Publish every humidifier fact as a number, with its unit and the setting it applies to. Noise needs dB plus the setting, distance and test method. Room coverage needs ft² and m². Mist output goes in mL/h, the tank in litres, and runtime in hours on a named setting. Then check what Shopify’s catalog actually carries for your product, because its generated spec lists can add, drop or change values.
When we searched Shopify’s agent catalog for “humidifier for dry winter air large room quiet top fill”, 4 of the 10 product descriptions printed a decibel figure. None of the four said which setting, at what distance or by what method. That tells you what was written. It doesn’t show that the numbers caused any ranking.
This guide covers what to publish. Our humidifier field guide covers the other side: how an AI shopping agent should judge the evidence you publish.
The record to publish
One row per fact, per variant. The template below is proposed: it’s how we’d build the record, not a format any platform requires.
| Fact | Publish as | Condition that has to travel with it | Source to keep |
|---|---|---|---|
| Noise | Value in dB, and say whether it’s dB or dB(A) | Setting (low, sleep, high), distance from the unit, test method | Test report or manual page |
| Room coverage | Area in ft² and m² | Setting the rating assumes; ceiling height or starting humidity if the maker states them | Manual or spec sheet |
| Mist output | mL/h, per setting | Which setting each figure belongs to | Manual or spec sheet |
| Tank | Litres (add US gallons if your buyers use them) | None needed | Spec sheet |
| Runtime | Hours, per setting | The setting; keep “up to” if the maker uses it | Manual |
| Fill and care | Top fill or not; filter or filter-free; cleaning method | Model-specific | Manual |
| Not for | A room size or use it doesn’t suit | Only what you can stand behind | Manual or your own testing |
Three rules make this record hold up:
- Same setting, same row. Runtime and output for the same setting go together. The longest runtime almost always belongs to the lowest output.
- Both units for area. ft² for US buyers, m² for everyone else. A single unit invites a conversion error further down the line.
- Unknown is a value. If the manual doesn’t give a decibel figure, write “noise: not stated by manufacturer”. Don’t fill the gap with a guess or an adjective.
For the full cross-category template, see the product specification template.
Ten real listings, field by field
Here’s what the ten product descriptions returned for that search actually said, field by field. This counts the description only, the text the merchant supplied.
| Field | Descriptions that gave a number | Of those, with a setting or condition | Words only, or nothing |
|---|---|---|---|
| Noise (dB) | 4 of 10 (positions 2, 3, 5, 7) | 0 | 3 in words only; 3 nothing |
| Room area | 4 of 10 | 0 | 1 in words only (“medium to large rooms”); 5 nothing |
| Runtime (hours) | 8 of 10 | 2 named the setting | 2 nothing |
| Tank size | 8 of 10 | Not needed | 2 nothing |
| Mist output (mL/h) | 2 of 10 (positions 3 and 7) | 0 | 8 nothing |
| Top fill | 7 stated it | Not needed | 3 descriptions said nothing; 1 of those says “Top Fill” in its title |
Tank size and runtime were nearly always there. Noise and room area, the two things the shopper asked about by name (“quiet”, “large room”), were the least supported. Mist output, the figure you need to check a runtime claim, appeared twice.
Counting the spec and feature lists Shopify generates as well, 6 of the 10 listings carried a decibel figure (positions 6 and 9 added one). None of the ten titles did. Every count on this page says which field it counts, because the two numbers mean different things: the 4 is what merchants wrote; the 6 includes what Shopify generated.
The 21 dB listing and the “quiet overnight” listing
Two listings in that top 10 answered “quiet” in different ways:
| Listing | What its description said about noise | What else it published |
|---|---|---|
| Position 2 | “21dB Quiet” | 11 L tank, 700 ft², 100 h runtime; no setting for any of them |
| Position 4 | “runs quietly enough to leave on overnight” | Up to 400 sq ft, up to 36 hours on low, 1.25-gallon tank |
The number is the more useful answer. A shopper or an agent can compare 21 dB with another listing’s 26 dB. All anyone can do with “quietly enough” is repeat it. We wrote up this contrast when we first saw it.
Two cautions keep this example honest.
It isn’t a ranking result. The listing at position 1, above both, said nothing about noise in its description. Position 4, with no decibel figure, sat above position 5, which printed 28 dB. Result order in one search can’t tell you what a number is worth, and we don’t claim it does.
