Hi everyone,

Every air purifier (well, most of them) with a sensor in it makes the same promise: leave it on auto and it'll sort itself out. You don't need to think about it, because the machine is automatically keeping the air quality ‘good’.

However, when you're actually relying on these devices, you start paying closer attention to what they're doing. When you start doing this, well, you notice that many of them have very different ideas of what ‘good’ air quality is.

This is particularly pertinent to my family at the moment, because fires in Indonesia have been burning for months and the smoke has drifted a long way, including over here in the Philippines. Most days now I look outside and the haze is just there, sitting over everything, and it hasn’t lifted for weeks.

It's changed quite a few things in our house. I've cut my runs right down, because going out and breathing hard in that air isn’t something I want to be doing. And we have a six-week-old in the house now, which has meant making some hard calls about where our purifiers go, which rooms we keep sealed, and what we do when the kitchen inevitably adds to the problem.

Anyway, that’s the background for this issue. Let’s discuss why you probably shouldn’t trust automatic modes on air purifiers.

The Issue With Automatic Mode

I started paying proper attention to this when I reviewed Xiaomi's purifiers earlier this year - the Mijia Smart Air Purifier 6, the 4 Lite and the 4 Compact. The three of them run in our house, and I wanted to know what they were doing when I wasn't watching. When were they ramping up? When would they reach maximum speed?

The first issue I ran into is that many manufacturers simply won't tell you. For example, I asked Levoit directly about the Vital 200S-P and was told they can't share the thresholds. The manual gives you four colours and no numbers, and the app gives you a word (‘good’, ‘poor’, etc) instead of a measurement. More on that one shortly, though, because this leads into another issue with automatic mode.

Xiaomi, to their credit, does publish thresholds. Here's what their documentation says the purifier is working to:

Band

PM2.5

What the fan does

Excellent (green)

1–20 µg/m³

Sleep speed

Good (yellow)

21–35 µg/m³

Level 1

Moderate (orange)

36–55 µg/m³

Between Level 1 and Level 2

Poor (red)

56 µg/m³ and above

Between Level 2 and maximum

I’m also happy to say that these numbers aren’t arbitrary: China's national ambient air quality standard rates PM2.5 from 0 to 35 µg/m³ as Excellent, so Xiaomi's top band at 20 is actually tighter than their own country's standard requires. These aren’t random numbers (as I’ve seen all too often). The trouble starts when you put them next to the health guidance instead, because GB 3095-2012 (the Chinese standard these purifiers are following) is an outdoor standard for 24-hour averages over cities.

The WHO's 2021 guideline values are 5 µg/m³ as an annual average and 15 µg/m³ over 24 hours. The US EPA's current annual standard is 9 µg/m³, with a 24-hour standard of 35. Keep in mind that air purifiers usually live in our homes - the areas where we spend the majority of our lives.

This means the ceiling Xiaomi calls "excellent" - 20 µg/m³, the level at which my purifier is still happily sitting on sleep speed - is already above the WHO's 24-hour guideline and four times their annual one. The machine will not go to maximum until the air hits 56 µg/m³, which is nearly four times the WHO 24-hour figure (and even then, it sits below the maximum speed you can manually set the device to). As with most air purifiers, automatic mode can’t be adjusted, so these thresholds are fixed.

For what it's worth, I try to keep our house under 5 µg/m³, which is where I think the target should sit (if not lower) if you've gone to the trouble of buying purifiers in the first place. I appreciate that's stricter than many people’s target number, but I'd argue almost everyone should be aiming well below 20.

What Is Everyone Else Using?

That's just Xiaomi. I went looking at a few other brands to see whether Xiaomi was an outlier, and the picture is a bit of a mess. Let’s take a look at some of the most famous air purifier brands on the market.

Brand (model*)

Published thresholds?

The bands

Can you change it?

Coway (Airmega Icon)

Yes

Good 0–15, Normal 16–35, Unhealthy 36–75, Very unhealthy 76+ µg/m³

Yes - Sensitive, Normal or Insensitive

Winix (5500-2, AM90)

No

Three colours, no numbers at all

No

Levoit (Core 600S)

Yes

Blue below 35, green 36–75, orange 76–115, red above 115 µg/m³

No

Airthings (Renew)

Yes

One fixed trigger at 25 µg/m³

No

*model indicates which model the thresholds were sourced from.

Coway is the clear winner here. They print the concentrations in the manual, their "good" band tops out at 15 µg/m³ (which actually matches the WHO's 24-hour guideline) and you can shift the whole scale with a Sensitive or Insensitive setting if their default doesn't suit your home. I would still argue that 15 µg/m³ is too lenient for most homes, but this is a big improvement and the ‘sensitive’ setting brings this down (although I can’t find to what exact value).

Coway’s PM2.5 thresholds

Winix is the opposite end. Three colours, no numbers, nothing published anywhere I could find. Interestingly, on several models, including the very popular 5500-2, the "Smart Sensor" is described as detecting ‘odorous gas’ (VOCs), with the fan speed in auto mode adjusting to the amount of gas it detects. I haven’t yet tested Winix devices, so I don’t know if it also reacts to particles, but if you've been running a 5500-2 on auto expecting it to react to smoke or haze, this is worth keeping in mind.

Interestingly, Coway's bands - 15, 35, 75 - are South Korea's national air quality standard. Each company has reached for whatever scale its home regulator publishes, and your purifier's idea of clean air depends rather a lot on where it was designed. This does make sense, and it leads to a bigger issue that is outside the scope of this newsletter - AQI scales themselves aren’t even comparable!

The Second Problem: The Sensor Doing the Measuring

‘Ghost’ spike from the Levoit 200S

Everything above assumes the purifier knows what the air quality actually is. After a few years of comparing these sensors against monitors I trust, I think that's a generous assumption. They're inexpensive parts sitting behind an intake that fills with dust, and most of them aren't close to a dedicated monitor.

