CAT normalisation is minimally affected by the small number of non-aspirants (teachers, curious ex-students) taking the exam - perhaps 500-1000 out of 3 lakh takers. The commenter that_little_witch made the mathematical case: "if 100 percentile is 110 marks, if people like him and coaching sirs attends the exam JUST to understand the exam difficulty, then someone with 109 might get 100 percentile." This effect is real but statistically small at aggregate level.
At IIM Ahmedabad (Rs 27.5L fees, Rs 35.22 LPA avg, 99%+ CAT cutoff) and other top IIMs, cutoff boundaries are where impact happens - marginal scores at 99.
6 vs 99.7 may shift by 1-2 marks. For 98-99 percentile aspirants, the impact is negligible.
The counter-argument by Dark_sun_new: "If the people with 110 marks aren't applying for a seat, how does it affect the one with 109?" - this is partially true. If top scorer doesn't fill a seat, the seat goes to next ranked candidate.
The issue is not seat allocation but score calibration - normalisation uses entire exam distribution. The practical concern for aspirants: your percentile is affected by everyone taking the exam, regardless of their applicant intent. Don't stress this; focus on absolute score improvement.
For fairness: CAT could implement "only applicants" rule but this conflicts with open exam policy. The ethical burden is on non-aspirants to self-reflect on their participation - but that's not your problem as an aspirant. Check your eligibility at collvera.