Are Technical Interviews Going Back to In-Person?

Partly, and in a specific shape: several large employers have added one in-person round to an otherwise remote loop, usually near the end, rather than returning to a full onsite process. The stated reason is verification — confirming that the person answering is the person who applied and that the fundamentals are actually theirs — not a judgment that remote interviewing failed. Most technical interviews in 2026 are still conducted over video.

That distinction matters for how you prepare. A single in-person round late in a loop changes what you need to be able to do unaided, but it does not invalidate a remote-shaped preparation routine. This article covers what the evidence actually says as of August 2026, which stage the in-person round tends to occupy, and what carries over from remote practice into a room.

What the evidence actually shows

The clearest public statement came from Google. On the Lex Fridman podcast in June 2025, CEO Sundar Pichai said Google would "introduce at least one round of in-person interviews for people, just to make sure the fundamentals are there," as reported by Axios on 12 August 2025. Note the scope: one round, added to an existing process, framed around fundamentals.

Two other named employers moved in the same direction:

• Cisco reinstituted in-person interviews alongside enhanced verification and background checks. Scott McGuckin, Cisco's VP of Global Talent Acquisition, framed the underlying rule directly: "If a candidate is not explicitly invited to use AI during the assessment process, then it should be considered off-limits" (Computerworld, 26 August 2025).
• McKinsey & Company began asking hiring managers to schedule at least one in-person meeting with a candidate before extending an offer, a change that predates the 2025 reporting wave by roughly 18 months (Entrepreneur, 18 August 2025).

On the demand side, the recruiting firm Coda Search/Staffing told the Wall Street Journal that in-person interview requests among its clients rose from 5% to 30% year over year, also relayed in the Axios reporting above. That is one firm's client base, not a market measurement, and should be read as a directional signal rather than a rate.

The broadest figure available comes from Gartner, which Computerworld reported as finding that 72.4% of recruiting leaders are conducting in-person interviews to counter candidate fraud — Gartner's definition of which is a candidate pretending to be someone else or having another person complete an interview for them. Gartner's own July 2025 survey release adds two candidate-side numbers from a 2Q25 survey of 3,000 job seekers: 6% admitted to interview fraud, and 62% said they would be more likely to apply to an organization that requires in-person interviews. Gartner separately predicts that by 2028, one in four candidate profiles worldwide will be fake.

That last figure is worth sitting with. The in-person round is not purely a burden imposed on candidates; a meaningful share of candidates report preferring it, plausibly because it is harder to lose a role to someone who is not who they claim to be.

Why employers moved when they did

The trigger was measurable, and it came from the assessment platforms.

CodeSignal reported on 25 February 2026 that the cheating and fraud attempt rate across its proctored assessments rose from 16% in 2024 to 35% in 2025. In entry-level hiring the rate went from 15% to 40%. Regionally, the company put Asia-Pacific at 48% and North America at 27%. Among flagged 2025 assessments, CodeSignal attributed 35% to frequent off-screen referencing, 23% to unusually linear typing where complex solutions appeared with minimal pauses or debugging, and 15% to elevated similarity with known or leaked answers.

These are vendor-reported detection figures, and they measure attempts that a detection system flagged, not confirmed cheating. A vendor selling proctoring has an obvious interest in a rising number. Treat the trend as credible and the precise level as uncertain.

Even discounted, the direction explains employer behavior. If a third of proctored assessments trip a fraud signal, an assessment score stops functioning as evidence on its own, and the cheapest way to restore confidence in a hiring decision is to put the candidate in a room once before signing. An in-person round is not a detection technology. It is a way of making detection unnecessary for one round.

The countercurrent: the same companies are also opening interviews up

The tidy narrative — employers panicked, so they closed everything down — is wrong, and Google is the clearest counterexample. The same company that added a mandatory in-person round is also piloting a software engineering round in which candidates are expected to use AI, with the resulting prompt transcript reviewed alongside the code.

These two moves are consistent once you see what each is measuring. The in-person round asks whether you personally hold the fundamentals. The AI-permitted round asks whether you can direct a model, check what it returns, and own the result. Neither replaces the other, and a 2026 loop can contain both.

The practical consequence is that "is AI allowed?" now has a per-round answer rather than a per-company one. Our guide to what AI-assisted coding interviews actually grade covers the permitted-use format in detail, and how to verify the AI rules before a technical interview covers the confirmation step. Cisco's rule remains the safe default in the absence of an explicit invitation: tool availability is not permission.