Here are three devices from my own reviews, each wrong in a different way.

Levoit Vital 200S-P - invents pollution. That graph above is the Vital reporting a spike above 400 µg/m³ while my dedicated air quality monitor, sitting right beside it in a closed room, showed 20. During real spikes my monitors put it at 100 to 150 µg/m³, it read above 600. I only noticed because the fan kept ramping for no reason I could see. If you run Home Assistant, the VeSync integration exposes the raw value the app itself hides from you.

Provirtec Pawtent Pro - can't see clean air. In two months I never once saw it read below 5 µg/m³, while our house sits under 1 for much of the day. At the other end it under-reports badly: 153 µg/m³ while my AirGradient and PurpleAir Zen were both north of 500.

Mijia 6 - the best of the three, and still not a control input. It tracks trends well, which is more than I expected. But it over-reported 425 µg/m³ against my three AirGradient monitors averaging 277, and throws occasional spikes from a steady 15-20 up to around 50 with no cause I could find.

To be fair, better parts exist and some manufacturers fit them. Airthings puts a proper laser sensor in the Renew, and Levoit's own Core 400S and 600S get one instead of the infrared part in the Vital. So this is often a cost decision rather than a technical limit. The trouble is you can't tell which you're buying, because nobody publishes a sensor accuracy figure the way they'll publish a CADR.

So there are two problems stacked here: a threshold that's too lenient for an indoor space, and a reading feeding it that might be out by a factor of three or more.

Does anyone get it right?

While I haven’t yet had the chance to test Mila air purifiers, they look to get this part right!

Coway, as above, is the one that comes closest. Two others are worth a mention for getting part of the way there.

Mila: hands you the target

Mila is the only purifier I've found that puts a Target AQI on the device itself, right next to the current reading, and there's a mode called Bubble Boy that effectively means "keep the air as clean as you can and don't worry about the noise". This is the right philosophy, because clean enough should be a decision for the person in the room. The catches are that it's an AQI score rather than µg/m³, and at around $300 plus $150 to $200 a year in filters you're paying a lot for less airflow than my Mijia 6 moves.

Airthings: so close

Airthings builds air quality monitors for a living, and the Renew has a proper laser sensor to show for it. They then set the auto mode trigger at 25 µg/m³ and locked it. That's the tightest fixed threshold I found anywhere and better than 35, but it's still above the WHO 24-hour guideline and there's no way to move it. The part that really gets me is that Airthings will happily let you set your own PM2.5 threshold through IFTTT - just not on the purifier itself.

So Is Automatic Mode Useless?

No, and this is important to clarify.

What the evidence says

An intervention study in Building and Environment put auto-mode air cleaners into real homes and compared three scenarios: filters removed, auto mode, and an adjustable mode where the occupant set the speed themselves. Auto mode cut indoor PM2.5 by 31% during cooking and 28% otherwise against the units with no filters. It also beat the user-adjusted mode, by 19% during cooking events and 4% the rest of the time.

So auto mode is better than nothing, and better than relying on a person to remember. My argument isn't that automation doesn't work, but rather that the target it's automating towards is too lenient, and a 28% cut from a bad starting point can still leave you above the guidelines. Bad automation beats no automation. Good automation beats both. Read the study →

Worth reading

I'm not the first to notice any of this. Smart Air ran their own auto mode tests years ago on Xiaomi and Philips units in Beijing and Bangkok, found the purifiers left indoor air in the unsafe range overnight, and published the raw data as a downloadable file. They also flagged the thing I ran into when researching Winix - that a good number of purifier "air quality" sensors are gas sensors rather than particle sensors.

What I like most is what they did next. Perhaps this is part of the thinking behind their devices like the Sqair I reviewed, which has no sensor and no auto mode at all.

What I Do

My current Home Assistant control panel.

So here's how we handle it in our house. The short version is that I've stopped treating the purifiers as smart devices.

AQ monitors decide, not the purifiers

Every room that matters has a dedicated air quality monitor in it, and those readings are the only ones that get a vote. All four of our purifiers can be reached from Home Assistant, so fan speeds are set by the monitors and the purifier's own sensor is ignored entirely.

The target is 5 µg/m³

That's the number I want our house to stay under. It's stricter than most people need and much easier in a small house with four purifiers than a draughty one with a single unit - but it's the right direction, and it's a long way below the 20 µg/m³ my Xiaomi calls excellent.

Leave a gap between up and down

This is one aspect that many people get wrong on their first attempt. If the fan ramps up above 5 µg/m³ and drops back below 5 µg/m³, it'll sit on the boundary flicking between speeds all night, which is unbearable in a bedroom. Mine are set to ramp up at a higher number than they step back down at, with a few minutes of delay on the way down, so it settles instead of oscillating.

No monitors? Just pick a speed

If buying a monitor per room isn't feasible, this is the free version and it's most of the benefit: turn auto mode off and leave the purifier on a speed you can live with. A fan running steadily at a sensible setting will beat a smart one making bad decisions on a sensor that can't see straight. Level 2 on the Mijia 6 is our sweet spot, and speed 2 on the Levoit is both quiet and perfectly adequate.

Before You Go

That’s all for today! If you found this useful, please pass it on to someone who'd appreciate it. BreatheSafeAir is still just me - no team, no marketing budget, just a lot of test equipment and a house I keep burning incense in. Forwarding this to one person is the single most helpful thing you can do, and it costs you nothing.

Thanks for reading, and apologies this one ran a little longer usual - you might be able to tell, but this is a big issue I have with most purifiers!

Until next time,

Ethan

Keep Reading