BreatheSafeAir is an independent, one-person publication. I don't have a marketing budget or a team behind me - just a genuine interest in helping you make sense of air quality. If you find this useful, sharing it with someone who'd appreciate it means more than you'd think.

I get a fair few emails every week, and one question keeps surfacing in slightly different shapes. Can I trust the CO2 reading on this monitor? My VOC sensor is giving weird measurements, should I believe them? Should I trust consumer-grade PM readings? It's not the single most common message that lands in my inbox, but it's close.

This issue is my attempt to fix that. It's about which of the readings on these consumer devices you can actually rely on. A few of them are good enough to be taken at face value. Others are close enough to be useful, so long as you know how to read them. And then there are the ones I'd treat as little more than a rough guess (or worse!).

This is a real problem in air quality monitoring because a proper CO2 measurement and a cheap VOC estimate can sit right next to each other on a monitor’s screen, both looking equally certain. Let’s go through it sensor by sensor, starting with the readings you can trust.

Ah, and the needed disclaimer: There are some generalisations made in this issue. Please keep in mind that there are always outliers, but this is the general state of consumer-level air quality monitoring in 2026.

When it comes to CO2, what you want to see on the spec sheet is NDIR, or non-dispersive infrared. These sensors shine an infrared beam through the air and measure how much of it the CO2 absorbs, so they’re reading the gas directly. For this reason, NDIR has been the benchmark for years. The good news is how cheap the technology has become. You'll now find proper NDIR sensors in monitors as low as $30, but you’ll also find them in monitors from reputable brands like Aranet, AirGradient and Airthings. These can be considered very accurate.

Photoacoustic sensors are the other sensor type worth knowing, and Sensirion's SCD4x series has made them a good alternatives to NDIR when size is the focus (NDIR sensors are normally much larger, but SenseAir’s S12 might change that!). They’re accurate enough for any normal home, though, since they work by listening for tiny pressure changes from the gas, this means vibration and loud surroundings can throw them off. In most cases, however, this isn’t an issue. These sensors can also be considered accurate.

Newer still are thermal conductivity sensors like the STCC4, found in the Atmotube Pro 2 and the IKEA Alpstuga. They're a cheaper (and smaller) way to get a CO2 figure on screen and look okay for spotting trends, enough to tell you a room's gone stuffy, but they sit well below NDIR and photoacoustic on accuracy. I've been testing the Alpstuga and have plenty to say about its sensor, so I'll save that for the review coming shortly.

One thing to avoid outright is eCO2 as this figure doesn't come from a CO2 sensor at all. Rather, it's an estimate pulled from a cheap VOC sensor, and it'll happily shoot past 2000 ppm because someone used hand sanitiser nearby while the real CO2 hasn't budged. If you see eCO2 on a spec sheet, run for the hills.

There is also one caveat that I need to mention: Most CO2 sensors (NDIR, Photoacoustic and thermal conductivity) use automatic baseline calibration, which assumes they periodically see fresh outdoor air near 420 ppm and resets to match. This is fine in a normal home or for a monitor that goes outdoors often but in a bedroom that’s shut up overnight, where the sensor never sees that baseline, it can drift low and under-report what's building up. Some devices let you turn this off and calibrate by hand, which is what I'd do before leaving one in a room that rarely clears.

Radon is the other reading I'm comfortable leaning on, and the good monitors are accurate. The reputable brands land within about ±10% of the true value, and that figure holds up under independent testing. In Canada, the C-NRPP runs radon monitors through a calibrated chamber and publishes how they perform, and the devices that pass come from names like Airthings, Ecosense (the brand behind RadonEye) and Aranet. If a monitor performs well in C-NRPP's reports, you can take its numbers seriously.

However, as always, there’s a catch and it comes down to time. Radon detection is a counting exercise at heart. The sensor tallies the decay events thrown off by radon in the air, and because that decay is random, a short measurement window gives you a jumpy, noisy number. The longer it counts, the more those fluctuations average out and the closer the reading sits to reality. That ±10% you see quoted almost always carries a time condition, usually something like "after 7 days", because the device can't be that precise in the first few minutes - or hours.

So only treat early readings as a rough orientation and nothing firmer. Some monitors advertise decent accuracy over a day or a week, which is fine for a quick look, but the figure that matters is the long-term average. Radon shifts with the weather and the season, so a reading taken over a single evening can look alarming or reassuring for reasons that have nothing to do with your true exposure. C-NRPP's own advice is to leave a monitor running for at least three months before you trust the average it gives you.

Every PM sensor has its own traits. Some are known to overreport, others lean low, and two units from the same product line can don’t always agree. The general consensus, though, is that low-cost sensors like the Sensirion SPS30 and the Plantower PMS5003 are good enough for trend identification and usually show the same trends as reference equipment. Where they slip is in very high humidity, where they tend to overreport as moisture swells the particles they're counting.

What you want to look for is a laser scattering sensor. These work by firing a laser at the passing air and reading the light that scatters off particles, which is the approach behind almost every decent consumer PM sensor.

The caveat is that all of this applies to PM2.5. PM1 holds up reasonably well on some sensors too, but PM10 is mostly guesswork, and nearly every low-cost sensor struggles badly with it. The bigger particles that make up PM10 are exactly the ones these sensors are worst at sizing and catching, so the figure ends up being more of an estimate than a measurement. Treat PM10 with scepticism. PM2.5 and often PM1 are trustworthy enough (from reputable sensors) for spotting trends, but they'll show noise from reading to reading, so you're better off watching an averaged value than reacting to every twitch on the live number.

