The independent robotic mower lab

For most gardens between 800 and 2,000 m², our best overall pick for 2026 is the Mammotion Luba 2 AWD 5000 (9.2/10): wireless RTK GPS, AWD for slopes up to 38%, full-featured app. On a tighter budget, the Husqvarna Automower Aspire R4 (8.7/10, £1,000) remains the best value.

For 1,000 m², aim for 90 to 180 minutes of battery and a 22 cm cutting width minimum. Our three picks: Worx Landroid Vision M800 (£1,300, wireless via AI camera), Husqvarna Automower 320 NERA (£1,800, reliability), Stihl iMow 7 EVO (£2,000, 5-year warranty).

Yes, since 2023. RTK GPS models (Mammotion Luba, Husqvarna NERA, Segway Navimow) reach centimetre precision in open conditions. AI-vision models (Worx Vision, Ecovacs Goat) handle wooded gardens better but lose accuracy in low light. Plan 30 minutes of install vs 2–4 hours for a boundary wire.

Most entry-level models tolerate 25 to 30% (14 to 17°). Above 35%, you need AWD (4 driven wheels): Mammotion Luba 2 AWD (38%), Husqvarna 535 AWD (45%). Measure the steepest point with a level — average garden slope is not the right metric.

Plan £450 to £700 for a small garden (<500 m²) with a boundary wire, £1,000 to £1,800 for a family home (500 to 1,500 m²), £2,500 and up for a wireless large-area model or estate. Add £100 to £200 for accessories (shelter, spare blades) and around £70/year of average electricity.

Not mandatory (charging stations are IPX4 rated on every analysed model), but strongly recommended: a shelter doubles shell and battery lifespan by limiting UV exposure and thermal swings. Budget £70 to £220 for a universal-fit shelter.

Robotic wins on time saved (zero weekly chores), lawn quality (denser daily mulching) and noise. The traditional mower stays more cost-effective under 400 m² and remains essential for tall grass, exotic finishes or very complex terrain (stairs, multiple levels).

We buy every robot at retail, never on loan from manufacturers. Affiliate commissions (Amazon, Manomano, Leroy Merlin) are equal across brands and never influence scores or recommendations. Raw datasets and protocols are public — you can check the numbers. No manufacturer sees our verdicts before publication.