Write every temperature rating as a signed number with its scale, such as -30 °C (-22 °F). Then say which part of the garment it covers, the activity and layering it assumes, and who set it. For fit, give the body measurements each size is cut for, not only an age label.
When we searched Shopify’s agent catalog for “toddler snowsuit rated for minus 30”, two of the ten listings stated -30 °C. One of them applied the rating only to the hood’s insulation, and Shopify’s generated spec list for that record dropped the word hood. A missing scope or condition is an unknown you can fix. It isn’t proof that a garment is warm, or that it isn’t.
The ratings weren’t missing, either. Our own first text matcher reported both listings as having no rating, because it didn’t know “minus 30” and “-30 °C” are the same value. That was our mistake, and we’ve written up how temperature matching goes wrong. This guide is about the merchant side: what to publish so a rating survives being read, copied and compared.
What ten snowsuit listings said about warmth
Here’s every temperature or warmth statement in the ten results, with the field it appeared in. Store and brand names are withheld.
| Listing | Where | What it said | What’s missing |
|---|---|---|---|
| Position 2 | Title | “protects to -30 °C” | Scope, activity, who rated it |
| Position 2 | Description | “Temperatures ca. 0 - -30 °C” | Which end applies to which activity |
| Position 2 | Description | “generally being outside in temperatures around -10°C to -30°C”, in a list of activities | Ties temperature to activity; still no rater named |
| Position 9 | Description | “a hood with [brand] insulation rated to -30°C” | Rating for the rest of the suit; who rated it |
| Position 9 | Shopify-generated spec list | “Insulation: [brand] rated to -30°C” | The word hood |
| Position 1 | Description | “protects from even the harshest frost”, plus a test-jury quote | Any number |
| Position 1 | Shopify-generated selling point | “proven to keep children warm and dry in extreme winter weather” | “Proven” isn’t in the description |
| Positions 6 and 8 | Description | “Very warm”; position 6 “extra-snug padding guarantees warmth”; position 8 “extra-snug insulation guarantees warmth” | Any number |
| Positions 3, 4, 5, 7, 10 | Description | Warm in words (an insulated seat, “warm and dry”), or nothing about warmth | Any number |
Counted by field:
| Field | Listings with a temperature figure |
|---|---|
| Title | 1 of 10 (position 2) |
| Description | 2 of 10 (positions 2 and 9) |
| Shopify-generated metadata | 1 of 10 (position 9, without the hood scope) |
The other eight descriptions described warmth in words only, or not at all. Only position 2 linked temperatures to what the child would be doing. None of the ten descriptions said what the child should wear underneath.
Three statements, not one range
Position 2 is the most complete listing, and it shows why a rating needs its conditions spelled out. Its title, its spec block and its activity guidance make three different statements:
- “protects to -30 °C” (title)
- “Temperatures ca. 0 - -30 °C” (spec block in the description)
- “around -10°C to -30°C” for activities such as sleigh rides, sledging and sitting in a pushchair (description)
The description also explains that a small child may spend long stretches fairly inactive, which is exactly when a rating matters most. Blending these into “rated -30 °C” or “0 to -30 °C” throws that information away. Keep them as separate rows, each with its own condition.
The hood that became the suit
Position 9 states its rating carefully: the hood’s insulation is rated to -30 °C. Shopify’s generated spec line turned that into “Insulation: [brand] rated to -30°C”, and a generated feature line added that it “keeps kids safe from the elements”. A shopper asking for a suit rated to -30 °C could now match a garment whose only published rating covers the hood. The merchant didn’t overclaim. The copy did.
The same trap outside outerwear: a dropped minus sign
Temperature data breaks in ways that are easy to miss. In a search for “ice cleats for running shoes on icy sidewalks”, one description said the product “can be worn in temperatures as low as -41 Fahrenheit”. Shopify’s generated feature list said “Suitable for temperatures as low as 41°F”, and its generated spec list said “Temperature Range: Down to 41°F”. Both generated fields lost the minus sign. 41 °F is above freezing, so the copied rating now says almost the opposite of what the merchant wrote.
We can’t see how Shopify generated those fields. The lesson for merchants is the same either way: write the rating so that each part survives copying, and check what was received.
