After four years of remote-by-default hiring, the onsite interview has returned. Google reinstated at least one in-person round for technical hires. McKinsey now asks hiring managers to meet candidates face to face before an offer goes out. Cisco and others have made similar moves. A Gartner survey found that 72.4% of recruiting leaders reported conducting interviews in person to combat fraud.
The reason is not nostalgia for whiteboards or a quiet push to get people back into offices. It is that remote interviews stopped reliably telling employers what they needed to know, because a meaningful share of candidates were being fed answers in real time.
This piece covers what actually drove the change, what an in-person round tests that a video call does not, and how to prepare for one if you have spent your entire career interviewing through a webcam.
Key takeaways
- The trigger was AI-assisted cheating at scale. CodeSignal, analysing millions of proctored technical assessments, reported fraud attempt rates roughly doubling from 16% in 2024 to 35% in 2025, with the sharpest rise in the second half of the year.
- 72.4% of recruiting leaders told Gartner they are conducting in-person interviews to counter fraud.
- This is a filter, not a punishment. In-person rounds are cheap to add and they eliminate the entire category of real-time assistance at once, which is why they spread so fast.
- What onsite actually tests is recall under load: whether the knowledge is in your head, and whether you can think out loud while someone watches.
- The preparation that transfers is spoken, not written. If your practice consists of reading and typing, an onsite round will feel like a different sport.
- The upside is real. Face to face rounds are harder to fake and easier to be memorable in. Candidates with genuine depth are advantaged by this shift.
What actually happened
The remote interview had a structural weakness that took a few years to become obvious. When the entire assessment happens through a screen, the employer cannot see the rest of the room. That was tolerable when the worst case was a candidate glancing at notes. It stopped being tolerable when a second monitor could listen to the question and produce a polished, structured answer in about two seconds.
The scale is what forced the change. CodeSignal's data on proctored technical assessments showed cheating and fraud attempt rates rising from around 16% in 2024 to 35% in 2025. Separate interview-integrity analyses through late 2025 and early 2026 reported flag rates in a similar range, with technical roles at the high end.
Whatever the exact number, hiring teams reached the same conclusion at roughly the same time: they could no longer tell, from a video call alone, whether a strong answer meant a strong candidate.
Two responses were available. Build detection, or remove the opportunity. Detection tools exist and companies do use them, but they are imperfect, they generate false positives that damage good candidates, and they turn hiring into an arms race. Adding one in-person round removes the entire class of problem in a single step, which is why so many large employers arrived there at once.
The uncomfortable part for honest candidates
If you never cheated, this change costs you something anyway. Travel, time off, the expense of getting to an office, and the return of a format that disadvantages anyone for whom the logistics are hard: caregivers, people with disabilities, candidates who are not local, and anyone interviewing while employed full time.
That cost is real and worth naming. The industry has quietly shifted a burden onto every candidate in order to solve a problem created by a minority of them. If it feels unfair, it is, at least in part.
What it also does, though, is restore the value of actually knowing things. For the last few years the ceiling on a remote interview was partly set by who had the better tooling. In a room, that variable disappears. If you have genuine depth, this shift is in your favour, and it is worth approaching it that way rather than as an obstacle.
What an in-person round actually tests
It is not a harder version of the same interview. The evaluation changes shape.
Recall without retrieval. On a video call you can keep a document open. In a room you cannot. What you actually remember is what you have.
Thinking out loud under observation. The most common onsite failure is not being wrong. It is going silent. A remote interviewer sees a frozen video tile. An in-person interviewer watches you sit in silence for forty seconds, which feels vastly longer to you than it does to them, and which they read as either careful thought or panic depending entirely on whether you narrate it.
Follow-up depth. In-person conversations naturally go three or four questions deep on a single topic, because the rhythm is faster and there is no lag. This is the specific thing that surfaces borrowed answers. A prepared answer survives the first question. It rarely survives the fourth.
