How Appifire Uses Shopify Product Information to Answer Questions

You will understand which Shopify product information Appifire AI Chat uses to answer storefront questions. You will know what is synced, what “published only” means, how answers are built, and how to keep knowledge current.

When you finish, you can check your catalog against the fields Appifire actually reads, then improve thin products on purpose.

Direct answer

Appifire answers product questions from synced, published, active Shopify products. It turns product text into searchable knowledge snippets, finds the closest snippets for the shopper’s question, then writes a reply from that store context. It does not invent materials, fit, or prices that never appear in synced catalog text.

Other knowledge (collections, blog articles, website knowledge, FAQ) can also help. This guide focuses on products. For fixing weak copy, see How to Improve Product and Order Answers in Appifire AI Chat and How to Prepare Shopify Product Data for Accurate AI Answers.

Prerequisites

• Appifire installed with at least one product sync completed (Getting Started)
• Products published to your online store (not draft-only)
• Ability to open a product in Shopify Admin and the live PDP
• Optional: Appifire Data Sync access to re-run product updates after bulk edits

How product answers work (current behavior)

1) Shopify products sync into Appifire

On install, Appifire starts a product sync. Later, product create / update / delete webhooks keep many changes current. You can also refresh products from Data Sync (update products / full refresh paths available in admin).

For each published active product, Appifire stores fields such as:

Field	Used for answers?	Notes
Title	Yes	Retrieval + product card matching
Description (HTML stripped to text)	Yes	Main place for specs, fit, care
Product type	Yes	Category grounding
Vendor	Yes	Brand questions
Tags	Yes	Discovery signal
Status + published time	Yes (gate)	Draft / unpublished / archived stay out
Product page URL	Yes	Link context in knowledge
Featured image URL	Cards	Thumbnail on product cards, not the answer essay
Variants (up to 50 per product)	Yes	Variant title, price (store currency), SKU, inventory quantity as “In stock” lines

Not in this product pipeline today: product metafields are not part of the implemented product GraphQL sync/chunk path. If a fact only lives in a metafield, put it in the description (or another ingested source) if you want chat to use it.

2) Only published, active products enter chat knowledge

Appifire syncs with an active + published filter. Draft, unpublished, and archived products are removed from searchable knowledge. Zero inventory does not block a published product from knowledge.

That keeps chat aligned with what shoppers can buy on the storefront.

3) Product text becomes searchable snippets

Appifire splits product content into smaller knowledge chunks (overview + variant details). Long descriptions are split so search can find the right section. Those chunks are embedded for similarity search per shop.

Product (re)embedding uses the same AI credit pool as other AI work (logged as product knowledge usage).

4) A shopper question retrieves the closest snippets

When someone chats:

1. Appifire embeds the question (with short recent chat context when useful).
2. It finds the closest knowledge chunks for that shop (top matches; default top 5).
3. The language model answers using that store context (and optional focus / PDP boosts below).
4. If there are no knowledge chunks and no order context, Appifire returns a fixed “not enough information” style message instead of guessing.

There is no minimum similarity floor that drops weak matches. If your shop has knowledge, low-relevance chunks can still be retrieved. Good titles and clear descriptions reduce that risk.

5) Product cards use exact catalog titles

When the assistant names products in a structured way, the widget can show up to 6 product cards (image, title, price, View Product, Get More Info).

Card matching needs the full exact catalog title in the reply. Partial or fuzzy titles may not show a card. Get More Info focuses the next turn on that product’s knowledge.

6) On a product page, context can boost that product

On a PDP, the widget can send page product context (Shopify product id, handle, title, selected variant). Appifire boosts that product’s chunks for on-page questions. A Get More Info focus product overrides page product for that turn.

Steps: verify your store matches this pipeline

1. In Shopify Admin, open a best seller. Confirm it is Active and published to the online store.
2. Confirm title, description text, and variant titles/prices look correct.
3. In Appifire, confirm products have synced (re-run product sync from Data Sync after big edits if needed).
4. On the live PDP, ask a question whose answer is clearly in the description.
5. Compare the chat reply to the Admin fields. If chat is wrong and the page is right, re-sync and retest.
6. Ask the assistant to recommend a product using its exact title. Confirm a product card appears when expected.
7. Ask a question that only exists in a metafield (if you use metafields). Confirm whether chat knows it. If not, copy the fact into the description and re-sync.

What success looks like

• Published products answer from description / variant facts that exist in Shopify
• Draft products do not show up as sellable recommendations
• PDP questions about the current item stay on that product more reliably
• Exact-title mentions can show product cards
• After a catalog fix + sync, the same test question improves

Limits and non-goals

• Appifire does not read product metafields in the current product sync/chunk path
• Sync includes up to 50 variants per product
• Chat messages are capped (for example 2000 characters on the API)
• Inventory lines in answers come from variant quantities on product chunks; they go stale until the product is re-synced / updated
• A separate “stock levels” Data Sync resource (if shown in admin) is not the same as inventing live stock from thin air in every reply
• Collections, blogs, and Knowledge Hub website/FAQ content are separate sources (helpful, but not covered in depth here)
• Order status uses live Shopify order lookup, not the product catalog
• Appifire does not add to cart or complete checkout from product cards (View Product opens the PDP)

Why this matters (short)

Chat quality is mostly catalog quality. If the description is empty, retrieval has nothing true to say. Knowing the field list stops you from blaming “the AI” for missing Shopify text.

Related Appifire Guides

• Getting Started with Appifire AI Chat
• How to Improve Product and Order Answers in Appifire AI Chat
• How to Prepare Shopify Product Data for Accurate AI Answers
• Ecommerce Product Data Audit: Is Your Shopify Catalog Ready for AI?
• How to Customize the Appifire Chat Widget
• Appifire Installation & Setup Checklist

Last verified

2026-07-23 against Appifire product docs for product sync & ingestion, published-product retrieval, storefront product cards / PDP context, and related RAG notes. Re-check Data Sync labels in the live admin after releases.