Open your Shopify admin and go to Settings. There's a section for agentic storefronts, and Meta is now listed in it as an AI channel. Unless someone on your team has already been in there and switched it off, your products are being shared through Shopify Catalog by default. That's the short answer to the Shopify Meta AI channel question: it's on, it's a toggle, and the data going out is your existing product record, not a separate feed you built.
Whether you should leave it on is a different question, and for most US merchants the answer is yes — but only after you've looked at what's actually in your product data. Agents don't forgive a blank barcode field the way a human shopper forgives a thin description.
Where the agentic storefronts settings live
It sits under Settings in the admin, alongside the other AI surfaces Shopify has been adding. Each channel gets its own entry, and each entry has two separate controls worth understanding before you touch either:
- Whether your catalog is shared at all. This is the discovery switch. Off means the agent has no structured record of your products from Shopify.
- Whether the channel can complete a purchase. Shopify calls this direct checkout. Off means an agent can still describe and recommend your product, but the shopper lands on your PDP to buy.
Those are independent. You can be visible without handing over the transaction. Plenty of US brands we work with start there for a quarter, watch the order source data, then decide.
Shopify moves these settings around as the feature matures, so if the path in your admin looks different from what you read in a blog post three weeks old, trust the admin.
What Shopify Catalog actually sends
Nothing exotic. It's the product record you already maintain, which is exactly why messy catalogs get punished here. In practice what travels is the structured part of a product: title, description, product type, vendor, images, variant options, per-variant price and currency, inventory availability, SKU and barcode, the product URL, and the policies attached to your store — shipping and returns.
Shopify has not published a field-by-field schema I'd stake a season's revenue on, and the set has been expanding. So the useful way to think about it is this: if a field is a real field in your admin, assume it goes. If the information only exists as a sentence inside your description, or worse, as text baked into an image, assume it doesn't travel in a form an agent can filter on.
That last point is the whole game. A shopper searching an AI surface for "waterproof hiking boot, women's 8.5, under $180, ships in two days" is running four filters. Three of them are structured fields. If your size lives in a variant option you've named "Option 1" with values like "8.5 W", you're fine. If it lives in a size chart image, you're invisible for that query.
Meta Shop catalog vs Shopify Catalog: two different pipes
These get conflated constantly, and the confusion costs people a week of debugging.
The Meta catalog — the one in Commerce Manager, fed by the Facebook & Instagram sales channel — exists to serve ads and Shops. It's the thing powering Advantage+ catalog ads and dynamic retargeting. You've probably had it running for years. It has its own feed rules, its own diagnostics tab, its own habit of rejecting products for policy reasons.
The Shopify Catalog share under agentic storefronts is a different pipe to a different consumer. It's product data made available for AI shopping experiences, sent by Shopify, governed by a Shopify-side toggle. Turning off the Facebook & Instagram sales channel does not turn this off. Deleting a product set in Commerce Manager does not change it.
If your Meta ad catalog is clean and healthy, that's a decent signal your data will hold up here too. It isn't a guarantee, because the ad catalog tolerates gaps that an agent's filter does not.
Direct checkout on AI channels: what you keep
When direct checkout is enabled, the purchase completes inside the agent surface and the order lands in your Shopify admin like any other order. Shopify checkout still does the work underneath, which matters more in the US than anywhere else: sales tax is calculated against your nexus settings, whether you're on Shopify Tax or piping through Avalara. Enabling an AI channel doesn't create nexus and doesn't change your filing obligations. Your shipping rates, your carrier accounts, your USPS and UPS labels, unchanged.
What you give up is the first-touch experience. No upsell block, no bundle offer at cart, no post-purchase page. If 11% of your revenue comes from a post-purchase one-click offer, model that loss before you flip it on.
Customer data depends on consent in the flow, and the honest answer is that it's thinner than a direct order. Treat agent-sourced buyers as a list you may not get to email twice. Given how much US D2C has leaned on retention and SMS as paid acquisition got expensive, that's a real trade-off, not a footnote.
One thing that breaks: anything that isn't a simple variant purchase. Subscription-only SKUs, build-your-own configurators, made-to-order with a text field, products gated behind a B2B price list. In testing, agent checkout handles "pick a variant, buy it" well and very little beyond that. If that's most of your catalog, direct checkout is noise for you right now.
The arithmetic on leaving it on
Run your own numbers, but here's the shape of it. Say your AOV is $95 and your gross margin is 62%, so $58.90 of gross profit per order. Blended paid CAC on Meta at $38 leaves you $20.90. On an agent-sourced order with no ad spend against it, you're paying Shopify Payments at 2.9% + $0.30, which on $95 is $2.76 + $0.30 = $3.06. Even after a channel referral cost and higher-than-normal return rates — and returns will be higher where the shopper never saw your PDP — the contribution per order is multiples of the paid-social equivalent.
