Blind threads this month surfaced Meta reinstating in-person final rounds starting in September, and Google now requiring in-person onsites for "local candidates" on certain roles. If you have been running a remote loop strategy since 2022, the ground moved under you. The practical question is no longer how to prep for a Zoom onsite. It is whether your resume signals "local" hard enough to get invited without eating a $1,200 flight.
Why 72% of companies now require an in-person round
The share of employers requiring at least one in-person interview round hit 72% in 2026, up from 24% in 2022. That is a threefold jump in four years, and the mechanism is not RTO politics. It is fraud.
A Gartner survey found 72.4% of recruiting leaders currently conduct interviews in person to combat fraud, defined narrowly: a would-be hire pretending to be someone else, or someone else completing the interview. Google, Cisco, and McKinsey are named adopters. Meta's September pilot is the latest.
The specific numbers driving the reversal:
- 48% of candidates in technical roles showed signs of AI cheating, versus 12% in sales (Fabric's study of 19,368 interviews).
- Unauthorized AI use in interviews rose from 15% to 35% between June and December 2025.
- 81% of Big Tech interviewers said by end of 2025 they believed candidates were using AI to cheat in remote interviews.
- 61% of companies now run detection software during interviews (Greenhouse 2026 AI Hiring Report).
- 17% of hiring managers have encountered candidates using deepfake video.
Compared to 12% in sales. This asymmetry is why the onsite reversal is hitting engineers first.
The tools employers name explicitly are Cluely (Columbia founders suspended for using their own product mid-interview, $5.3M raised, $3M ARR, tagline "cheat on everything"), Final Round AI ($149 to $299 per month, marketed as "100% Invisible & Undetectable"), and InterviewCoder. Pindrop Security has already published a case where a deepfake candidate named "Ivan" was caught because his facial expressions ran slightly out of sync with his words.
What Meta's September pilot actually changes
Meta's September change is a final-round in-person mandate for certain product and infrastructure roles, paired with an AI-enabled coding round where interviewers grade how critically you evaluate the model's output rather than whether you use it.
Read that carefully. The reversal is not anti-AI. It is anti-hidden-AI. The interview is about your judgment layered on top of the AI, not your ability to code without it. Candidates who show up to the onsite and reflexively hide their AI use in the AI-enabled round will underperform candidates who narrate what the model got wrong.
Meta is a top-2 employer of engineers who list "distributed systems" in Refolk's index, alongside Glean and Databricks. Those are the exact profiles Meta is now flying in. Which raises the second question: are you one of the profiles they even find?
The resume location signal that decides your invite
Recruiter search filters at Google and Meta use city radius, not commute time. That is why "San Francisco Bay Area" as a location on your resume outperforms "Fremont" or "Sacramento" even when your actual driving time is identical. The location field is the single highest-leverage line you are ignoring.
In Refolk's index of 346,687 U.S. software engineer profiles, San Francisco is the top region at roughly 16% of the sample, followed by Seattle. When Google flips a role to "local candidates only," it targets that 16%. Everyone else needs the location signal to match, or the search filter drops them before a human ever reads the resume.
Two moves that actually work:
- Widen your location to the metro, not the suburb. "San Francisco Bay Area" or "Greater Seattle Area" instead of a specific town. This is not dishonest. It matches how recruiters query.
- Drop "remote" from your headline unless you are exclusively pursuing distributed companies. In Refolk's index, only 215 U.S. SWE and ML profiles include "remote" in their current-role headline. That is under 0.1%. It sounds like a differentiator. It reads to a Google recruiter as "will not come onsite."
Rewriting a location signal to match a specific posting is fiddly work, and it changes per company. That is the exact tailoring Refolk does automatically: paste the posting, and Refolk rewrites the resume with the location, keyword, and seniority signals that posting actually filters on.
The 4x cheating gap and who pays for it
Junior engineers cheat at roughly twice the rate of seniors, and the 48% versus 12% technical-versus-sales gap means in-person mandates hit new-grad and mid-level engineers hardest. That is also the group least able to eat unreimbursed travel.
Here is the segment view from Refolk's index next to the public cheating and adoption data:
| Segment | Metric | Source |
|---|---|---|
| U.S. Software Engineers, total | 346,687 profiles | Refolk index |
| SWEs in San Francisco (top region) | ~16% of sample | Refolk index |
| SWEs listing "distributed systems" skill | 23,519 profiles (~6.8%) | Refolk index |
| SWE/ML profiles with "remote" in current headline | 215 (~0.06%) | Refolk index |
| Top employers of "distributed systems" SWEs | Meta, Glean, Databricks | Refolk index |
| AI-cheating flag rate, technical vs sales | 48% vs 12% | Fabric / connectingpeople.io |
| Companies running AI-detection software | 61% | Greenhouse 2026 |
| Companies requiring at least one in-person round | 72% (up from 24% in 2022) | sherlock.sh |
If you are a junior engineer weighing a fly-out, the math is bleaker than for a senior. You are more likely to be flagged, more likely to be re-tested in person, and less likely to have the negotiating leverage to demand reimbursement upfront. Ask for it anyway. The worst answer is no.
