On low pile, yes. On deep pile, no machine here publishes figures that support it. The line between those two is where most disappointment in this category comes from.
Everyone in this category says they tested twenty vacuums. We have not tested any, and we say so. What we do instead is read the manuals, compile the published specifications, and work out what each machine costs to run for three years — the method is written out in full.
We earn a commission if you buy through the links on this page, at no extra cost to you. It never changes what we recommend or the order we recommend it in. Full disclosure.
Low pile: yes. Medium pile: adequately, for surface debris. Deep pile: no. Carpet cleaning needs agitation to loosen embedded grit and airflow to carry it away. Robots publish suction figures and, at best, an anti-tangle brush — none publish the figures that would support deep-pile extraction, and the physics of a 7-pound machine does not favor it.
Why carpet is hard for a robot specifically
Dirt on a hard floor sits on top and only needs to be picked up. Dirt in carpet is held between fibers, and getting it out takes two things: a brushroll agitating the pile to release it, and enough airflow to carry it into the bin before it settles back.
A full-size upright brings 800 to 1,200 published watts and a large driven brushroll to that job. A robot brings a battery, a small brush and, at best, a manufacturer's pascal claim — which, as covered throughout this site, is not comparable between brands and is not an airflow measurement at all.
The second problem is the robot's own weight. Deep pile compresses under a heavier machine and resists a lighter one. A 6 to 10 lb robot floats over thick pile rather than pressing into it, so the brush never reaches the fiber base where the grit is.
What the published specs support, pile by pile
Read against each manufacturer's own published surface recommendation.
Pile type
What a robot does
What it does not
Low pile and flat-weave rugs
Picks up surface hair, dust and crumbs well. This is a genuine use case.
Nothing much is left behind at this pile depth.
Medium pile
Surface debris, and daily hair maintenance if run frequently.
Embedded grit stays in the pile.
Deep and shag pile
Struggles to move across it. Several manufacturers publish 'medium pile' as their upper limit.
Cleaning, in any meaningful sense. Some robots will not traverse it at all.
eufy publishes the 11S MAX's surface recommendation as 'Hard Floor & Tile & Medium carpet'. Lefant publishes the M210's as 'Wood, Hard Floor and Low Pile Carpet'. These limits are the manufacturers' own.
The features that genuinely help on carpet
Carpet boost. Suction increases when carpet is detected. Lefant publishes it on the M2 Plus: carpet detection with auto boost when the mop holder is removed.
A real brushroll, and an anti-tangle one. Roborock publishes dual anti-tangle brushes — a JawScrapers main brush plus a zero-tangle side brush — on the Q7 M5+. Both matter: agitation for cleaning, anti-tangle for not stopping every week with scissors.
Multiple passes. Shark's Matrix Clean on the AV2501AE cleans in a grid with multiple passes over the same area, which partly compensates for lower per-pass agitation.
Higher published suction, within one brand. Roborock's 10,000 Pa Q7 M5+ against its own 8,000 Pa Q7 L5 is a meaningful internal comparison. Across brands it is not.
The realistic setup for a carpeted home
A robot for daily maintenance, a proper upright for the weekly clean. The robot keeps surface hair and dust from accumulating — which is genuinely most of what makes carpet look tired — and the upright does the deep work. That combination costs less than one flagship robot and it actually cleans the carpet.
If your carpet needs more than that, the tool is not a vacuum at all. A carpet extractor is what removes set stains and ground-in grime, and no robot is a substitute for one.
#
Product
Best for
Score
Price
01
Roborock Q7 M5+ Robot Vacuum and Mop10,000 Pa published and a 2.7-liter sealed dock bag rated at seven to nine weeks: the pet-hair answer, at a real price.
ILIFE A30 Pro Robot Vacuum and Mop10,000 Pa and a five-bag box that ILIFE rates at up to 280 days of hands-free running — the longest published interval on the site.
Scores are ours, from the published rubric on the How We Pick page. They are editorial judgments about published specifications — not test results. We have not handled these machines.
01 · Best self-empty for pet hair
Roborock Q7 M5+ Robot Vacuum and Mop
10,000 Pa published and a 2.7-liter sealed dock bag rated at seven to nine weeks: the pet-hair answer, at a real price.
JawScrapers main brush plus zero-tangle side brush
Mopping
Three adjustable water levels, simultaneous with vacuuming
Rated noise
65 dB
Every figure above is the manufacturer's own published number. Where a manufacturer publishes nothing, the row says "Not published" rather than an estimate.
