Every trick you have been told about works until the day it does not, and then it is misconduct.
A high similarity score is usually a citation problem, not a copying problem. Understanding which one you have is the first step, and the similarity report already tells you if you read it properly.
Similarity software reports matched text. It does not report plagiarism — that is a judgement a human makes. Before rewriting anything, go through the sources one at a time:
Two things, and only two: rewriting in your own sentence structure, and citing properly. A real paraphrase changes the syntax, not just the vocabulary. If the sentence still has the same shape with different words in the slots, the software will still match it, and a reader will still recognise it.
Character substitution using Cyrillic look-alikes, white or zero-point text, image-based text, inserted invisible characters, and synonym-spinning tools. Every one of these is detectable, most are flagged automatically now, and all of them convert an honest formatting problem into deliberate misconduct. The penalty for a high similarity score is a revision. The penalty for manipulating a similarity check is usually the degree.
Most Indian universities now have a position on generative AI in theses, and many require a declaration. Assume anything you did not write yourself must be disclosed, and check your own institution's rule rather than a general one. It is a fast-moving area, and "nobody told me" has never been an effective defence.
The number itself is rarely the problem. Not being able to explain where it came from is.
Scope mismatch causes more desk rejections than weak research does.
A review is an argument towards your gap, not a summary of everything you read.
None of them are about your findings. All of them are about your decisions.
Send your topic, your dataset or one draft chapter. We will tell you honestly what it needs — before you pay anything. The first consultation is free.