Where Furniture Product Data Hides: PDFs, Spec Images, Tabs and Dropdowns
Furniture facts are rarely missing. They're in a PDF, a spec image, a tab fetched on click or a dropdown. A tour of each hiding place and the fix that keeps the design.
The surprising thing about most furniture product pages is how much they know. The width is there. The fabric is there. The lead time is there. The problem is where. Each fact has found a home that suits the person who put it there and is harder for any reader that isn't a person with a mouse and patience.
Here are the usual hiding places, in roughly the order you'll find them on a page.
The PDF spec sheet
The supplier sends a PDF. It has everything: dimensions per size, materials, construction, weights, care, sometimes test results. The retailer links to it under "Download specifications" and considers the facts published.
For a machine, a PDF is a separate document, often a scan or a designed layout with tables that don't parse cleanly, covering a whole range rather than this variant. Search engines can index PDFs and some assistants will follow them; parsing the table and attaching the right row to the right variant is where it gets unreliable, so don't depend on it as the only copy. And it's the supplier's document, so it says "available in 12 fabrics" when you stock four.
Why it happens: it's one upload and it's authoritative. The fix: extract the facts once into attributes on the record, per variant. Keep the PDF as a download for the shopper who wants it.
The spec image
Dimensions as a diagram; features as an infographic; "what's in the box" as a picture. Covered in detail in dimensions in the image. The short version: pixels are harder to read than text. The fix: every number in the image also in a text table and on the record; alt text that carries the facts.
The tab that loads on click
"Description", "Specifications", "Delivery", "Reviews" as tabs. Two ways to build these. One puts all the content in the page and hides the inactive tabs with styling; a machine reading the page gets everything. The other fetches each tab's content from the server when it's clicked; a system that doesn't run scripts gets the first tab and nothing else.
You can't tell by looking, and a tab appearing on click doesn't mean its content was fetched on click. Ask your developer, or view the page source and search for a phrase that's in the second tab. If it isn't there, it's only available to systems that render the page's scripts. Google does [1]; others vary, so deliver it with the page rather than relying on rendering.
Why it happens: the second pattern is a common default in some themes and feels faster. The fix: have tab content delivered with the page. Usually a small change.
The variant dropdown
Size: 72 / 84 / 96. Fabric: twelve swatches. Leg finish: three. Pick a combination and the price, the dimensions, the images and sometimes the lead time update by script.
A machine reading the page sees the default combination. If the default is the 84-inch in Oat, the assistant may report your sofa as 84 inches wide when the shopper asked for something under 80 and you make a 72. Variant-level facts that only appear after selection are harder for a machine to attach to the right variant, and the default's values can end up standing for the product.
Why it happens: it's how configurable products are presented, and it's good for the shopper who's clicking. The fix: per-variant facts in the product feed and structured data, and a plain table on the page listing each size with its dimensions and price. For deeper problems with configurable ranges, the existing piece on why AI shopping assistants struggle with configurable furniture covers them.
The shipping calculator
"Enter your ZIP code for delivery options." Accurate, personalized and invisible until someone types. An assistant asked "ships in under four weeks" gets nothing. The fix: a plain statement per product ("Ships in 2 to 3 weeks" or "In stock, delivers in 5 to 10 business days") on the page, in addition to the calculator.
The popup and the policy page
Returns, warranty, assembly, white-glove delivery: often in a modal or on a separate policy page. Fine for the shopper who looks. For a machine comparing products, "does it come assembled" needs to be on the product. The fix: the product-level answer as a line on the product page and, where it varies by product, as an attribute.
The category word
Not a hiding place on the page, but a hiding place in the language. "Performance fabric", "solid wood", "premium foam", "fast shipping". Each is a category standing in for a fact. The fact exists (100% polyester at 50,000 double rubs; kiln-dried hardwood; 2.0-density high-resilience foam; 2 to 3 weeks) and was replaced by something that sounds nicer and says less. The fix: keep the nice word if you like; add the fact next to it.
The reviews
Oddly, the one place facts are often readable and present. "Fits through our 30-inch door with the legs off." "Deeper seat than it looks." Machines read reviews. The problem is that these are facts the product should state, found by shoppers after purchase. The fix: mine them for questions, verify the answers against your own data, then put the verified facts into the product data.
The pattern behind all of them
Every hiding place is a reasonable decision by one person for one reader. The supplier's PDF is authoritative, the diagram is clear, the tabs are tidy, the dropdown is how configurators work, the calculator is accurate. Each one is locally right and collectively they produce a page that knows everything and tells a machine much less than it knows.
The fix is also the same each time: the fact lives on the product record as an attribute, per variant, and the page, the feed and the structured data read it from there once each is mapped to it. The PDF, the diagram, the tabs and the calculator can all stay. They just stop being the only copy.
That's a catalog decision, not a page decision, which is why why fixing one product page doesn't fix the range comes next. For the full argument, see being found by AI.
A fifteen-minute audit
Open your top product. Find the width, the fabric composition, the lead time and the price for the smallest size. For each, write down where you found it: text on page, image, PDF, tab, after a dropdown, after a ZIP code. Anything that isn't "text on page" is on this list.
