Robot Vacuums for Carpet and Hardwood: How to Read a Suction Claim

We have tested none of these machines and make no claim about which cleans a rug better. This article reads the fine print the makers attach to their own numbers and names the hardware differences visible on a spec sheet. A model appears only if it is sold in the US, only if the claim quoted came from the maker’s own page opened on October 1, 2026, and only after we searched the CPSC recall database for its brand that day.

Amazon’s search box completes “robot vacuum for carpet and hardwood,” “robot vacuum for carpet and pet hair” and “robot vacuum for carpet deep cleaning” before it offers a brand. The mixed-floor home is the default case, and the one where a single suction number is least useful.

Trend Signal

60/100 Steady momentum

1. Emerging
2. Rising
3. Mainstream
4. Peaking
5. Cooling

Published suction figures passed 30,000 Pa with no shared carpet test behind them

Roborock publishes 22,000 Pa for the Saros Z70 from its own internal testing under IEC 62885-2:2021 clause 5.11, Dreame publishes 12,000 Pa for the X40 Ultra from tests in its own lab, and the Carpet and Rug Institute Seal of Approval database listed 141 certified vacuums on October 1, 2026 with no robotic category among them.

Our editorial reading of the cited data, not a forecast. How it is scored

By the numbers

• 22,000 PaSuction Roborock publishes for the Saros Z70, from testing it ran itselfRoborock
• 141Vacuums certified in the Carpet and Rug Institute Seal of Approval database, none roboticCarpet and Rug Institute
• $2.21BSharkNinja cleaning appliance sales in 2025, growth credited to extraction and cordlessSEC filing
• NoneFloor-cleaning robot vacuums in the CPSC recall database, our count October 1, 2026CPSC data, our count

The suction number and what sits behind it

Pascals measure a pressure difference at the appliance, not dirt removed from a carpet, and there is no common US test that every maker runs before printing the figure. The footnotes alone establish that. Roborock’s Saros Z70 page carries 22,000 Pa with the note that it is “based on internal testing carried out by the manufacturer according to IEC 62885-2:2021/5.11 standards, using Max+ mode with a full battery.” Dreame’s X40 Ultra page carries 12,000 Pa with the note “based on tests in Dreame Lab. Actual performance may vary depending on usage.” The eufy Omni E35 page calls its 35,000 Pa a peak figure and carries a line saying testing was by TÜV SÜD under laboratory conditions.

Three makers, three bases, one unit. The figure does not even order one maker’s own range: on the Dreame listing we read on October 1, 2026, the X Series X50 Ultra is tagged 20 kPa and the L Series L60 Pro Ultra 35 kPa, while the two X Series models above them, including the one the page badges Flagship, carry no pascal tag at all.

The two brands with the longest US history skip the unit. iRobot advertises “4X more power-lifting suction” on the Roomba Max 715 and footnotes that claim twice on one page, once against the Roomba 105 and once against the Roomba Max 705. Shark advertises “4X stronger pickup” on the Navigator 2-in-1 and footnotes it “in Eco mode vs. Normal mode,” the same robot on a different setting.

Model	Published figure	The basis the maker states
Roborock Saros Z70	22,000 Pa	Internal testing by the manufacturer under IEC 62885-2:2021 clause 5.11, Max+ mode, full battery
Dreame X40 Ultra	12,000 Pa	Based on tests in Dreame Lab; actual performance may vary with usage
Anker eufy Omni E35	35,000 Pa peak	Page labels it peak suction and carries a TÜV SÜD laboratory testing note
iRobot Roomba Max 715	No pascal figure	4X more suction, footnoted on the same page both against the Roomba 105 and against the Roomba Max 705
Shark Navigator 2-in-1	No pascal figure	4X stronger pickup, footnoted in Eco mode versus Normal mode

A pascal figure is comparable to itself and to nothing else.

What carpet certification covers, and what it skips

There is a third-party carpet program in the US and robot vacuums are not in it. The Carpet and Rug Institute runs the Seal of Approval scheme, which says certified products “effectively remove soil without adversely affecting the appearance or the performance of the carpet.” Its certified vacuum database listed 141 certifications when we pulled it on October 1, 2026, and the vacuum type filter offers only backpack, canister, rider, upright and wide area.

The listed companies are commercial floorcare names such as Nilfisk, Karcher, ProTeam, Sanitaire and Tennant. No robot vacuum is listed. That is not a verdict on robots; it means the one independent carpet-cleaning seal a US shopper can look up has nothing to say about them, so the pascal figure stands in for a certification that does not exist here.