The temperature checklist
Run every temperature statement on your product through these eight checks before publishing. The checklist is proposed: it’s our working list, not a published standard.
| Check | Write it like this | Not like this | Why |
|---|---|---|---|
| 1. Sign | “-30 °C” with the minus attached | “30 below”, “minus thirty” alone | A dropped or spelled-out sign is the easiest error to make and the costliest |
| 2. Scale | “-30 °C (-22 °F)” | “-30”, “-41 Fahrenheit” with no symbol | Give both scales so nobody converts for you |
| 3. Scope | “Whole suit”, “jacket only”, “hood insulation” | “Rated to -30 °C” | A part’s rating isn’t the garment’s rating |
| 4. Activity | “Active play”, “standing”, “sitting in a stroller” | No activity | The same suit keeps an active child warmer than a still one |
| 5. Layering | “With a base layer and fleece mid-layer” | No layering | A rating that assumes layers can mislead a shopper who buys only the suit |
| 6. Kind of number | “Maker’s guidance”, “insulation maker’s rating”, “independent test, [name, year]” | A bare number | Who set it changes how much weight it can carry |
| 7. Ranges | One row per statement, each with its own condition | Merging separate figures into one range | Each figure was given under different conditions |
| 8. Consistency | Identical wording in title, description, spec field and metafields | Different numbers or scopes in different places | Every copy is a chance for the scope to drop |
Check 6 is where most listings were silent. Neither -30 °C listing said who set the figure. If the number comes from the insulation maker, say so. If it’s your own guidance for typical use, say that instead. Both are honest; an unattributed number isn’t something anyone can check.
Conversions to use: -10 °C is 14 °F, -20 °C is -4 °F, -30 °C is -22 °F, and -40 is the same on both scales.
Fit: sizes need measurements, not just labels
Fit was the second gap. All ten snowsuit listings offered a size option, but they used five different kinds of label (month ranges, T sizes, year sizes, “years” and bare numbers), in four combinations:
| Labelling style | Example | Listings |
|---|---|---|
| Month ranges plus T sizes | 9-12M, 2T, 3T | Positions 1, 4, 10 |
| T sizes plus year sizes | 2T, 4T, 5Y, 10Y | Positions 3, 5, 6, 7, 8 |
| Years | 1.5 years, 2 years, 3 years | Position 2 |
| Bare numbers | 4, 5, 6 … 16 | Position 9 |
None of the ten descriptions gave the body height or chest measurement a size is meant for. Product pages may have size charts that the catalog response didn’t include, but an agent reading the catalog record can’t use a chart it never received.
Generated lists didn’t always agree with the size options, either. Position 9’s options run from 4 to 16. Its generated spec list said “Available sizes: 5-6 years, 7-8 years, 6 US, 16 US”.
The search also asked for a toddler snowsuit. Four titles said toddler, five said kids’, and one said neither and started at size 4. Age range is a fit fact, and it belongs in the record.
OpenAI says ChatGPT’s virtual try-on images don’t guarantee fit or size, and tells shoppers to check the merchant’s measurements, product details and return policy before buying. If the measurements aren’t published, there’s nothing to check.
What a fit record should hold:
| Fact | Publish as | Example |
|---|---|---|
| Size label | Your label, and the system it uses | 2T (US toddler) |
| Body measurements | Height and chest range per size, in cm and in | [height range] cm / [height range] in |
| Cut | Straight, regular, roomy | Straight fit |
| Room for layers | Whether the size allows a fleece underneath | [Yes / size up for thick layers] |
| Growth features | What adjusts, and by how much | Sleeves extend 4 cm; hems adjust 2 levels |
| Age range | The youngest and oldest size | 9 months to 3 years |
Position 9 is the one listing that put a number on growth: sleeves that “grow 4 cm” and hems with two adjustment levels. That’s the kind of fit fact a parent can act on. The product specification template has the same structure for any category.
