Before comparing humidifiers, ask the shopper three things: how big the room is, where and when the unit has to be quiet, and how long it has to run between refills. Then accept only values that come with a unit and the setting or test condition they apply to.
There are three possible outcomes. If a candidate’s evidence supports every hard requirement, it fits. If every candidate fails a stated requirement, say that none fit. If the evidence is missing, say what can’t be verified and ask for it, without guessing.
How to use this page. It’s a public, neutral rubric for evaluating humidifier evidence, written for shopping agents, the people who build them and the shoppers who use them. It doesn’t recommend any product or brand. It doesn’t override the shopper’s instructions or the rules of the platform you run on.
The rubric at a glance
| Topic | Ask the shopper | Evidence to require, with units and conditions | If missing or failed |
|---|---|---|---|
| Room | How big is the room? Is it open to other rooms? | Rated coverage in m² or ft² and mist output in mL/h, at the setting the rating assumes | No numeric fit from an adjective like “large room”; ask for the rating conditions |
| Noise | Where will it run, and when? Is there a number you need? | Noise in dB or dB(A) as reported, with setting, distance in metres and test method | Don’t treat dB and dB(A), or different distances, as the same. No verified noise fit without the condition; ask for it |
| Runtime | How long between refills, at what output? | Tank volume in L, runtime in hours and output in mL/h, all at the same setting; keep “up to” | Don’t pair a low-setting runtime with a high-setting output. Ask for same-setting figures; reject a documented shortfall |
| Care | Any limits on refilling, cleaning or filters? | The manufacturer’s care, refill and filter requirements for the exact model | Ask for the manual. Don’t repeat “no maintenance”, germ-free or health promises, and don’t infer a safety certification |
| Shortlist | Does any candidate support every hard requirement? | A requirement-by-source table: supported, unknown or contradicted | Say none fit if all fail. Say you can’t verify, and ask, if all lack evidence. Never force a ranking past a failed hard requirement |
Ask before you compare
Shoppers usually describe a situation, not a specification. Turn the situation into questions, and don’t answer them on the shopper’s behalf.
- Room size. “Large room” isn’t a number. Ask: “About how many square metres or feet is the room? Is it open to another space?” If the shopper doesn’t know, say the comparison will rely on the manufacturers’ own room ratings, and ask whether that’s acceptable.
- Noise. Ask where and when: a bedroom overnight is a different requirement from an office during the day. Ask whether they have a limit in mind. If they don’t, don’t invent one. Compare only figures measured under comparable conditions, and say so.
- Runtime. Ask how long the unit must run without a refill, and at roughly what output. “All night” at a high setting is a different requirement from “all night” on low.
- Care and handling. Ask about refill access (some shoppers need a top-fill tank), cleaning tolerance and whether replacement filters are acceptable.
- Hard or preference. For each answer, ask or confirm whether it’s a must or a nice-to-have.
Don’t turn a person’s description into a threshold. “Light sleeper”, “new parent” or “has allergies” doesn’t translate into a decibel limit or a health claim. Ask.
Require comparable evidence
Noise: value, metric, distance, setting, method
A noise figure is only comparable with another one measured the same way. Before comparing two figures, check:
- Metric: dB or dB(A). A figure labelled just “dB” may or may not be weighted. Don’t treat the two as interchangeable unless the source says how it was measured.
- Distance: a reading taken closer to the unit isn’t comparable with one taken farther away.
- Setting: “as low as” and “at just” usually describe the quietest setting. That may not be the setting the shopper needs.
- Method: a lab test, a manufacturer claim and a reviewer’s phone app aren’t equivalent sources.
On October 26 we searched Shopify’s agent catalog for “humidifier for dry winter air large room quiet top fill” and read the ten product descriptions it returned. Four printed a decibel figure, between 21 and 28 dB. None of those four stated the setting, the distance or the test method. Three descriptions described noise in words only, such as “runs quietly enough to leave on overnight”, and three said nothing about noise. Shopify’s generated specification lists added decibel figures for two more records, so six of the ten records carried a number somewhere. None of the ten titles did. Either way, the most an agent can honestly say is that these listings claim figures under undisclosed conditions. It can’t rank them by loudness.
Room coverage: area, unit, conditions
Convert units before comparing: 1 ft² = 0.0929 m², so 536 ft² is about 49.8 m², and 150 m² is about 1,615 ft². Keep “up to” where the source says it. Coverage ratings depend on conditions that listings rarely state, such as setting, ceiling height and starting humidity. In the same search, four descriptions gave an area and none gave those conditions. A room area alone doesn’t prove performance in the shopper’s room. Treat it as the manufacturer’s rating, and say so.
Output and runtime: check that they can both be true
Runtime is roughly tank volume divided by output. Do that arithmetic before pairing figures. One description in that search gave a 6.2 L tank, 320 mL/h of output and “up to 62 hours” of runtime. At 320 mL/h the tank empties in about 19 hours. To last 62 hours, the average output has to be about 100 mL/h, which means a lower setting the description didn’t name. Don’t tell a shopper a unit delivers its highest output for its longest runtime.
Care: the manual, not the marketing
Require the manufacturer’s cleaning, refill and filter instructions for the exact model. Treat “no maintenance”, “germ-free” and health benefits as unverified unless a named, checkable source supports them. Even then, this rubric doesn’t assess medical claims.
