I personally think all IELTS assessment except one type is unreliable.

AI, instructors, former examiners, examiners -- they are all unreliable, in my humble opinion. All scores by all these people (and machines) are just opinions that don't guarantee anything.

Think about this: If your score is changed after a remark, what does it mean? It means the original assessment -- assessment by a trained, paid examiner -- was wrong. If even this assessment can be unreliable, then how can assessment by people (or machines) who never got examiner training or who got it a decade ago be reliable?

The only reliable IELTS assessment, in my humble opinion, is the assessment by a current senior examiner.

In the face of this overall unreliability -- you are unlikely to have consistent access to senior examiner assessment -- I think it makes more sense to focus on skills, not scores.