Worked example: rebuilding the hood-rated record
Here’s position 9’s published warmth and fit data as a merchant record. Brackets mark facts the merchant would take from their own documents; we don’t know them, and you shouldn’t invent them.
| Fact | Value and scope | Condition | Source and status |
|---|---|---|---|
| Hood insulation | -30 °C (-22 °F), hood only | [activity and layering, if stated by the rater] | [insulation maker or garment maker]; stated in description |
| Whole-suit rating | Unknown | — | Not published; say “not rated” or publish the real figure |
| Jacket lining | Polar fleece | — | Stated in description |
| Sleeve growth | +4 cm | — | Stated in description |
| Hem adjustment | 2 levels | Snap buttons | Stated in description |
| Sizes | 4 to 16 | [body height per size] | Option list; measurements not published |
Then compare the record with what the catalog carries. For this listing the comparison flags two fixes: the generated spec line that dropped “hood”, and the generated size list that doesn’t match the size options. Shopify’s Agentic section in the admin lets you review how product details are mapped to its catalog and preview how products appear in Catalog search. Start there, and put the scoped wording into a structured field you control as well as the description.
Check your record against the questions shoppers ask
For each question, your published record should hold the evidence in the second column, avoid the trap in the third, and say plainly when the garment doesn’t fit the need.
| Shopper question | Evidence your record must hold | Unit or condition trap | Rejection case to publish |
|---|---|---|---|
| “Is it warm enough at minus 30?” | Signed temperature in °C and °F, scope, rater | Sign dropped; °F read as °C; a hood rating read as a suit rating | “Whole suit not rated below [temperature]”, if true |
| “Will my toddler be warm sitting in a stroller?” | The activity the rating assumes | A rating for active play applied to a still child | “Rating assumes active play”, if that’s what it assumes |
| “What does she need underneath?” | The layering the rating assumes | No layering stated at all | “Rating assumes [layers]; without them, expect less warmth” |
| “Will it fit a child who’s 92 cm tall?” | Height range per size, in cm and in | Age labels in different systems; generated size lists that differ from the options | “Runs small; size up for thick mid-layers”, if true |
| “Is it waterproof for wet snow?” | A figure with its unit, such as water column in mm | “Waterproof” with no figure; one statement covering both shell and zips | “Not seam-sealed”, if that’s the case |
Position 2 shows what the last row can look like: its description gives a water column “over 10 000 mm” and breathability of at least “8 000 g/m2/24h”. Those are figures with units that a parent or an agent can compare.
A “not for” line is worth writing. A shopper who needs a suit for long, still stretches at -30 °C is better served by “rating assumes active play” than by a refund later. Only publish limits you can stand behind. Spec-sold products shows how to turn shopper questions like these into category requirements, and agent-ready product data covers the per-variant record behind them.
What the evidence supports
The catalog searches show disclosure, not warmth. They show what ten snowsuit listings wrote and what Shopify’s catalog carried on one day, with no shopper context. Six of the ten came from one seller, so the result partly reflects one brand’s writing habits.
The generated-field changes are real but unexplained. A dropped hood, an added “proven” and a lost minus sign appeared in Shopify-generated fields. Shopify says those fields are marked Inferred and may vary in accuracy. We can’t see what they were generated from.
Nothing here measures what assistants do. We can’t see what ChatGPT or any other assistant tells a shopper about these products. Our matcher error shows that a checker can miss a stated fact. It doesn’t show that any assistant does.
How we checked
We sent both searches to Shopify’s agent catalog on October 26, 2026, with no shopper context, and saved the ten results for each exactly as received. On October 27 we re-read every snowsuit listing’s title, description, size options and generated spec, feature and selling-point lists. We recorded every temperature figure and warmth claim by field, every mention of activity, layering, body measurements and size labels, and every difference between the description and the generated lists. For the ice cleats, we compared the one listing that stated a Fahrenheit rating with its generated fields. The two -30 °C statements and our matcher error are also listed in our receipts for the snapshot; the other observations on this page are new. We did not query the catalog again, open product pages or test any garment.
Your next step
Take your warmest jacket or suit and write its rating on one line: signed number, both scales, scope, activity, layering and who set it. Add the height range for every size. Put that wording in your description and in a structured field, then preview the product in the Agentic section of your Shopify admin and check that the scope, the sign and the sizes came through unchanged.