What the numbers say
+Highest published suction figure on the site, at 10,000 Pa
+2.7-liter dock bag is larger than the iRobot's and rated for a comparable interval
+Anti-tangle on both brushes, which is the pet-hair spec that matters
+150 minutes plus recharge-and-resume covers a large floor
Where it stops
−Dock bags are a permanent consumable
−The dock is large and 65 dB when emptying is genuinely startling the first time
−Pa figures remain incomparable between brands, however impressive 10,000 sounds
Who it is for
Pet households with real square footage, where the on-board bin is the daily annoyance.
Skip it if
You live in a studio. This is a whole-floor machine with a dock the size of a wastebasket and you will resent the footprint.
Our score: 8.5 out of 10
Published suction9.5
Navigation9.0
Runtime8.5
Bin and dock9.5
Running cost6.0
Scored against the published rubric on How We Pick. These are judgments about spec sheets, not measurements.
02 · Best mapping under $300
iRobot Roomba 105 Vac
LiDAR mapping, keep-out zones and a 200-minute rating from the brand that will still be supporting the app in five years.
70x more power-lifting suction, compared by the manufacturer against the Roomba 600 series
Control
Roomba Home app, on-device buttons, voice assistants
Wi-Fi
2.4 GHz required for setup and operation
Suction in Pa
Not published
Every figure above is the manufacturer's own published number. Where a manufacturer publishes nothing, the row says "Not published" rather than an estimate.
What the numbers say
+200 minutes rated, the longest published runtime of any robot on this site
+Real LiDAR mapping, keep-out zones and room targeting at a budget price
+Recharge-and-resume, so a large floor completes without you
+iRobot's app and parts supply are the most durable bet in the category
Where it stops
−The 70x suction claim is measured against iRobot's own decade-old 600 series, not against a competitor
−No self-empty dock in this configuration — you empty the on-board bin
−No published bin volume
Who it is for
Multi-room homes that want scheduled, mapped cleaning without a self-empty dock's price.
Skip it if
Nobody in the house will install an app. Half of what you are paying for is mapping, zones and scheduling, and none of that exists without the phone.
Our score: 8.0 out of 10
Published suction6.5
Navigation9.0
Runtime10.0
Bin and dock6.0
Running cost7.5
Scored against the published rubric on How We Pick. These are judgments about spec sheets, not measurements.
03 · Longest dock interval
ILIFE A30 Pro Robot Vacuum and Mop
10,000 Pa and a five-bag box that ILIFE rates at up to 280 days of hands-free running — the longest published interval on the site.
2.5 liters, five in the box, rated up to 280 days hands-free in total
Tank
2-in-1 integrated: 200 ml dustbin plus 200 ml water tank
Navigation
LDS LiDAR with SLAM mapping, no-go zones and virtual walls
Weight
10.4 lb
Rated noise
72 dB
Every figure above is the manufacturer's own published number. Where a manufacturer publishes nothing, the row says "Not published" rather than an estimate.
What the numbers say
+Five bags in the box is real money that the competition charges separately for
+Full LiDAR SLAM mapping with virtual walls at a budget price
+10,000 Pa published, matching the dearest robot here
Where it stops
−72 dB is the loudest robot figure on the site
−200 ml dustbin is the smallest here — the 2-in-1 tank splits capacity with the water
−The 150-minute rating is explicitly the gentle-suction number
Who it is for
Buyers optimizing for the longest possible gap between touching the thing.
Skip it if
You want to run it while you are home. At 72 dB it is a run-it-when-you-are-out machine.
Our score: 8.2 out of 10
Published suction9.0
Navigation8.5
Runtime8.0
Bin and dock8.5
Running cost7.0
Scored against the published rubric on How We Pick. These are judgments about spec sheets, not measurements.
04 · Best under $150
eufy 11S MAX Robot Vacuum
No app, no LiDAR, 100 minutes and 55 dB — the cheapest robot on this site and the one with the fewest things to go wrong.
Infrared obstacle sensing and drop sensors — no mapping
Connectivity
None. Manufacturer states no Wi-Fi and no app; remote control and on-device buttons only
Filter
Washable
Suction in Pa
Not published
Every figure above is the manufacturer's own published number. Where a manufacturer publishes nothing, the row says "Not published" rather than an estimate.
