Do robot vacuums actually work on carpet?

Suction in pascals is the figure every listing leads with, and it is the one that answers this question least well.

This question has a real answer, and it is not the one the product listings imply. The listings compete on a suction figure that measures the wrong thing, and the property that actually determines carpet performance — how much air the machine can move through a mat of fibres — is a hard ceiling set by the machine’s size. Below is what independent testing shows, why the headline number misleads, and the brush decision that most buyers never see.

How this was researched

We sourced this from the published results of independent vacuum testing programmes — including the testers who buy every unit themselves and the consumer organisation that measures embedded debris on a standard carpet — alongside published suction measurements that separate static pressure from airflow, and manufacturer specifications.

We did not run a robot vacuum on a carpet, and we have not measured pickup on any model. Where a figure comes from a testing organisation, a published measurement or a manufacturer’s own specification, we say which. Where testers disagree about the cause, we report the disagreement.

The answer, by pile height

Carpet performance is not one question, because carpet is not one surface. The useful split is by pile — the height and density of the fibres — because that is what changes how much air the machine can pull through.

The middle row is where most households actually live, and it is where the honest answer gets uncomfortable. Independent testers running a large library of robot vacuums describe the best carpet performer they have measured as strong on low and high pile for larger debris, but still liable to struggle with finer material that has been worked deep into carpet fibres. That is not a criticism of one model. It is the shape of the category.

Why the Pa number does not answer this question

Every listing leads with a suction figure in pascals, and those figures have climbed absurdly — from a couple of thousand a few years ago to five figures on current flagships. Read that number as though it predicts carpet cleaning and you will buy the wrong machine.

Pascal ratings describe static pressure: the maximum vacuum a motor can pull when the airway is completely sealed. The standard illustration is pinching the end of a garden hose — pressure against your thumb climbs, but no water is moving. Static pressure is the pressure with nothing moving.

Cleaning requires moving air. Debris is carried by flow, and flow is measured in cubic feet per minute. Here the category’s ceiling appears:

That gap is the honest explanation for why an expensive robot does not deep-clean thick carpet the way a cheap corded upright does. It is not a quality problem that more engineering will fix. A machine that has to carry its own battery and fit under a sofa cannot move the same volume of air.

There is a second measurement that matters, and almost no manufacturer publishes it. Sealed suction, quoted in kilopascals, is the pressure measured at the machine’s cleaning head with the whole air path in place — dustbin, filter and brush assembly included. It is the difference between an engine’s output and what reaches the wheels.

The published comparison is worth sitting with: a flagship rated at 16,600 Pa was measured at the highest sealed suction of its generation at 2.76 kPa — while a unit rated at 20,000 Pa delivered about 2.0 kPa, and one rated at 22,000 Pa delivered about 1.8 kPa because its air path was less efficient. The headline number ranked the three machines in the wrong order.

What actually decides carpet performance

If pascals are the wrong number, what is the right one? In practice, two things: how much air the machine moves, and what the brush does to the carpet.

On a hard floor, a well-designed brush sweeps debris toward the intake and airflow carries it away. On carpet, the brush has to physically beat and agitate the fibres to loosen material that has settled below the surface — and then airflow has to carry it out. The two jobs are different, and that is where the trade-off lives.

This is the decision most buyers never see, because it does not appear in the headline specs. Most current flagships have standardised on rubber primary rollers plus some form of anti-tangle mechanism — a choice driven by the pet-hair and long-hair complaints that dominate owner reviews. The consequence is a category of machines that is very good at not wrapping hair around a roller, and less aggressive on carpet than a bristle design would be.

Independent testing shows the same split from the other direction. The tester’s top carpet pick uses bristled rollers — and the same reviewer notes those rollers “really struggle to pick up pet hair on carpeted surfaces.” A different model praised for lifting pet hair thoroughly from carpet is not the carpet pick. You are choosing which failure mode you would rather live with.

What a real carpet test actually measures

It is worth knowing what the testing organisations put a machine through, because it explains why their verdicts read more cautiously than a product page.

The consumer organisation that publishes vacuum ratings describes its carpet test as measuring how much embedded sand or flour a vacuum lifts from a medium-pile carpet, and its pet-hair test as embedding fur into medium-pile carpet and weighing what is left behind after several passes. That is a test of material below the surface — the hardest case, and the one where the airflow ceiling above bites.

Two details about how those programmes work matter for reading the results. First, they calibrate the difficulty by machine type: full-size uprights and canisters are put through a tougher version of the same test than smaller machines. Second, for robot vacuums specifically, the same organisation built a dedicated test space with the obstacles robots actually meet — carpet fringe, cords and debris — and maps each machine’s movement through it.

That second point is easy to skip past. For a robot, getting to the carpet and staying on it is part of the result, not a preliminary.

The carpet problems that are not about suction at all

Some of the most annoying carpet failures have nothing to do with cleaning power. They are navigation and design issues, and they are worth checking before purchase because they cannot be fixed later.

The dark-carpet one deserves emphasis because it is counter-intuitive and it is reported for real models. A cliff sensor works by emitting light and watching for its reflection; a very dark surface reflects little, so the machine concludes the floor has ended and turns away. Owners of dark rugs have reported exactly this, and the workarounds — covering the sensors or manually marking the area — are not obvious from a listing.

One more thing about that suction figure

Even taken at face value, the pascal number on a listing is a maximum, and it is not what the machine runs at most of the time.

