How we evaluate robots

What we have and haven't done

We research products from manufacturers' own pages, regulatory filings, recall databases and named independent testers. We have not yet run our own hands-on tests. You won't find claims of our own testing anywhere on this site until we've done it, and when we do, we'll publish how. Where we rely on someone else's testing, we name them and date it.

The 12-question sellability check

Before a product can be recommended, we try to answer:

1. What problem does it actually solve?
2. What's in the box?
3. What are the setup steps?
4. What does it require — Wi-Fi band, hub, phone, subscription, cellular?
5. How long does setup realistically take? (We label estimates as estimates.)
6. Is there usable support — manuals, videos, English-language help?
7. What happens in an emergency? (See emergency labels below.)
8. What's the ongoing maintenance — charging, consumables, updates?
9. What are the strongest reasons not to buy it?
10. Which of those objections can a good guide actually fix?
11. How friendly are checkout, returns and trial periods?
12. Who is it realistically right for?

A product that fails the requirements, support or emergency questions can't be labeled Ready to buy. A product whose setup is beyond an ordinary household is Specialist only and is never recommended to consumers, whatever it pays.

Setup classes

• Ready to buy: an ordinary household can set it up in under an hour or so with the included instructions.
• Guided setup: doable for most people, but it takes planning — mapping a yard, placing sensors, fitting a device.
• Specialist only: needs professional installation, programming or clinical fitting. Not recommended to consumers.
• Emerging: announced, preorder, beta or prototype. Lives on the tracker, never in buying advice.

Availability labels

On sale now means the seller listed it for purchase in the US on the date shown. Stock unclear means the listing showed out-of-stock or contradictory messages. We don't recommend preorders, waitlists or "notify me" products.

Evidence and confidence

• High confidence: read on the manufacturer's own US page on the checked date.
• Medium confidence: price from an official page, with some details from a dated independent review or major retailer — named on the card.
• Conflicting figures are shown as conflicts. When a company's two pages disagree, we say so.
• Company claims (suction numbers, success rates, waitlist sizes) are labeled as company figures.

Every product's source, date and confidence is in our public data file at /data/products.json.

Prices

We publish static, dated prices — never live price widgets. "List" prices are shown only for context, because several brands list prices they rarely charge. Prices change; the date on each card is the day we read it.

The 3-year cost

Price + one-time fees + 36 months of any subscription + required add-ons with published prices + an upkeep estimate (consumables, blades, batteries). Upkeep is our estimate, built from official part prices where published, and labeled that way. Required items without a published price are listed as "price varies."

Emergency labels

Every senior-safety product states who responds:

• A — The person asks for help by button or voice and reaches a trained operator.
• B — Family alerts: a notification to family or caregivers. Never described as emergency dispatch.
• C — Automatic detection: the device detects a fall, and we name exactly who it contacts.

We never present a robot dog or camera robot as fall-detection or emergency hardware unless the configuration you can buy verifiably does that. None currently does.

Shutdown risk

Robots depend on companies. We track products that died, stalled or were orphaned in the Robot Graveyard, flag ownership changes (like iRobot's), and note when a product's core features depend on a company's cloud.

Corrections

When we get something wrong, we fix it and log it in Updates with the date and what changed. We don't silently edit. Send corrections to [email protected].

How we use AI

We use AI tools to help gather sources, organize data and draft pages. A human owner reviews what we publish and is accountable for it. We don't use AI-generated author personas or fake bylines; pages are published under the Robot Uncovered name.

Money

Commissions never decide rankings, and unconfirmed commission rates count as zero when we plan. See How we make money.