Ordinary human presence. Whether you make eye contact, whether you listen or wait to talk, whether you are someone the team wants in the room. This was always part of hiring and it partly disappeared on video.
How to prepare
The preparation that works is different from remote preparation in one specific way: it has to be spoken, and it has to be interrupted.
Practice out loud, standing up, without notes. Reading your story bank silently builds recognition, not recall. Saying it out loud with nothing in front of you is the only thing that tests whether it is actually in your head. This is uncomfortable and that discomfort is the point.
Build a story bank you can access without prompts. Four to six stories, each covering a different competency, each with the situation compressed and the result specific. Our guide on building a story bank covers the structure. For onsite specifically, the test is whether you can start any of them within two seconds of a question landing.
Rehearse the fourth question, not the first. For each story, have someone ask you "why did you do it that way?", then "what would you do differently?", then "what did that cost?". Depth is what in-person rounds are for, and it is where prepared answers run out.
Learn to narrate your silence. "Let me think about that for a second" buys you all the time you need and reads as confidence. Silence without narration reads as being stuck. This one sentence is probably the highest-leverage thing on this list.
Practice on a whiteboard if the round is technical. Writing code by hand is a genuinely different motor skill. Syntax you type fluently disappears when there is no autocomplete. Do it badly a few times in private rather than for the first time in front of a panel.
Handle the nerves deliberately. In-person rounds bring back physical symptoms that video calls muted: the walk in, the handshake, the room. Our guide on calming interview nerves covers what actually works, which is mostly preparation and breathing rather than positive thinking.
What this means for AI in your preparation
There is an obvious tension in an article on this site: we build an AI interview tool, and I am telling you the industry is moving to in-person rounds specifically to keep AI out of interviews.
Both things are consistent, and the line is the same one we have written about before in using AI during a live interview. Using AI to prepare is ordinary practice, no different from a mock interview with a friend or a coach. Using AI during the interview is misrepresenting your ability to someone making a decision about you.
The shift to in-person rounds actually makes preparation more valuable, not less. When you cannot lean on anything during the interview, everything depends on what you did beforehand. The candidates who struggle in a room are the ones whose preparation consisted of reading. The candidates who do well practiced out loud, got interrupted, and were made to defend an answer.
That is the entire argument for rehearsal, and this shift has strengthened it.
FAQ
Are all companies bringing back in-person interviews?
No. The change is concentrated among large employers and roles where the cost of a bad hire is high, particularly technical and consulting roles. Gartner found 72.4% of recruiting leaders conducting in-person interviews to combat fraud, but that typically means adding one round to an otherwise remote process rather than returning to fully onsite loops.
Will the whole process be in person now?
Usually not. The common pattern is a remote recruiter screen and remote early rounds, with one in-person round added before an offer. Google's reinstated round is for technical hires, and McKinsey's requirement is one in-person meeting before an offer goes out. Budget for one trip, not five.
Does this mean AI interview prep is now pointless?
The opposite. When you cannot use anything during the interview, preparation is the only variable you control. What changes is what good preparation looks like: spoken rehearsal with follow-up questions, rather than reading model answers.
What if I cannot travel to an in-person round?
Ask. Many companies will accommodate a genuine constraint, particularly for candidates who are not local, and some will substitute a proctored or supervised remote round. The ask is normal and asking it early is better than discovering the constraint late.
How do I practice the "thinking out loud" part?
Record yourself answering a question you have not seen before, and force yourself to narrate every pause. Play it back and count the silences longer than three seconds. Most people are shocked by how many there are, and the fix is a single sentence: say what you are thinking about rather than thinking about it quietly.
Practice the round you are actually walking into
The specific thing an in-person interview tests is whether your answers survive being questioned, out loud, without notes.
Round Zero runs a live mock interview by voice that asks follow-up questions when an answer is thin, then scores the transcript and quotes your own words back to you so you can see exactly where the depth ran out. It is the closest thing to being pushed in a room, and the free tier includes one full scored session. Practice before, so you do not need anything during.