The hardware that actually differs on carpet

Published specs are most useful on the mechanics, which are described rather than scored. Automatic carpet detection is common and named: iRobot calls it Carpet Boost and says it raises suction on carpet or a rug; Shark calls it Floor Detect and says it senses floor type and adjusts suction to keep carpets dry by vacuuming only.

Mop handling is the other carpet-specific difference, and the one that causes damage if it is wrong. Dreame publishes a 10.5-millimeter height difference on the X40 Ultra between mop down and mop lifted, and says medium and high-pile carpets can be flagged in the app. Roborock publishes a liftable FlexiArm mop on the Saros Z70 plus app-enabled automatic mop removal. Dreame’s Aqua10 Ultra Roller listing describes an AutoSeal that closes on carpet and lifts the roller 0.55 inch.

Brush material is published plainly too: Roborock describes dual all-rubber brushes on the Saros Z70, iRobot patented anti-tangle Dual Rubber Brushes that flex across floor types. If the floor is still being chosen, where rigid vinyl plank went in 2026 covers the hard-surface side.

Carpet detection, mop lift height and brush material are published facts; pascal figures are self-reported numbers.

What the sales mix and the recall record say

SharkNinja is the one large floorcare company that files with the SEC. Its Cleaning Appliances line, which it defines as including handheld and robotic vacuums alongside steam mops and carpet extraction, reported $2,205.8 million of net sales in 2025, up 6.9%. The 10-K filed on March 2, 2026 credits that to “the carpet extraction and cordless vacuums sub-categories.” The 10-Q filed on August 5, 2026 repeats it for the first half of 2026: $1,038.6 million, up 10.1%, driven again by the carpet extractor and corded and cordless vacuums sub-categories.

Robots are not what is pulling that line, which is worth knowing before a robot replaces a plug-in machine.

On safety, our October 1, 2026 search of the CPSC recall database returned no recall of any household floor-cleaning robot vacuum under any brand we checked, including every brand named in this article. The five household-robot recalls since 2022 are pool robots and lawn mowers, all battery, adapter or dock-circuit hazards.

What to check before you buy for a mixed-floor home

Start with the mop, because that is where carpet goes wrong. Confirm the robot lifts or removes the pad and that the lift height is published, then confirm carpet detection is automatic rather than a mode you set per room.

Treat filtration as a separate question from suction, since the bin and base filter decide what goes back into the air; our explainers on HEPA labeling and how MERV numbers work apply here. If the concern is dander rather than grit, our home grooming piece belongs in this decision.

What to watch next

• Whether any maker publishes a suction figure verified by a named third-party laboratory rather than its own test
• Whether the Carpet and Rug Institute adds a robotic category to the Seal of Approval vacuum database at carpet-rug.org
• Whether SharkNinja filings ever credit cleaning growth to robots rather than extractors and cordless vacuums

Where to start

Paid links: we earn a commission if you buy through the buttons below, at no extra cost to you. As an Amazon Associate I earn from qualifying purchases.

More: Smart Home trends and how the Trend Signal is scored.

Questions readers ask

Is a higher Pa number a better robot vacuum for carpet?

Not demonstrably. Roborock attributes 22,000 Pa on the Saros Z70 to internal testing under IEC 62885-2:2021 clause 5.11 in Max+ mode, Dreame attributes 12,000 Pa on the X40 Ultra to tests in its own lab, and the eufy Omni E35 page calls 35,000 Pa a peak figure. Different bases and modes are not comparable.

Is there an independent carpet test for robot vacuums in the US?

Not one a shopper can look up. The Carpet and Rug Institute Seal of Approval database listed 141 certified vacuums on October 1, 2026, and its vacuum type filter offers only backpack, canister, rider, upright and wide area. No robotic category exists and no robot vacuum is listed.

Why do iRobot and Shark not publish a pascal rating?

Both compare with their own products instead. iRobot’s Roomba Max 715 page claims 4X more suction, footnoted against the Roomba 105 in one place and the Roomba Max 705 in another. Shark’s Navigator 2-in-1 page claims 4X stronger pickup, footnoted as Eco mode versus Normal mode on the same machine.

Will a robot damage carpet with its mop?

That depends on whether the pad lifts or detaches, and how far. Dreame publishes a 10.5-millimeter difference on the X40 Ultra between mop down and mop lifted and says carpets can be flagged in the app, Roborock publishes a liftable mop and app-enabled mop removal on the Saros Z70, and Dreame cites a 0.55-inch roller lift on the Aqua10 Ultra Roller.