New research guide: The Confidence Economy, how AI is changing fashion commerce.

Read it →

Read it →

/

/

Agentic Commerce, Shopify, News

Agentic Commerce, Shopify, News

·

October 29, 2026

Shopify Channel Mapping and Knowledge Base: A Fashion Brand Guide

Shopify's October 28 channel mapping and October 26 knowledge base change what AI agents see about your colours, variants and returns. What fashion brands should do this week.

Mehul Agarwal, Founder & CEO

Quick answer: On October 28, 2026, Shopify launched channel mapping, a page that lets you change the product data each sales channel receives (titles, colour, brand, and Shopify Catalog grouping) without editing your products. Two days earlier, on October 26, it moved the knowledge base, the store facts AI agents use for non-product questions, into the admin under Sales channels > Agentic > Knowledge base. For a fashion brand, these two pages now decide how AI shoppers see your colours, your variants and your returns policy.

Both changes come from Shopify’s changelog and Help Center. Here is what changed, why it matters more for apparel, and five things to do before Black Friday traffic arrives.

What changed on October 26 and 28

A shopper on a sofa chatting with an AI shopping assistant on her phone, with a lilac knit and a tote bag beside her

Knowledge base (October 26). Shopify’s Help Center describes the knowledge base as “the source of truth for AI agents for all non-product information about your store”. It is published through the Storefront MCP and can be used by AI channels including Shop and third-party platforms. Shopify scans your storefront and drafts facts in categories such as returns and warranty, gift services, fulfillment, payment, contact and second hand. Scanned facts stay unpublished until you approve them, and you can add your own facts as paragraphs, yes or no answers, or tags. The limit is 1,000 facts, and the standalone Knowledge Base app is retired on January 15, 2027.

Channel mapping (October 28). A new Channel mapping page shows, for each eligible sales channel, which of your product fields feeds each channel field, with a preview and a coverage percentage per field. You can point a channel field at a different source (a product field, variant field or metafield), combine up to 10 sources with simple Liquid, or override a value for one product on one channel. Shopify says most channels that use contextual product data are eligible, naming Google & YouTube and Facebook & Instagram, while Shop and Shopify Collective are not. For Shopify Catalog, which feeds Shop and Shopify’s agentic storefronts, you can map titles and descriptions, choose whether price or compare-at price is the list price, and group separate products as variants of one listing.

Why it matters more for fashion brands

A clothing rail holding the same wrap dress in five colourways: navy, plum, ivory, sage and rust

Apparel catalogs are messy in exactly the places these tools touch. Three common examples:

  • Colour names. “Midnight”, “Ink” and “Petrol” are good merchandising and bad data. If your Google colour field is fed from the variant name, an agent asked for a navy dress may not match it. Channel mapping lets you feed a standard colour metafield to Google while your storefront keeps the creative name.

  • One product per colourway. Many fashion stores list each colour as its own product for photography and SEO. Shopify Catalog’s custom variant grouping lets you group them back into one listing for AI channels, using a title delimiter, a metafield or a tag prefix, so an agent sees one dress in five colours instead of five unrelated dresses.

  • Returns and fit questions. Shoppers ask AI assistants “can I return a dress I wore once?” or “do you offer exchanges after Christmas?”. Those answers now come from your knowledge base. If your scanned returns facts are stale or unpublished, the agent either says nothing or says the wrong thing.

There is also a pricing point. Shopify notes that AI agents might use the list price to advertise strikethrough pricing, so the price versus compare-at choice in Catalog mapping is worth checking before your Black Friday markdowns go live.

What to do this week

Five tasks, roughly an afternoon for a mid-size catalog:

  • Review and publish your knowledge base. Go to Sales channels > Agentic > Knowledge base. Read every scanned fact in Returns & warranty, Fulfillment and Gift services against your live policy pages, fix anything wrong, then publish. Remember that scanned facts are not live until you publish them.

  • Add the facts fashion shoppers ask about. Use Additional facts for things a scan will miss: whether you offer exchanges for size, your holiday returns window, alterations or hemming, gift wrapping (a yes or no fact), and fabric or certification tags. Each fact can carry a reference URL, so link the matching policy page.

  • Open Channel mapping and sort by coverage. For Google, check colour, size, gender, age group and material. Anything below full coverage on a required field is a listing the channel may not show properly. Use the Fix products shortcut to bulk-fill gaps.

  • Map clean sources. Point the Google colour field at a standard colour metafield, add your brand name to titles if the channel benefits from it, and keep creative names on your storefront. Changing a metafield changes every channel that uses it, so check before you edit.

  • Group colourways for Shopify Catalog. If you split colours into separate products, set up custom variant grouping, and confirm which price Catalog shows as the list price before markdowns start.

Shopify warns that the preview shows what a channel receives, not whether it approves or displays your listing, and that changes can take time to appear. Make changes now rather than on November 26.

The part channel mapping cannot fix

Clean data gets your products found and answered correctly inside AI assistants. It does not help the shopper who lands on your product page and still asks, “what do I wear this with to a winter wedding?” That is where AI-referred visitors either build a complete look or bounce.

Elara is an AI stylist for Shopify fashion brands. It speaks first on your product pages, answers occasion questions, builds complete looks from your live catalog, offers full-outfit virtual try-on, and measures its impact with a holdout lift report, so you can see what it adds against shoppers who did not see it. Industry benchmarks put conversion up to 4x higher with AI shopping assistants (Neuwark 2026) and average order value 25% higher (Rep AI 2025); your own holdout result is the number that matters.

For the bigger picture, read our guides to Shopify Agentic Storefronts for fashion brands and how to get your fashion brand recommended by ChatGPT and AI Mode.

FAQ

What is Shopify channel mapping?
A page in the Shopify admin, launched October 28, 2026, that lets you choose which product fields, variant fields or metafields feed each sales channel’s fields, combine sources, and override values per product, without changing your product data.

Where is the Shopify knowledge base for AI agents?
In the Shopify admin under Sales channels > Agentic > Knowledge base. It holds up to 1,000 facts about your store that AI agents use for non-product questions such as returns, shipping and gift services.

Do I need to do anything if I used the Knowledge Base app?
If you had 1,000 or fewer facts, Shopify moved them over automatically. If you had more, add the ones you want to keep before the app is retired on January 15, 2027.

Does channel mapping work for Shop and ChatGPT?
Shopify says Shop and Shopify Collective are not eligible for channel mapping. For Shopify Catalog, which feeds Shop and Shopify’s agentic storefronts, you can map titles and descriptions, the list price and variant grouping, but not the product category.

Related guides

Want AI-referred shoppers to leave with a complete look? Book a demo or start Elara’s 30-day free trial.

Written by

Mehul Agarwal, Founder & CEO

October 29, 2026

ShareX ↗LinkedIn ↗
Thirty days on us

Give your store a stylist that actually sells.

Live in under an hour. Measured against a holdout from day one. Keep it only if the number moves.

No credit card · Cancel any time · No developer