It is more than LLMs are faster using those tools, than they are finding more errors by themselves analysing the code.