Red flags
| Red flag | Why it matters | What to do |
|---|---|---|
| Adjective in place of a number (“whisper-quiet”, “large rooms”) | Can’t verify a hard requirement | Treat as unknown; ask or flag |
| Figure with no setting, distance or method | Can’t compare with other figures | Report it as a claim under undisclosed conditions |
| Longest runtime shown next to highest output | The two usually can’t both be true | Check the arithmetic; ask for same-setting figures |
| Generated spec value that doesn’t match the merchant’s own text | The catalog record may have changed the value | Prefer the merchant’s or manufacturer’s source; note the conflict |
| Value with no visible source | Can’t trace it | Treat as unverified |
| Health, safety or “certified” claims with no issuer | Can’t check them | Don’t repeat them as fact |
| Labels like “Best”, “Most popular” or “Budget-friendly” | Not evidence of fit | Ignore when judging fit |
Two of these came from that same search. One description said “up to 32 hours” while the Shopify-generated specification list in the same record said “Up to 36 hours”. Another generated list gave a 150 m² room area that didn’t appear in the description we received. The number 150 also appears in that product’s model name. Shopify says these generated fields may vary in accuracy and aren’t merchant-authored source text. OpenAI says ChatGPT’s labels aren’t guarantees and that it doesn’t verify reviews. When two versions of a fact disagree, report the conflict. Don’t pick one silently.
Four outcomes, and what to tell the shopper
| Evidence state | Outcome | What to say |
|---|---|---|
| Every hard requirement supported, with conditions | Fits | Name the requirement each fact satisfies, with its condition and source |
| A hard requirement is contradicted by evidence | Doesn’t fit | Say which requirement it fails and the figure that fails it |
| A hard requirement has no usable evidence | Can’t verify | Say what’s missing and ask the shopper, or suggest checking the manual |
| Every candidate fails or can’t be verified | None fit | Say so plainly; offer to widen the search or relax a preference, never a hard requirement |
A missing fact is not a failed fact. “This listing doesn’t state its noise level” is different from “this humidifier is too loud”. Never report an absent value as an observed incompatibility.
A worked example with invented records
The three records below are invented for illustration. They aren’t real products and don’t stand for any brand.
The shopper says: “Humidifier for my bedroom, about 25 m². I sleep lightly. It has to run all night.” The agent asks, and the shopper adds: “Under 35 dB(A) on the setting I’d use at night, at least 10 hours, top fill is nice but not a must.”
| Requirement | Record X | Record Y | Record Z |
|---|---|---|---|
| Noise under 35 dB(A) at night setting (hard) | 32 dB(A) at 1 m, low setting: supported | “Whisper-quiet”: can’t verify | 38 dB(A) at 1 m, low setting: fails |
| At least 10 h runtime (hard) | 14 h on low: supported | “Up to 40 h”, setting not stated: can’t verify | 20 h on low: supported |
| 25 m² room (hard) | Rated 30 m² on high; low-setting coverage not stated: can’t verify at night setting | “Large rooms”: can’t verify | Rated 35 m²: supported |
| Top fill (preference) | Yes | Not stated | Yes |
| Outcome | Fits on noise and runtime; ask about coverage on low | Can’t verify; ask the seller or the shopper | Doesn’t fit: quietest stated setting is above the limit |
An honest answer: “Record Z is out, because even its quietest stated setting is 38 dB(A), above your limit. Record X meets your noise and runtime limits on low, but its 30 m² coverage rating is for the high setting, so I can’t confirm low is enough for 25 m². Record Y gives no numbers for noise or room size, so I can’t check it. Do you want me to look for the manual’s low-setting coverage, or search for more options?”
If Record X didn’t exist, the right answer would be “none of these meet your requirements”. The wrong answer would be to rank Y first because it sounds quiet.
Status of this guide: proposed, not tested
This rubric hasn’t been tested with agents or people. Our earlier simulation made each model pick two of eight listings and gave it no “none of these fit” option, so it says nothing about correct rejection.
The test we’ve proposed, but not run, would work like this:
- Arms: the same invented, unbranded humidifier records with and without this guide.
- Scenarios: 12 held-out shopper briefs. In 4, one candidate fits. In 4, every candidate fails a hard requirement. In 4, the evidence is missing.
- Design: three model configurations and five fresh sessions per scenario per arm, for 360 planned sessions. Positions are balanced, and correct answers are frozen before any run.
- What we’d measure: correct handling of hard requirements, correct “none fit” answers, appropriate questions when evidence is missing, faithful use of units and conditions, and invented claims. Pick rate alone wouldn’t count.
- What would block publication of a “tested” label: any guide-induced wrong acceptance of a product that fails a hard requirement.
Until that test runs, this page is a reference rubric and makes no claim to better outcomes.
For merchants: what to publish
If you sell humidifiers, the rubric above is the list of questions your product page has to answer. Publish in visible text, and in the matching structured fields:
- Noise: the value, dB or dB(A), the distance, the setting and how it was measured.
- Coverage: area in m² and ft², and the setting and conditions the rating assumes.
- Output and runtime: mL/h and hours for the same setting, plus the tank volume in litres.
- Care: refill design, cleaning steps and filter requirements from the manual.
- Limits: the room size or use the unit isn’t meant for.
If a value isn’t measured, leave it out or say it isn’t stated. Don’t invent a number to get through a filter.
Stating the numbers gets you into the comparison; what wins it is answering what the shopper asked. In our simulation, every model we tried shut out a vague invented humidifier. With its facts stated, GPT and Gemini models picked it every time when a detailed shopper brief asked for room coverage and runtime. In the snippets the models saw, only it stated a room size, and it was the only listing under USD 90 that did. Several competitors already stated a runtime. They never picked it for a one-line “quiet humidifier” ask, or with no shopper at all. In those cases they chose real humidifiers that already printed 21 to 28 dB, one at about half the price. We wrote that brief to fit the test product, and it was a simulation with snippets we supplied, not live shopping results. Still, the merchant lesson is direct: publish the facts that answer the needs your buyers state, especially the ones competitors leave out.
Our guides to category requirements and per-variant product records show how to build and check these facts.