Where the in-person round lands in a loop

Based on how these employers describe the change, the added round tends to have three characteristics:

• Late in the process. McKinsey's version is explicitly before an offer. A company is unlikely to fly candidates in at the screening stage; the cost only makes sense once the pipeline has narrowed.
• Verification-weighted rather than novel. Pichai's framing was "make sure the fundamentals are there" — confirming a signal the loop already produced, not testing something new.
• Additive, not substitutive. Nothing in the public reporting suggests the earlier remote rounds were removed. Expect the same number of remote sessions plus one.

Two conditions make an in-person round more likely for you specifically: a role attached to an office, and a large employer with existing site infrastructure. Robert Half's Q2 2026 analysis of job postings across more than 450 job titles found 87% of postings fully on-site, 10% hybrid, and 3% fully remote — a sharp move from 65% fully on-site in Q4 2025. Those figures describe where the job is performed, not where interviews happen, but the correlation is intuitive: a company hiring for a desk in a building has a low-friction reason to use it once during hiring.

What transfers from remote preparation, and what does not

Most of a well-built remote preparation routine survives the move into a room, because the substance being evaluated has not changed.

Transfers cleanly:

• Fundamentals held in your own head — data structures, complexity reasoning, the language you claim on your resume.
• Verbal structure under questioning. A system design answer that follows a repeatable sequence works identically at a whiteboard; our walkthrough of answering system design questions step by step is format-agnostic on that point.
• Behavioral narratives you can tell without notes.
• Rehearsal against a scoring rubric. Running an AI mock interview with a rubric before an onsite is arguably more valuable than before a remote round, because there is no fallback in the room.

Does not transfer:

• Any live-assistance workflow. This applies to Control as much as to anything else: Control is a desktop application for Windows and macOS built for remote interviews, assessments, and screen-share sessions, and it has no role in an interview room. There is no version of an in-person round where a desktop tool is part of your answer. Anyone telling you otherwise is selling something.
• Second monitors, open reference tabs, and pre-written snippets.
• Typing. A whiteboard or a shared laptop removes the editor, the autocomplete, and often the compiler.
• Recovery time between rounds. An onsite loop is typically back-to-back with a short break, and fatigue in round four is a real variable that remote scheduling usually lets you avoid.

How to prepare for the room specifically

Four adjustments cover most of the gap:

1. Practice without an editor. Write one medium-difficulty solution per day on paper or a whiteboard, then type it in and see what broke. Off-by-one errors and syntax you thought you knew are the usual casualties.
2. Rehearse a full loop in one sitting. Four consecutive 45-minute problem sessions with 10-minute breaks. The purpose is to find out how your explanation quality degrades over three hours, because that is what an onsite measures that a single remote round does not.
3. Practice thinking out loud with a person present. The social pressure of a human standing beside you is different from a webcam. Ask someone to sit in silence and watch you work; it is unnervingly effective preparation.
4. Confirm the logistics as part of the prep. Round count, duration, whiteboard versus laptop, whether a lunch or informal session is evaluative, and whether the same-day debrief exists. Asking a recruiter these questions is normal and gives you the shape of the day.

Where this does not apply

If you are interviewing at a distributed-by-default company, a startup without a local office, or for a role in a different country from the hiring team, an in-person round is unlikely and the trend above is largely irrelevant to you. The same is true for most contract and agency placements.

It is also worth stating plainly: the reporting behind this trend is concentrated in a handful of very large employers, and much of it dates to mid-2025 with 2026 confirmation that it continued. There is no public dataset showing what share of all technical interviews now include an in-person round. The honest summary is that a real, verification-driven shift is underway at large employers, its size is not precisely known, and remote interviewing remains the default nearly everywhere else.

What to do with this

Assume any loop at a large employer with an office near you may include one in-person round late in the process, and prepare so that outcome is unremarkable rather than disruptive. That means holding the fundamentals unaided, practicing without an editor, and building endurance for a multi-hour day. It also means checking the AI policy per round rather than per company, because the same employer now plausibly runs a closed in-person round and an AI-permitted remote one in the same loop.

If you want a concrete next step, run a full timed loop against a rubric before your next onsite and see where your explanations thin out — the AI mock interview walkthrough sets up a version you can repeat weekly.