, plus ratings and reviews. There’s no availability, brand, GTIN, MPN, item condition, shipping or return policy, and Magento’s core storefront templates contain no JSON-LD. Hyvä’s default theme adds availability and priceValidUntil to the Offer but drops the SKU and description properties, and its product page BreadcrumbList JSON-LD is added by JavaScript.

Google lists availability, brand, GTIN, MPN, shippingDetails and hasMerchantReturnPolicy as recommended for merchant listings, and its variant guidance uses ProductGroup with hasVariant and variesBy, which maps to configurable products. Stock, delivery and returns are the facts a purchase turns on, so add them on either theme.

Magento also ships with canonical tags off for categories and products. Turn on both Use Canonical Link Meta Tag settings under Stores > Settings > Configuration > Catalog > Catalog > Search Engine Optimization.

robots.txt, sitemaps and layered navigation

robots.txt lives at Content > Design > Configuration. Edit the Global or website row, open Search Engine Robots, and use “Edit custom instruction of robots.txt File”, which is set per website. The same section’s Default Robots should read INDEX, FOLLOW on production. On Adobe Commerce on cloud, robots.txt is generated on request and indexing can only be enabled in Production.

Be careful with Reset To Defaults. It replaces your instructions with Magento’s default block, which includes Disallow: /*?. That keeps crawlers out of layered navigation URLs, as Google’s faceted navigation guidance recommends when you don’t need filtered pages indexed. It also blocks any product URL with a query string, including feed links with tracking parameters, and OpenAI’s feed best practices suggest adding utm_medium=feed. Test the exact URLs your feeds send against your rules.

For sitemaps, open Stores > Settings > Configuration > Catalog > XML Sitemap. Set Generation Settings > Enabled to Yes so cron rebuilds the file, set Enable Submission to Robots.txt to Yes, and pick Base Only or All for Add Images into Sitemap. Create one sitemap per store view under Marketing > SEO & Search > Site Map.

Bot protection on Fastly and Cloudflare

On Adobe Commerce on cloud, Adobe and Fastly manage the WAF, and Adobe lists bot mitigation as something it doesn’t provide. Blocking a crawler there takes a custom VCL snippet or edge ACL, which can match on the User-Agent header.

Cloudflare’s docs list the Block AI bots setting as deprecating on 15 September 2026, replaced by AI bot policies that treat Search, Agent and Training bots separately. From that date, new domains block Training and Agent bots on pages that display ads and allow Search. Blocking training, with either setting, also blocks crawlers that do both search and training. AI Crawl Control lets you allow or block single crawlers, so use it to keep the search crawlers below allowed.

Allow the crawlers that put you in answers: OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot and Storebot-Google. Google’s AI Overviews and AI Mode only link to pages that are indexed and eligible for a snippet. ChatGPT-User and Perplexity-User fetch pages for a user’s question, and OpenAI and Perplexity say robots.txt may not apply to them, so only a CDN rule blocks them. Training crawlers such as GPTBot and ClaudeBot are a separate decision. Google-Extended is a robots.txt token, not a crawler, so only robots.txt controls it. Google says it doesn’t affect Search inclusion or ranking, but blocking it also stops the Gemini app grounding answers in your pages.

Allow: search and user fetch
  • OAI-SearchBot, PerplexityBot, Claude-SearchBot
  • Googlebot and Storebot-Google
  • ChatGPT-User and Perplexity-User: only a CDN rule stops them
Decide separately: training
  • GPTBot, ClaudeBot
  • Google-Extended is a robots.txt token, not a crawler
  • Blocking it also stops Gemini grounding in your pages
Where to check
  • robots.txt per website in Content > Design > Configuration
  • Cloudflare AI bot policies and AI Crawl Control
  • Fastly VCL snippets and edge ACLs on Adobe Commerce on cloud

Feeds after the Google Shopping ads Channel

With the Google Shopping ads Channel deprecated since 2.3.5, Magento stores feed Merchant Center through a Marketplace extension, a feed tool or a custom export. If yours pushes products through the API, check it uses the Merchant API. Google’s Content API for Shopping reached sunset on 18 August 2026. Google says requests from clients without an active extension intermittently fail with HTTP 410 from 1 September, with full shutdown in early 2027.

ChatGPT shopping is live in the US, Shopify and Etsy catalogs are integrated already, and everyone else applies. OpenAI says feed onboarding is “currently available to approved partners” and that it’s “moving away from a standalone Instant Checkout experience” toward merchant-owned checkout. Copilot Checkout takes applications from PayPal and Stripe merchants. Google’s UCP checkout shows a Buy button only on products with the native_commerce attribute, and Google plans to roll it out in Canada and Australia, then the UK. Field-level detail is in our guide to product feeds for AI shopping.