What the numbers say
+600 ml bin, the largest on-robot bin here, which matters when there is no self-empty dock
+55 dB rated, quiet enough to run while you are in the room
+2.85 inches tall, so it fits under more furniture than most
+No app means no account, no firmware updates, and nothing to stop working when a server retires
Where it stops
−No mapping — it cleans in a semi-random pattern and takes longer to cover a room
−No no-go zones, no room targeting, no scheduling from a phone
−Single side brush
−No published suction figure at all
Who it is for
One-floor apartments, mostly hard floor, where the job is daily dust and pet hair rather than a scheduled whole-house clean.
Skip it if
You have more than about 800 square feet or several rooms. Without mapping it will miss areas, and you will not be able to tell it not to.
Our score: 6.6 out of 10
Published suction4.5
Navigation3.5
Runtime9.0
Bin and dock7.0
Running cost9.0
Scored against the published rubric on How We Pick. These are judgments about spec sheets, not measurements.
05 · Quietest robot here
Lefant M210 Robot Vacuum
52 dB and 2.99 inches tall, with a brushless suction port that sidesteps hair tangle by not having a brushroll at all.
Up to 120 minutes in low suction mode, manufacturer-stated
Height
2.99 inches
Diameter
11 inches
Dustbin
500 ml
Rated noise
52 dB
Suction port
Brushless — no roller brush
Navigation
FreeMove 3.0, 6D infrared anti-collision — no LiDAR mapping
Battery
Lithium iron phosphate
Suction in Pa
Not published
Every figure above is the manufacturer's own published number. Where a manufacturer publishes nothing, the row says "Not published" rather than an estimate.
What the numbers say
+Brushless port cannot tangle, because there is nothing to tangle on — the only true fix for long hair
+52 dB is the quietest published figure on this site
+11-inch diameter gets between chair legs a 13-inch robot cannot
+Lithium iron phosphate chemistry is more cycle-durable than standard lithium-ion
Where it stops
−No brushroll also means no agitation, so carpet pickup is weak by design
−The 120-minute figure is explicitly the low-suction number
−No LiDAR mapping
Who it is for
Hard-floor homes with long-haired occupants or pets, and anyone who has cut hair off a robot brushroll once too often.
Skip it if
You have carpet you actually want cleaned. Without a brushroll it lifts surface debris and leaves embedded grit; this is a hard-floor machine.
Our score: 6.6 out of 10
Published suction4.0
Navigation4.0
Runtime9.0
Bin and dock6.5
Running cost9.5
Scored against the published rubric on How We Pick. These are judgments about spec sheets, not measurements.
What we cannot tell you
How well any specific robot performs on your specific carpet. Pile depth, density and fiber type all matter enormously, no manufacturer publishes a carpet pickup test, and pascal figures cannot be compared between brands. What we can do is report each manufacturer's own stated surface recommendation — which is the most honest guide available, and which several of them state more conservatively than their marketing implies.
Common questions
Do robot vacuums work on carpet?+−
On low pile, yes. On medium pile they handle surface debris. On deep or shag pile, no — several manufacturers publish 'medium carpet' as their own upper limit, and no robot here publishes figures supporting embedded-grit removal from thick pile.
Which robot vacuum is best for carpet?+−
Look for a real brushroll with an anti-tangle design plus carpet boost. Roborock's Q7 M5+ publishes dual anti-tangle brushes and its highest suction figure; Shark's AV2501AE publishes Matrix Clean multi-pass coverage, which partly compensates for lower per-pass agitation.
Can a robot vacuum replace an upright on carpet?+−
No. The published gap — 800 to 1,200 W of upright motor with a large driven brushroll, against a battery-powered robot with a small brush — is too large. The realistic setup is a robot for daily maintenance and an upright for the weekly clean.
Why do brushless robot vacuums struggle on carpet?+−
Because a brushless design has no agitation by construction. Lefant's M210 suction port is a genuinely good answer to hair tangle, and the same feature means nothing disturbs the pile, so embedded grit stays where it is.
Does higher suction help on carpet?+−
Within one brand's own range, yes — Roborock's 10,000 Pa machine against its own 8,000 Pa machine is a real comparison. Across brands it is not, because pascals are measured differently by every manufacturer and none publish the method.
Sources
Manufacturer-published specifications — Every runtime, wattage, capacity, weight, filtration and warranty figure quoted on this page is the manufacturer's own published number for that model, retrieved September 1, 2026. Where a manufacturer publishes no figure, the page says 'Not published' rather than estimating one.
Sweep & Suction — How We Pick — Our published method, including the scoring rubric, the three-year running-cost formula and an explicit statement of what published specifications cannot tell you.
