I would also add that I know of a few citizen science databases that have expert verification and where I often encounter them is people uploading observations here that they got the ID from an expert and it was wrong. One platform in particular that I did examine had a more than 10% inaccuracy rate overall and for one of the more challenging species jumped up to a rather impressive 40%.
On iNat you get something that isn’t on those other platforms: every observation is up for review all the time. Mis-ID’s only require one person to see them to get fixed. In other databases, there is often no way to even contact someone to get mistakes re-reviewed, let alone able to run through reviewing the entire dataset.