A bare number is only half an answer. “21 dB” with no setting or distance might be the quietest mode measured from across the room. Position 4 is actually ahead on one field: it’s one of only two descriptions that named the setting behind its runtime. Neither listing is “good” or “bad”. Each one is complete on some facts and missing others.
Worked example: rebuilding the “quiet overnight” record
Position 4 is worth rebuilding because Shopify’s catalog carried no generated spec or feature list for it at all. For an agent reading the catalog, the description was the record. Here’s every fact it published, and what the merchant would add:
| Fact | What the description says | Status | What to add |
|---|---|---|---|
| Noise | “runs quietly enough to leave on overnight” | Words only | dB or dB(A), setting, distance, method, from the manual or test report |
| Room coverage | “rated for single rooms up to 400 square feet” | Number, no condition | The setting the rating assumes; 37 m² alongside the ft² |
| Runtime | “up to 36 hours on low” | Number with setting | The runtime on high |
| Tank | “1.25 gallons” | Number | About 4.7 L, alongside the gallons |
| Mist output | Not stated | Missing | mL/h on low and on high |
| Humidistat | “Manual mist controls rather than a programmable humidistat” | A stated limit | Keep it. It tells an auto-mode shopper to look elsewhere |
| Fill and filter | Top fill, “no filter to replace” | Stated | Nothing |
Check the arithmetic before you publish. 1.25 US gallons is about 4.7 L. Running 4.7 L in 36 hours averages roughly 130 mL/h, so that is roughly what “low” has to mean. If the manual gives a low-setting output, publish it. If the manual’s figure doesn’t fit the runtime, one of the two numbers needs correcting before it goes live.
The other listings show the same trap from the other side. Position 3’s description stated a 6.2 L tank, 320 ml/h of output and up to 62 hours of runtime. At 320 mL/h, 6.2 L lasts about 19 hours. The 62 hours needs an average near 100 mL/h, which means a lower setting the description never named.
The finished noise row should look like this, with your real values in place of the brackets:
| Fact | Value and unit | Condition | Source |
|---|---|---|---|
| Noise | [value] dB(A) | [setting] at [distance], [method] | [manual or test report, page] |
If no test exists, write “noise: not measured”. That’s still more useful than an adjective, because an agent can see you checked.
What Shopify’s catalog added, dropped and changed
Your description isn’t the only text that reaches an agent. Shopify says some Global Catalog fields, including technical specifications, top features and selling points, are generated or enriched by Shopify, marked Inferred, and may vary in accuracy. In this result, those generated lists did more than tidy up the merchants’ text:
| Listing | What the merchant’s description said | What Shopify’s generated list said |
|---|---|---|
| Position 1 | Nothing about noise | “Quiet operation ensures minimal disruption in any home environment” |
| Position 6 | Nothing about noise | “Noise Level: As low as 26 dB” |
| Position 5 | “4-inch mist outlet”; “up to 32 hours” | “Mist Output: 4 inches (10.16 cm)”; “Up to 36 hours” |
| Position 10 | No room size | “Maximum Room Area: 150 m²” (150 is also in the model name) |
| Position 2 | “21dB”, “100H” | No noise or runtime value at all |
Read each row carefully. Positions 1 and 6 gained a noise claim the description never made. We can’t see Shopify’s inputs, so these values may come from merchant data we didn’t receive. Position 5’s nozzle size landed in the field where an output rate belongs; an output in inches can’t be compared with anyone’s mL/h. Position 10’s 150 m² is about 1,615 ft², more than twice the largest area any description in this result stated, for a 3-litre unit. It may be right, but nothing visible supports it. Position 2’s two most specific claims didn’t make it into the generated list at all.
The fix is the same in every case. Put the value, unit and condition in your description and in a structured field you control, and keep them identical. 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, so check your mapping there. If a generated value contradicts your own, your published source is what a careful agent or shopper can check against.
