Assistants Are Answering Fit Questions Directly — and Image-Only Size Charts Are Invisible to Them

Assistants have started answering fit and sizing questions directly. Ask a shopping assistant whether a jacket runs large, and it increasingly answers rather than deferring to the product page gallery. That is a meaningful shift for apparel and footwear merchants, because fit is the one product attribute buyers cannot evaluate from a photo — and it is the attribute most often missing from structured product data.

The published guidance coming out of the major platforms is consistent: buyers ask assistants comparative questions ("does this run small," "will this fit someone 180cm"), and the assistants answer from whatever text they can find that speaks to the issue. Where a merchant publishes a real size chart, a measurement table in both centimetres and inches, and model measurements, the assistant has material. Where the merchant publishes photos only, the assistant either declines or repeats marketing adjectives that answer nothing.

There is a second layer to this, and it is where most catalogs fall short. Fit information that exists only as an image — a sizing chart as a JPEG — is invisible to retrieval systems. Assistants do not read pixels. A size chart rendered as a graphic is functionally absent, no matter how clearly a human can read it.

## The measurement gap

Three problems recur across cross-border apparel catalogs:

Unit ambiguity. A chart that says "38" without stating centimetres or inches is not usable. Cross-border catalogs frequently mix units across product lines, or inherit a chart from a supplier in a different system.

Measurement target ambiguity. Chest, waist, and length are measured differently by different brands, and "size M" is not a standard. An assistant comparing two merchants cannot reconcile them without a stated measurement basis.

Return-policy disconnect. Sizing and returns are the same conversation for a buyer. A store that publishes detailed measurements but describes returns vaguely gives the assistant conflicting signals about how much risk to attach to the purchase.

## What this means for cross-border merchants

The practical work is unglamorous and clear. Publish fit data as text, not only as images. State units explicitly for every dimension. State what each measurement refers to. Keep the sizing table and the returns policy describing the same purchase experience. Where a supplier provides measurements, normalize them before publishing rather than passing them through.

This is a case where the usual temptation — add more schema and hope — is not sufficient. A Product record whose size attribute is just the letter M carries no fit information. The text on the page is what answers the question, and it is what gets quoted.

Visora's read on this: fit is becoming a citation battleground because it sits at the intersection of what buyers ask and what merchants fail to publish. A free scan at /audit flags whether your sizing content is machine-readable or trapped in images, and whether your returns wording agrees with your measurement claims. Measurement semantics are covered at /faq.