Checklist for Magento Open Source, by impact

  1. Confirm AI crawlers reach product pages. Read each website’s robots.txt, your Cloudflare AI bot policy and any WAF rules, then check logs for 200 responses to OAI-SearchBot, PerplexityBot and Googlebot.
  2. Check the HTML before JavaScript. Fetch a product page with curl and look for name, price, stock status and key specs. Luma and Hyvä pass. PWA Studio needs server-side rendering or prerendering.
  3. Complete the Product markup. Add availability on Luma, or SKU and description on Hyvä, then brand, GTIN or MPN, itemCondition, shippingDetails and hasMerchantReturnPolicy. Use ProductGroup for configurable products, and keep every value identical to the page and the feed.
  4. Fix the feed. Submit the price any buyer pays, real GTINs and the availability the page shows, through the Merchant API or a scheduled file.
  5. Turn on both canonical settings and keep layered navigation parameters out of the crawl. If a filtered view deserves to rank, make it a real category.
  6. Enable cron-built XML sitemaps per store view, with images, and submit them in robots.txt.
  7. Run Varnish for full-page cache. Adobe strongly recommends it in production over the built-in file or database cache.
  8. Apply to the surfaces you sell into: ChatGPT merchants, Microsoft Merchant Center and, if you take PayPal or Stripe in the US, Copilot Checkout.

Checklist for Adobe Commerce, by impact

  1. Run steps 1 to 4 of the Open Source list. On cloud, read every custom VCL snippet and edge ACL for rules that catch AI crawlers.
  2. On the Edge Delivery storefront, deploy Commerce Prerender first and follow Adobe’s SEO metadata guidance so product metadata renders server-side. Without it, crawlers that don’t run JavaScript miss your product details.
  3. On Luma, apply the same canonical, markup and sitemap fixes as Open Source.
  4. If you license Brand Visibility, connect it to Commerce and ask your account manager for the catalog enrichment beta. Review each rewrite first, because enrichments write to the catalog and flow into your advertising and LLM feeds.
  5. Get Adobe’s UCP and ACP plans for your edition in writing: surfaces, countries, dates and the documentation page.

In an Agentic SEO audit we check these in this order, because markup can’t help a page crawlers can’t reach. To measure the visits and sales these fixes bring, use our GA4 guide.

What can an AI agent see in a B2B catalog?

Company accounts, shared catalogs, negotiable quotes and category permissions come with Adobe Commerce and its B2B extension. Magento Open Source’s core has none of these modules. A crawler, or an agent shopping for someone, arrives as a guest and sees what your NOT LOGGED IN group sees.

  • With shared catalogs on, guests see the public shared catalog, which Adobe says “by default, it is empty and does not contain any products.” If nobody assigned products to it, your catalog is invisible to every crawler.
  • Category permissions can hide prices from NOT LOGGED IN, so guests see products with no price.
  • Custom shared catalog prices show only to that company’s logged-in buyers.
  • Negotiable quotes start from the cart of an authorized company buyer or from a sales rep, so an agent acting for an unknown buyer can’t open one.

Without a price, an assistant can describe your part but can’t compare you, and Google’s merchant listings require a price above zero. Merchant Center’s price rules allow a sign-up only if it’s free and straightforward and can be completed on the landing page or during checkout, and a price that changes after sign-up belongs in loyalty_program. An account that waits for manual approval won’t meet that bar.

  1. Publish a public list price wherever you can, a price any buyer can pay, and keep contract pricing in custom shared catalogs. Feeds, markup and agents get a real number, and logged-in buyers still see theirs.
  2. Where you can’t, keep everything else public: specs, part numbers, compatibility, stock status, lead time and minimum order quantity. The Product markup still carries those facts, though the page won’t qualify for merchant listings.
  3. Give request a quote its own plain, crawlable page, linked from every product page. Say what you need and when you reply, and don’t require a company account, which means a separate form from the built-in B2B quote.
  4. Keep Allow Browsing Category open to NOT LOGGED IN and assign products to the public shared catalog.

What to leave alone for now

  • llms.txt and other AI-only files. Google says you don’t need new machine-readable files or AI text files to appear in AI Overviews or AI Mode. Fix the product HTML first.
  • Building your own UCP or ACP checkout on Open Source. UCP checkout is a US pilot, ChatGPT onboarding is for approved partners, and OpenAI is stepping back from standalone Instant Checkout. Apply now and build once you’re accepted.
  • AP2. There’s nothing for a merchant to configure yet.
  • Replatforming to Adobe Commerce as a Cloud Service for AI features alone. Commerce MCP is still coming soon, I found no UCP or ACP setup docs, and the Edge Delivery storefront needs prerendering before AI crawlers can read it.
  • Blanket AI crawler blocking on product pages. Decide on training crawlers separately. For the wider picture, see what AI assistants need from your store.

If you want help

Our Agentic SEO audit covers Magento and Adobe Commerce stores, takes two to three weeks and costs $1,500. The full audit fee comes off the first month if a retainer starts within 30 days. Book a 30-minute call.

Find out what assistants can read on your Magento store

The Agentic SEO audit covers Magento and Adobe Commerce stores: crawler access, rendering, Product markup, feeds and the surfaces you can apply to, in priority order. $1,500, two to three weeks. The fee comes off the first month if a retainer starts within 30 days.

Book a 30-minute callAgentic SEO

Sources

Facts checked on 5 October 2026.
Contact Us