When to say no to the fly-out
Refusing a fly-out onsite is a much cheaper signal in 2026 than it was in 2022. With 72% of companies requiring at least one in-person round, saying no to one company routes you to the 28% still running remote loops or to companies that reimburse. It no longer eliminates you from the market.
Candidates over-index on FOMO here. Run this checklist before you buy a ticket:
- Is travel reimbursed? Get it in writing. If the recruiter dodges, that is your answer.
- Is this a final round or a "come meet the team" round? A final round with a decision within 5 business days is worth flying for. A meet-and-greet is not.
- Do you have a competing offer or late-stage loop? If yes, use it to negotiate a hybrid loop: one in-person round, rest remote.
- What is the base-salary delta vs your current comp? Under 15%, the expected value of eating $800 to $1,500 in travel is negative.
- Is the role local-only going forward? If yes, factor relocation into your decision now, not after the offer.
The in-person mandate is not anti-AI. It is anti-hidden-AI. The interview grades your judgment on top of the model.
The "verified human" counter-signal
Volunteering for a proctored or in-person round early in the process is now a positive signal, not a burden. With 61% of companies running detection software and detection lagging the tooling, candidates who pre-clear suspicion get moved through faster.
Concretely, when a recruiter reaches out for a phone screen:
- Offer to do the technical screen on camera with screen-share, without asking.
- If you live within driving distance of the office, offer to come in for the first technical round, not the last.
- In your cover letter, name one thing you use AI for in your current job, and one thing you deliberately do not. This signals judgment, not avoidance.
Drafting a cover letter that lands this "verified human plus judgment on AI" note without sounding defensive is fiddly, because the framing has to match the company's own AI stance. It is one of the specific rewrites Refolk handles when it drafts a cover letter for each posting.
The hybrid loop is the new normal
The realistic 2026 shape is one in-person round bolted onto a mostly-remote loop, not a full return to onsite days. RTO-style mandates raise time-to-hire by 23%, senior turnover by 36%, and shrink candidate pools by 340%. Companies know this. They will keep the phone screen and the hiring-manager chat remote, and pull the technical final in-person.
What that means for your application strategy:
- Front-load applications to companies within 90 minutes of you. Local-only filters reward proximity over pedigree at the top of the funnel.
- For dream-tier out-of-market companies, get to a referral first. A referral bypasses the location filter almost entirely because the recruiter is briefed to route around it.
- Score your fit before you apply. If you are a 60% match on a role that requires a fly-out, do not apply. Refolk scores how well a candidate fits each posting before the effort goes in, which is how you avoid $1,200 flights for roles that were never a real match.
The candidates who will do well in the next twelve months are not the ones who master remote-interview cheating tools. Those tools have a half-life measured in weeks, and Meta's September change is the leading indicator. The candidates who do well are the ones whose resume location signal is tight, whose fly-outs are negotiated with reimbursement upfront, and who show up in person ready to narrate what the model got wrong.
FAQ
Should I put "San Francisco Bay Area" on my resume if I live in Sacramento?
Yes, if you can realistically commute or relocate for the role. Recruiter searches use city radius, and "San Francisco Bay Area" matches the queries Google, Meta, and mid-stage startups actually run. Sacramento often falls outside those radius filters even when it is inside your practical commute range. The tradeoff: you need to be ready to answer "when can you be onsite" honestly on the phone screen. If the answer is "never," widening your location misleads the recruiter and wastes both of your time.
Is Meta really requiring in-person interviews for all roles starting September?
No. Based on Blind threads and public reporting, Meta's September change targets specific product and infrastructure roles at the final-round stage, plus an AI-enabled coding round. It is not a company-wide reversal of remote interviewing. Google's parallel move applies to certain teams with "local candidates" language in the JD. Expect this to expand through 2026 as the 48% technical cheating flag rate holds, but treat any recruiter claim of "we still do everything remote" as a role-by-role question, not a company-wide one.
How do I get a company to reimburse fly-out travel?
Ask directly, in writing, before you book anything. The line that works: "Before I put dates on the calendar, can you confirm travel is covered and share the reimbursement policy?" Companies flying in serious final-round candidates almost always cover it. Companies that dodge are either early-stage (fine, negotiate a stipend) or not treating you as a real finalist (bad sign). Junior engineers under-ask here. The 48% cheating flag rate means employers know in-person is now table stakes for trust, and the good ones budget for it.
Does using AI during a remote interview automatically disqualify me?
Not if you are transparent about it. Meta's AI-enabled coding round explicitly grades how critically you assess the model's output. What gets you disqualified is hidden use: Cluely-style invisible overlays, Final Round AI copilots, or someone else typing off-screen. 61% of companies now run detection software, and 17% of hiring managers have caught deepfakes. If a role allows AI, narrate what you are doing. If it does not, do not use it. The middle ground, secretly assisted, is the one shrinking fastest.