I think Denis' suggestion is the way to go. In a matrix question you would add a "3rd dimension" to the concept and this will be probably confusing and hard to understand for respondents.

You could still use the matrix format, but use only one subquestion per matrix (so 3 matrix questions with 1 subquesiton each).

Then in the header you use the scale and below the scale in smaller with a line break the description.

This would be a mix between Denis' suggestion and the matrix. But your description texts are fairly long, not sure how good it would look in the scale header (answer options) of a matrix.

Trying to force a paper&pencil questionnaire 1:1 into online almost never works well and the impact of changing this a bit in my experience (done a couple of those methodology tests) is basically zero. We could almost never notice a difference in response when we tested. Might be different for you, so maybe run a methodology check, where you apply RANDOMLY the paper version or the online version and then compare the two samples and if there are signifcant differences between the answers.