Keyword difficulty data is always imprecise and of limited use. It is helpful to filter a list of 100,000 keywords down to 100 interesting ones. It is not a relevant metric to evaluate individual keywords.

No SEO knows how to accurately measure keyword difficulty, so no tool can build a perfect solution. Both SEMrush and ahrefs give a good-enough estimate of KD, which is right more often than it is wrong. Neihter can be trusted blindly.

And unless you have knowledge of the internal workings of these scores from ahrefs and SEMrush, who is to say 20 KD in ahrefs isn’t exactly the same as 50 KD in SEMrush? There are different ways to scale a score from 0 to 100.

Source: I worked on this topic at an ahrefs/SEMrush competitor (Searchmetrics).