A friend of mine did 11 mock interviews before her Google onsite. She felt ready. She bombed the behavioral rounds anyway, because every single one of her mocks was with a close friend who couldn’t bring himself to say “that answer was vague and unconvincing.”
Mock interviews are a good idea in principle. Most people execute them badly. The gap between “I practiced” and “I practiced effectively” is wide, and it explains a lot of otherwise surprising rejections.
The feedback problem
This is the root of most failed prep. Practicing without useful feedback is just rehearsing. You reinforce whatever you’re already doing, good and bad.
Useful feedback has to be specific (“your answer to the conflict question ran 4 minutes, and you never named what the conflict actually was”) and it has to come from someone willing to say uncomfortable things. That’s usually not a friend. That’s not most AI tools either, because they default to encouragement.
Recording yourself is the fastest workaround. Watch the playback. You’ll notice things your friends would never tell you: that you say “like” 47 times in a 3-minute answer, or that you tilt your head down and lose eye contact when you’re uncertain. A recording doesn’t pull punches.
Stranger-to-stranger platforms like Pramp or interviewing.io are the other option. You get someone who has no social obligation to protect your feelings. That friction is useful.
You’re probably only practicing half the interview
Most prep skews heavily toward technical questions: coding, system design, maybe a few leetcode patterns. Behavioral and situational questions get a quick pass the night before, if they get anything at all.
This is backwards at companies that weight culture heavily. According to LinkedIn’s Economic Graph research, hiring managers at top tech companies consistently rank communication and judgment as harder to assess than technical skills, and harder to correct post-hire. Those are the things that sink candidates in final rounds, not algorithm knowledge.
The fix is simple: alternate question types. One session of technical, one session of behavioral, one session mixing both in the order an actual panel might ask them. The switching cost is part of the prep.
Simulating pressure (and why most people skip it)
Here’s the part nobody likes to hear. Your kitchen mock interview, casual clothes, phone nearby, interviewer-friend who lets you restart sentences whenever you want, bears almost no resemblance to the real thing.
The Zoom waiting room sitting in silence. A stranger’s face with the expression that gives you nothing. The question you didn’t see coming. That’s a different cognitive state than “relaxed kitchen run-through.”
A few things actually simulate pressure. Hard time limits with someone calling them out. Interviewing in the clothes you’d wear. Camera on, light correct. No pause-and-restart. The interviewer stays quiet after you finish and says nothing for 5 seconds. That silence is a real thing that happens and it derails people who haven’t felt it before.
I’ll be honest: I’m not certain these environmental mimics fully replicate the anxiety of a real interview. But candidates I’ve talked to consistently say the closer the setup is to the real conditions, the less jarring the actual interview feels.
How many sessions do you actually need?
The honest answer is: it depends on the gap between where you are and where the role requires you to be.
If you interview regularly and you’re targeting a similar-level role, 2 or 3 quality sessions is probably enough to shake off rust. If you’ve been out of the interview market for 3 years or more, double that. If you’re career-changing or interviewing at a level above your current title, assume you’ll need 8 or more sessions before you’re genuinely comfortable.
The thing that matters more than count is what you do with each session. A 30-minute debrief after every mock, reviewing what the answer structure was, where it lost focus, what the follow-up questions exposed, is worth more than two additional sessions done without reflection.
Memorized answers are a trap
There’s a version of prep that creates perfectly structured 3-minute STAR answers for 15 possible behavioral questions. This is tempting because it feels thorough. It fails in two ways.
First, the interviewer asks a variant you didn’t prepare for, and your memorized answers don’t quite fit, and the seams show. Second, interviewers probe. They ask follow-ups. Memorized answers break under follow-up questions because the underlying thinking wasn’t yours, it was a script.
What works better: know your 8 to 10 best stories from your career well enough to deploy them for different question types. Know the shape of the story (context, decision, outcome, what you’d change), not the exact words. This gives you flexibility that scripts don’t.
One thing worth knowing before you book sessions
A decent AI interview copilot gives you a middle ground between recording-yourself-alone and paying $200 per session for a human mock. Craqly runs live mock interview sessions with real-time feedback on your answers, which is useful for getting quick response on phrasing and structure before you graduate to human mocks. It won’t replicate a stranger’s silence or a recruiting coordinator’s poker face. But for the first 5 rounds of deliberate practice, it’s a lower-friction starting point than most people use.
The mistake isn’t doing mock interviews. It’s doing them in a way that confirms what you already know instead of exposing what you don’t.