CO is the happy exception among the electrochemical sensors (we’ll get to those soon). Before the gloom about NO2 and ozone, here's a gas measured with an electrochemical cell that comes out fine, because the technology has had decades to mature around it. Even budget parts like the Figaro TGS5342, which I found inside the Sensereo when I opened it up, are properly selective to CO and carry UL2034 recognition for alarm use.

The catch (there’s always one, isn’t there?) sits at the low end. These sensors produce a tiny current, so at the handful of ppm you'd see in normal indoor air the signal almost disappears into noise and baseline drift. A cheap monitor showing you 2 ppm of CO is likely wrong. Where the same sensor earns its keep is higher up, at the tens to hundreds of ppm that mark an actual problem, because there the current is larger and easy to read with confidence. That's the range CO alarms are tested against, and it's why a certified CO alarm is what you should rely on for safety, rather than the CO figure on a general air quality monitor.

Getting accurate CO down at low, everyday concentrations is possible, but you're into the territory of a higher-grade electrochemical cell and proper calibration, which costs far more than the part in a $40 monitor. For most people that's overkill. Treat the cheap CO reading as a coarse safety backstop, watch whether it's climbing, and don't read much into whether it says 1 or 4 on a quiet day.

Measuring formaldehyde well doesn't cost a fortune, but you do need to be particular about the sensor. An example of a decent sensor at the moment is Sensirion's SFA30 (now the SFA40), an electrochemical part rated to ±20 ppb or ±20% of the reading, whichever is larger. Its trick is barely reacting to ethanol, so unlike the cheap semiconductor formaldehyde sensors, it won't jump just because someone used hand sanitiser or sprayed cleaner nearby.

You'll mostly meet these sensors through Sensirion's all-in-one modules - the SEN68 and the SEN69C, which build an HCHO channel into the same package. The cheaper end of the market still leans on semiconductor parts, so formaldehyde on a budget spec sheet is worth a second look before you trust the number.

Two VOC sensors show up in most consumer monitors: Sensirion's SGP series and Bosch's BME680 (the BME688 is the newer version). Both companies now report a relative index instead of an actual concentration and they've more or less admitted that pinning VOCs down properly at this price is impossible.

The reasons stack up. These sensors are cross-sensitive, so they react to a whole range of gases at once and can't tell you which one set them off, and they respond more strongly to some compounds than to others. They also drift over time, and their baseline wanders, so the same air can read differently depending on how long the sensor's been running and what it's been breathing in lately.

A true absolute reading is possible, but it costs far more than any consumer device is charging. What you get instead is a sense of the trend. Are VOCs climbing, dropping or holding steady? That part is still useful. It'll catch the spike when you start cooking or open a tin of paint, then show it settling once the room clears. Just read it as a relative scale, because that's all it is.

If you ever see a consumer-grade VOC sensor spitting out a figure in ppb or ppm, be sceptical of it. That number is dressed up to look like a real measurement, and it's almost certainly wrong, sometimes by a wide margin.

NOx sits in much the same place as VOCs, and for the same underlying reasons. The main thing to know is why it shows up so often. The SGP41 that handles VOCs on a lot of monitors has a NOx channel built into it, so you get a NOx reading almost as a byproduct of the VOC sensing. Sensirion's all-in-one SEN6x series bundles it in too, and since that module turns up in plenty of devices, NOx comes along for the ride.

As with VOCs, it’s a relative trend and nothing more, and it's not even a strong one. In my experience it's a step down from the VOC index on both reproducibility and accuracy, so two readings taken under similar conditions can disagree more than you'd like. Useful as a rough nudge that something's changed, not as a figure to take seriously - especially in typical indoor concentrations.

To measure either NO2 or O3 with any real accuracy you need a quality electrochemical sensor, and those often run north of $100 for the sensor alone, before any of the rest of the device gets added. This means that monitors with dedicated (and decent) NO2 and ozone sensors normally only exist in the > $1000 price bracket.

It gets more difficult (and expensive), because you usually want both at once. NO2 sensors are cross-sensitive to ozone, so any ozone in the environment skews the NO2 figure unless you can account for it. The usual fix is to run an ozone sensor alongside and correct one against the other, so you end up buying the pair. A monitor that measures NO2 properly tends to carry two costly cells, and the price tag shows it.

So when a cheap monitor claims NO2 or ozone, I treat it with a heavy dose of suspicion. At consumer prices there's no room in the parts budget for the sensors it would genuinely take, so whatever number it shows is almost certainly coming from something cruder, or estimated outright.

If you take one thing away, let it be a rough hierarchy of trust. CO2 from a real NDIR sensor and radon from a C-NRPP-tested monitor are the readings you can act on directly. PM2.5 sits just below, good for trends as long as you remember it gets worse in high humidity and that PM10 readings are not reliable. The VOC and NOx indexes only tell you which way things are heading, so read them as a direction and not a value.

Formaldehyde is doable when the monitor uses a proper sensor like the SFA30, while NO2 and ozone are the parameters to treat with real suspicion on anything affordable, since measuring them well costs more than most consumer devices will spend. CO is its own case, reliable at the high levels that signal danger and vague at the trace levels that don't.

It’s important to know what each number on your air quality monitor is worth, so you know what you should trust and what you should take with a grain of salt. I hope this email has helped clarify which values you can take at face value and which you should approach with caution.

Until next time,

Ethan