It's a vital tool for stopping bots from spamming websites, but reCAPTCHA can also be a pain in the neck for users. Over the years, it's been simplified and made less conspicuous for innocent humans, to the point where it should be invisible for most. The reCAPTCHA v3 API is now out of beta and ready to be deployed.
We first heard about the invisible reCAPTCHAs last year, and the v3 beta was launched back in March. While reCAPTCHA v1 required reading and inputting text, and v2 mostly just had a box to be checked (unless it was suspicious, at which point it would give you an image recognition challenge). Version 3 tracks your mouse movements and page interactions to give you a score, but it does this behind the scenes and human users will probably never be made aware of it at all.
For website developers, a new "Action" tag can be used to run a risk analysis on site visitors, and it's recommended that this is added to multiple pages for the best performance. The data produced by this can then be used to give you an overview and help identify which exact pages are most often targeted by bots.
In reCAPTCHA version 3, you get more control over how the information is used. You can set a simple target that needs to be met so that no further action is required, you can combine the score with your own metrics for better results, plus you can use the score to train your machine learning model. This should mean reCAPTCHA v3 gets better over time for the most accurate and frictionless implementation yet. More information can be found in Google's blog post or on the dev site.