That's the argument for leaving it on. Not that AI shopping is a large channel yet. It isn't. It's that the incremental cost of being present is a settings toggle and some data hygiene, while the incremental cost of a Meta customer keeps climbing.
Product data requirements for AI shopping agents
Here's what we fix first on a US catalog, roughly in order of payoff:
- Barcodes. GTIN, UPC or EAN on every variant. This is how an agent knows your Vibram-soled boot is the same object another retailer is listing. Missing barcodes are the single most common gap we find, and on apparel catalogs it's often most of the SKUs.
- Variant options named like options. "Size", "Color", "Material" — not "Option 1". Values consistent across the catalog: pick "Navy" or "Midnight Blue" and stop switching.
- Attributes in metafields, not prose. Fit, material composition, dimensions, weight, care, compatibility, certification. If a customer service rep gets asked it twice a week, it belongs in a field.
- Accurate inventory and a real lead time. An agent that recommends an out-of-stock product once will deprioritize you. Shopify tracks availability per variant; make sure your 3PL sync isn't lagging by six hours.
- A specific return policy. "Contact us for returns" reads to a buying agent as risk. Days, who pays shipping, condition requirements, stated plainly.
- Descriptions written for retrieval. First 150 characters carry the product type, the defining attribute and the use case. Keyword stuffing does nothing here; omission does damage.
Most of this is the same work that makes on-site filtering and search good, which is a nice coincidence — the attribute metafields that let a shopper narrow a 2,000-SKU collection to eleven products are the fields an agent uses to answer a question. If your collection pages still only filter by price and availability, that's the tell. FilterPro is how we usually get structured filters and search live on a large catalog without a theme rebuild, and the metafield discipline it forces pays off on both sides.
There's a second layer that has nothing to do with Catalog: whether AI crawlers can read your storefront at all. Check robots.txt, check that your PDPs carry valid Product schema with price, availability and GTIN, and check that your critical copy isn't rendered client-side in a way a crawler drops. Catalog gets you the structured record. Crawl access gets you everything else on the page. We cover the markup side in our SEO and content work, because it's the same job.
How to opt out, and when you actually should
Opt-out is the same screen. Turn off catalog sharing for the channel and the structured product data stops flowing. If you want to stay discoverable but keep the transaction, leave sharing on and disable direct checkout.
Per-product control behaves like channel publishing — the same mechanics you use to keep a product off Google Shopping. Worth testing with one SKU before you assume it works the way you want across a 5,000-product catalog.
The cases where switching it off is the right call:
- You enforce MAP or selective distribution and can't have an agent presenting your price next to a grey-market listing.
- Regulated categories where the compliance review hasn't happened. Supplements with structure-function claims, anything with age gating, firearms accessories.
- Your catalog is wholesale-priced or B2B-gated and a public price would undercut your own dealers.
- Your product data is genuinely bad and you'd rather be absent than be recommended wrong. Fair position. Fix the data and come back.
What switching it off costs you is discovery you can't buy back. An agent building a shortlist from structured retailer data doesn't fall back to your brand because it likes you. If your products aren't in the set, you aren't in the comparison, and the three competitors who did the barcode work are. For a brand with little unaided demand — most D2C, honestly — that's the whole pitch for being there.
Questions we keep getting
Does enabling this duplicate my Meta ad catalog? No. They're separate systems. Products can appear in both, and a rejection in Commerce Manager doesn't carry across.
Will agent orders mess up my attribution? They'll show as their own source in Shopify analytics. Where it gets murky is a shopper who discovers you through an agent, then buys direct a week later — same problem you already have with every channel. Don't rebuild your MTA model for a channel that's single-digit percent of orders.
Do I need Shopify Plus for this? No. The agentic storefront settings are in the standard admin. Plus matters for checkout extensibility, B2B and volume, which is a separate conversation entirely.
Should I turn it on before Black Friday? Turn on catalog sharing now and spend the time between now and November on data. Direct checkout, we'd hold until after BFCM — you don't want an unfamiliar order path during the one week your ops team has no slack.
Where to start this week
Export your products as CSV and count the rows with an empty barcode column. That number, as a percentage of your SKUs, is your readiness score for every AI shopping surface, not just this one. Under 10% and you're in decent shape. Over half and the toggle doesn't matter yet.
If you want a second pair of eyes on the catalog before BFCM, we do a free store audit that covers product data structure alongside the usual speed and conversion work. US merchants can also see how we work from our US engagements page.