Known limitations

We would rather tell you what our data can't do than have you find out. This page is updated when a limitation is fixed or a new one is found. Last verified 30 Sep 2026.

Coverage

• "Robot 500" is the name of the ranking, not a count. We track 240 robots. 158 of them are ranked in 11 categories today; the other 82 are on the watchlist because they aren't commercially available yet or their makers have published too little.
• Some important robots aren't in the database yet. For example, the Agility Digit that works at customer sites today is the generation before the Digit 5 we list; it will get its own record.
• Small categories. A few categories have fewer than 5 ranked robots. We show their scores but no positions.

What the ranking measures

• No hands-on testing. Ranks rest on published specifications and real-world proof (sales, deployments, certifications), not on our own tests. Two robots with similar published figures can perform very differently in use.
• Autonomy and AI quality aren't scored. There is no common, published measure for them yet, so a robot's software can't lift or lower its rank.
• Commercial status favours robots on sale. A robot on sale earns more real-world-proof points than one in a pilot with customers. That is deliberate, but it means a well-known pilot robot can rank below a less famous one you can buy.
• Close scores. Published figures are measured differently by different makers, so gaps of less than 1 point are shown as ties (=2), not as a real difference.
• Several editions in one record. When a maker sells editions with different hardware (for example, with or without dexterous hands), the record lists each edition, and the ranking uses the best published figure for each measure. That can favour robots sold in several editions. Separating editions is a proposal for the next annual review.

Where the data comes from

• Mostly makers' own figures. 93% of our 4,393 sourced figures come from the makers. Makers choose what to publish and how to measure it; we label each figure's source and don't adjust it.
• Status needs an official source. A robot is shown as on sale or in a pilot only when the maker or a named customer says so. A retailer listing or a news report is noted for checking but can't change the status, so a few robots you can find in shops still show no status until the maker confirms.
• Figures go out of date. Prices and availability change often. Every figure shows the date we last checked it, and the linked source is authoritative.
• AI-assisted research. Research and QA are carried out by AI-assisted desks under fixed sourcing rules, with a separate QA pass and human editorial oversight. Mistakes still happen; every correction is published in the corrections log.

The method is new

The Robot 500 launched in September 2026. The current method (v2.0) replaced two launch-week drafts that combined all categories into one list; a single list couldn't be made fair. v2.0 is now fixed until the annual review in January, and proposed changes will be published at least 30 days before they take effect. We are inviting independent robotics researchers to review the criteria.

Help us fix these

If you know a figure is wrong or missing, email corrections@rankedrobot.com with a link to the source. Makers can send new figures to data@rankedrobot.com. Edited by Tilanka Dahanayake, Founder and Editor.