The two numbers that tell the real story are 74 and 33.5. In the 2024 ResumeBuilder survey, 74 percent of hiring managers said they can identify AI-written CVs. In the TopResume blind test of 600 US hiring managers, actual detection accuracy was 33.5 percent. The gap between confidence and accuracy is where most of the bad advice on this topic lives.
What gets you caught is the visible pattern: phrases like "spearheaded", "leveraged data to inform decisions", "collaborated cross-functionally", "drove strategic initiatives". Bullets that all start with the same verb shape. Numbers that look round and convenient ("increased revenue by 47 percent") but were not in your inputs. Opening lines identical across applicants from the same prompt.
The 49 percent auto-dismiss figure applies to suspected AI output, not to AI-assisted output that has been edited. The 62 percent rejection figure applies to AI CVs lacking personalisation. Both gates are about what the recruiter sees, not what tool wrote it. Edit the output and you walk past both gates.
For the workflow that protects you, see can ChatGPT write your resume?. For the line between "obvious AI" and "good copy", see the comparison in ChatGPT vs purpose-built AI for resume review.