Is Your Furniture Catalog Ready for AI Shopping?

A shopper can now ask for a washable corner sofa that fits a particular wall, compare the options and narrow the list without opening 10 product pages.

Your product can only enter that comparison if the service can reach it and understand the facts that matter. For furniture, that means more than a product name and a price. Dimensions, orientation, material, lead time and the exact variant all affect whether an answer is useful.

The practical work is familiar: complete the catalog, make product pages crawlable, keep feeds current and expose important facts as text. AI shopping has made that work more valuable. It hasn't replaced it with a secret new file.

What information do AI shopping services use?

The mix varies by platform, but two sources matter most:

1. Product pages: the visible content and structured data on your website
2. Merchant or platform feeds: structured product information supplied directly or through an ecommerce platform

OpenAI says ChatGPT shopping uses structured metadata from first- and third-party providers. Merchants can apply to provide a direct product feed, while Shopify product data is integrated through Shopify Catalog.[1]

Google says the normal foundations of search still apply to AI Overviews and AI Mode. Product pages need to be indexable, important information should exist as text, structured data should match the visible page and Merchant Center information should be current.[2]

Pages and feeds support one another. A feed carries standardized fields and frequent changes. A product page can explain furniture-specific details that don't fit neatly into every destination's schema.

The five catalog checks

1. Can the service reach the product?

Check that:

• Product pages are available without a login.
• Important content is present in the page that crawlers receive.
• Canonical URLs and internal links lead to the products you want discovered.
• Your robots, CDN and bot-management settings reflect the traffic you intend to allow.
• Old or discontinued variants don't remain as apparently purchasable pages.

Google states that there are no special technical requirements for appearing in its AI search features beyond the usual search requirements.[2] An llms.txt file is not a substitute for crawlable product pages, Merchant Center data or valid structured data.

Cloudflare users should review their settings rather than rely on a default. Cloudflare says that from September 15, 2026, new domains will default to blocking bots classified as Training or Agent on pages with ads, while Search remains allowed.[3] That date and behavior are platform-specific, so confirm the live setting in your own account.

2. Can it identify the exact product and variant?

Every sellable option needs a stable identity. Keep:

• Product and variant SKUs
• Product-family or group ID
• GTIN where valid and available
• Color, material, finish, size, orientation and configuration as separate values where they affect the offer
• Variant-level price, availability, lead time and imagery

Google's product-variant guidance supports ProductGroup, hasVariant, productGroupID and variesBy. It allows both single-page and multi-page implementations. In a single-page setup, a URL can load the relevant option preselected.[4]

That detail matters for configurable furniture. If the same URL always opens the default sofa, a service may not be able to distinguish the blue left-hand corner unit from the cream three-seater a shopper requested.

3. Can it answer the furniture question?

Put the decisive facts in visible text, even when they also appear in a diagram or filter.

For a sofa, that could include:

• Overall width, depth and height
• Seat height and depth
• Orientation and configuration
• Cover material and care
• Assembly or access requirements
• Lead time and delivery method
• What is included in the stated price

Google Merchant Center supports typed fields including product length, width, height, weight and material.[5] Use the fields a destination provides, but don't assume every shopping service accepts the same model. The furniture product data model covers the underlying record those fields are drawn from.

A dimension diagram helps the shopper. It shouldn't be the only place the measurements exist.

4. Do the page, feed and imagery agree?

Conflicting facts create a bad answer before they create an SEO problem.

Compare several high-value and highly configurable products across:

• The catalog record
• The live product page
• Structured data
• Merchant feeds
• Marketplace listings
• Product and variant images

The selected colorway should show the right product. The price should belong to that SKU. The lead time should not change between the feed and the page.

Generated imagery needs human approval. Review silhouette, proportions, construction, material, finish, color, orientation and configuration against the reference. Once approved, connecting the image to an exact SKU or colorway requires confirmation in Furniture Connect.

5. Can you see what is working?

Choose a small set of commercial questions your customers actually ask, then review the results regularly. Examples might include a sofa that fits a specific width, a dining table for a named number of seats or an outdoor set in a particular material.

Record:

• Whether your eligible products appear
• Whether the named product and variant are correct
• Which source pages are cited or linked
• Whether price and availability are current
• Referral traffic and resulting sales where your analytics can identify them

Don't treat one prompt as a ranking report. Results can change with the wording, user context, location and product availability. The useful signal is a repeated pattern across commercially important questions.

Discovery and checkout are different decisions

Being understood by an AI shopping service does not require you to support an agent-completed purchase.

Google's Universal Commerce Protocol is an open standard for agentic commerce, and Google's current implementation begins with direct buying through AI surfaces. Access includes an application process, and the merchant remains the merchant of record.[6]

For furniture, decide whether the purchase should finish in the assistant or move to your own site. Delivery access, made-to-order options, swatch decisions and high return costs can make a referral to your site the better experience for some ranges.

Fix discovery first. A checkout protocol cannot compensate for a catalog that has no dependable dimensions or variant identity.

What should a furniture team fix first?

Use this order:

1. Confirm that product pages can be crawled and indexed.
2. Give each sellable variant a stable identity and retrievable state.
3. Complete dimensions, materials, lead times, delivery and returns.
4. Make visible page text, structured data and feeds agree.
5. Add accurate imagery for the variants shoppers can choose.
6. Measure a defined set of shopping questions and correct the gaps.

This is the same upstream catalog work needed for ecommerce and retailer distribution. The furniture ecommerce readiness checklist covers the broader go-live review, and how furniture product content reaches retailers covers the routes to every other destination.

Furniture Connect helps teams structure supplier data, identify missing catalog content and build variant imagery in batches.[7][8] It does not currently promise a direct feed into every AI shopping service, automatic image approval or automatic channel-rule enforcement.

The honest value is upstream: more products with enough trusted data and imagery to be represented wherever customers compare furniture next.

Frequently asked questions

Sources

1. OpenAI Help Center: Shopping with ChatGPT Search
2. Google Search Central: AI features and your website
3. Cloudflare: Block AI bots
4. Google Search Central: Product variant structured data
5. Google Merchant API: Product attributes
6. Google: Universal Commerce Protocol
7. Furniture Connect Help Center: AI enrichment
8. Furniture Connect Help Center: Batch product and material imagery

---

If AI shopping is on the roadmap, start with one difficult range. In a Furniture Connect demo, we'll work through the facts and variant images it needs before any destination can represent it properly.