Check your record against the questions shoppers ask
This is the bounded test for your own listing. For each question a shopper might ask, your published record should hold the evidence in the second column, avoid the trap in the third, and say plainly when the product doesn’t fit.
| Shopper question | Evidence your record must hold | Unit or condition trap | Rejection case to publish |
|---|---|---|---|
| “Is it quiet enough to sleep next to?” | dB or dB(A), the setting, the distance, the method | dB and dB(A) aren’t the same; a figure at 1 m and one at 3 m aren’t comparable; “as low as” means the quietest setting | If the sleep setting is louder than the figure you advertise, say so |
| “Will it cover my 450 sq ft bedroom?” | Rated area in ft² and m², at a named setting | ft² vs m² conversions; coverage on high quoted for a unit run on low | “Not for rooms over [area] on the sleep setting” |
| “How often will I refill it?” | Tank in L, runtime in hours and output in mL/h, all on the same setting | Longest runtime paired with highest output; gallons vs litres | Runtime on high, if it’s short |
| “Is it top fill? Does it need filters?” | Fill design; filter type and replacement interval | “Easy to fill” isn’t the same claim as top fill | “Requires [filter] every [interval]” |
| “Does it hold a set humidity on its own?” | Humidistat or auto mode, yes or no; sensor accuracy if stated | “Smart” isn’t a humidistat | “Manual controls only; no humidistat” |
The last column is the one merchants skip. A true “not for” line costs you a sale you would probably have refunded. It also gives an agent a reason to trust the rest of your record. Position 4’s “manual mist controls rather than a programmable humidistat” is a good model: plain, specific, and useful to the buyer who needs auto mode.
How an agent should weigh these answers, including when to say none of the candidates fit, is in the humidifier field guide. Spec-sold products shows how to build this question list for any category.
What the evidence supports
The catalog search shows disclosure, not performance or ranking. It shows which facts merchants wrote and what the catalog carried. It doesn’t show how loud any unit is, or why any listing ranked where it did.
A simulation shows that facts can be matched, and when they win. We placed an invented humidifier among seven real catalog listings and asked six AI models from Claude, GPT and Gemini to pick two of the eight, so chance alone picks it 25% of the time. With vague copy it was picked in 3 of 180 Claude sessions (three models), 0 of 60 on GPT-6.1-Sol, 6 of 60 on Gemini 3.8 Flash and 3 of 54 on Gemini 3.1 Pro. With its noise, coverage, tank and runtime written as a spec line: 134 of 180, 20 of 60, 19 of 59 and 15 of 55.
On GPT and Gemini that result split completely by what the shopper asked. With a detailed brief (a light sleeper, a 450 sq ft bedroom, under $90, hates refilling daily), they picked it every time: 20 of 20, 19 of 19 and 15 of 15. With a one-line ask or no shopper at all, 0 of 40 on each model. Claude picked it in 59 of 60, 32 of 60 and 43 of 60 sessions across those three briefs. In the snippets the models saw, ours was the only listing under USD 90 that stated a room size; one competitor above USD 90 stated 700 ft². Several competitors stated runtimes (100, 32 and 62 hours). Competitors also already printed 21, 26 or 28 dB, and the 28 dB one cost USD 39.99 with 310 ratings against our product’s 18.
We wrote that brief to fit the test product, so this shows facts meeting stated needs, not what a random shopper does. The lesson for your record: vague copy got excluded by every model, facts got the product considered, and it won when its facts answered a need the shopper stated that competing listings didn’t. Once every listing prints a decibel figure, the noise figure alone stops setting you apart. The full write-up has the detail and the limits. It’s a simulation with snippets we supplied, not live ChatGPT results.
The documentation supports the method, not a promise. OpenAI says ChatGPT shows products it judges relevant to the shopper’s query and context. Shopify says its generated fields may vary in accuracy. Neither publishes ranking weights, and neither says a complete record gets picked.
How we checked
We sent the search above to Shopify’s agent catalog on October 26, 2026, with no shopper context, and saved the ten results exactly as received. On October 27 we re-read every listing’s description and its generated spec, feature and selling-point lists, and counted each field separately: decibel figures, room areas, runtimes and whether a setting was named, tank sizes, mist output in mL/h, and fill design. The tank, output, generated-noise and mist-outlet observations on this page are new; the others are also listed in our receipts for the snapshot. We did not query the catalog again, open any product page or test any unit. Store and brand names are withheld, and quotes are kept short.
Your next step
Open your best-selling humidifier’s manual and fill five rows: noise, coverage, output, tank and runtime, each with its unit, setting and the page it came from. Put each value in your description and in a matching structured field. Then preview the product in the Agentic section of your Shopify admin and mark any value that differs from your own.