Published guidance on robot vacuum power modes puts a machine’s default “standard” setting at roughly 40–60% of its rated suction, with the headline figure reserved for a turbo or maximum mode. Running at maximum also drains the battery substantially faster, which is why machines that boost power on carpet do so for a bounded time rather than throughout a run.

The practical consequence: two machines advertised at 20,000 Pa and 35,000 Pa may differ by a few thousand pascals in the mode they actually clean in, which is a difference most owners would not detect. Once you are above the level needed for your floor type, additional pascals buy very little — while brush design and airflow buy a lot.

What to check before buying

• What pile is your carpet, honestly? This is the whole decision. Low and medium pile: a robot is a reasonable primary vacuum. Thick plush or frieze: plan on keeping a corded vacuum for the carpet, whatever you buy.
• Does the manufacturer publish sealed suction or airflow? If the only figure is pascals at the motor, you are being shown the number that is easiest to inflate.
• Bristle or rubber rollers? Bristle agitates carpet better and tangles hair more. If you have shedding pets and carpet, you are choosing between two compromises — decide which one you can live with.
• Is your carpet dark? Check whether the model has known cliff-sensor issues on dark surfaces before you buy, not after.
• Can it lift its mop pads? If it mops and you have carpet, mop lift is what stands between you and a damp rug.
• Can it climb your thresholds? Look for a stated climbing height and compare it against the actual lip between your rooms.

Rated suction, and what the machine actually runs at

The headline number on the box is the turbo figure. Published guidance puts the default “standard” mode at 40–60% of it. Applied to the 76 kPa example on this page, that is 30 to 46 kPa in the mode most people leave the machine in — our multiplication, on the guide’s percentage range and the manufacturer’s rating.

It matters on carpet, because carpet is where the gap between the two figures shows up.

Working backwards from the number you need

The earlier section divides the rated figure down to the standard mode. Buyers need the opposite: given the suction a carpet actually requires, what has to be on the box? The inversion is our arithmetic, on the guide’s 40–60% range:

Divide the figure you need by 0.4 and by 0.6. A carpet that wants 40 kPa in the mode you will actually use means a rated figure somewhere between 67 and 100 kPa. Need 30 kPa on carpet and the box has to read 50 to 75 kPa.

The 40–60% range is the guide’s; the division is ours. It is also why a comparison table of headline suction figures can rank two machines in the wrong order — the modes behind those figures are not necessarily the same.

The honest answer

On low and medium pile, yes — a robot vacuum does the job, and for a lot of homes it genuinely can be the only vacuum. On anything thicker, the honest position is that it will maintain the surface and leave the depth to something with a cord. That is a limitation of moving air through fibres with a machine the size of a dinner plate, and no pascal figure on a box changes it.

If your reason for buying is carpet, buy on brush design, airflow and the navigation details above. If your reason is daily maintenance on mixed flooring, the suction race barely concerns you, and the cheaper machine is often the right one.

Questions this page answers

Do robot vacuums actually work on carpet?

On low and medium pile, usually yes — well enough that it can be your only vacuum. Surface debris, pet hair and tracked-in grit all come up. On thicker carpet it is only partly true.

What do independent testers find on thick carpet?

Testers who have run more than 95 models report that even their best carpet pick can struggle with fine material worked into the pile.

Does a higher Pa rating mean better carpet cleaning?

Not by itself. Pascals measure static pressure with the airway closed, not the airflow that actually lifts debris.

What trade-off does no robot vacuum spec sheet list?

Brush design. Bristle rollers agitate carpet better but tangle hair; rubber rollers resist tangling but agitate less. Most flagship robots now use the second kind.

The documents behind this

The five sources below were read on 1 October 2026. The test methods come from a major independent consumer testing organisation that buys every unit at retail. A second independent testing publication — which also states it buys what it tests, and has evaluated more than 95 robot vacuums — supplies the carpet-fibre, pet-hair and climbing findings. The suction figures come from a specialist guide that publishes its own measurements; it links to retailer listings, so we used the measurements and not its rankings. Anything we worked out rather than read is marked in the text.

• Carpet test method, pet-hair test method, and how difficulty is calibrated by machine type: the published “how we test vacuums” documentation of a major independent consumer testing organisation, which buys every unit at retail.
• Robot-specific test conditions, including the dedicated obstacle space with carpet fringe and cords: the same organisation’s robot vacuum test documentation.
• That the best carpet performer in a large test library can still struggle with fine material worked deep into carpet fibres; that its bristled rollers struggle with pet hair on carpet; and that climbing thick carpet and thresholds is a tested behaviour: an independent testing publication that states it buys every unit it tests and has evaluated more than 95 robot vacuums.
• Static pressure versus airflow, the CFM comparison, sealed suction in kPa, the 16,600 Pa / 2.76 kPa example, and the 40–60% figure for standard mode: a specialist robot vacuum guide that publishes suction measurements alongside manufacturer ratings. That publisher links to retailer listings; we have used its measurements and explanations, not its product rankings.
• Mop-pad lift, carpet sensors, dark-carpet cliff sensor behaviour and climbing limits: the independent testing publication above, reported per model.

One note on a number that circulates in this category: the CFM ranges quoted for corded uprights and robot vacuums are typical figures published by a third party, not manufacturer specifications, and manufacturers generally do not publish airflow at all. Treat the comparison as directionally reliable rather than exact — which is itself part of the argument.