Fake recruiter check: 8 signs before you send a CV

What happened when AI agents went undercover to get scammed?

On 14 October 2026, an Australian startup called Apate.ai went public with an unusual tactic: it deploys thousands of AI agents whose entire job is to get scammed. The agents answer scam calls, reply to shady emails and messages, follow phishing links and join sketchy social media groups, posing as people of different ages, genders, accents and languages so the fraudster on the other end stays on the hook. The company grew out of research at Macquarie University in Sydney and says its agents have burned roughly 17 years of scammer time across 2.5 million conversations. Founder Dali Kaafar started down this road after keeping one scam caller busy for 44 minutes himself, then wondering whether the stalling could be automated.

What the agents are fishing for is the detail victims normally only learn far too late: bank account numbers, crypto wallet addresses, burner phone numbers, malicious websites and the brands being impersonated. That intelligence goes to banks, telecoms, utilities and government agencies so the infrastructure can be blocked before the next person picks up the phone. Apate closed an $8.15 million seed round in September 2026 and is pitching this week at Startup Battlefield 200 in San Francisco. Strip away the novelty and the lesson underneath is blunt. Scams are run as an industry, with scripts, staff, call centres and reusable infrastructure — and recruitment is one of the highest-converting scripts in the whole catalogue, because the mark is motivated to hand over documents.

You don't have a fleet of undercover agents, but you can borrow the posture. Those agents work because they assume every inbound contact is hostile until evidence says otherwise, and they give away nothing of real value while they collect it. That's precisely the stance a job seeker needs, because your CV is the single most useful document a fraudster can talk you into sending. It arrives voluntarily, it's formatted for easy reading, and depending on your market it may carry your home address, dates that pin your age, a national ID number, and the names and direct mobile numbers of two referees. Which makes the live question not "is this job any good?" but "who am I actually sending this to?"

Which eight signals separate a real recruiter from a scam?

Eight signals do almost all the work in a fake recruiter check: the sender's domain, whether the role exists on the employer's own careers page, whether the person has a verifiable history, which channel they push you toward, which direction money flows, when they ask for sensitive data, how they behave on camera, and how hard they lean on urgency. Treat the first one as a hard gate. A scammer can copy a logo, a name, a headshot and an email signature in seconds, but they can't send mail from a company's real domain — so free webmail or a near-miss address disqualifies the message on its own, before you read a word of the job description.

Read domains from the right, not the left, because the rightmost part is the real owner: careers.acme.com belongs to Acme, while acme.com.jobs-portal.net belongs to jobs-portal.net. Watch for hyphenated add-ons like acme-careers-secure.com, swapped endings (.co or .ai in place of .com), and single-character homoglyphs that survive a quick glance. Then check the vacancy itself rather than the pitch. Genuine openings nearly always sit on the employer's own careers site with a matching title and requisition number, and you should reach that page by typing the company name into your browser, never by clicking the link you were sent. A domain registered a few weeks ago, with no company registration and no footprint beyond the recruiter messaging you, is the highest-risk combination going.

Money direction is the cleanest test of the eight. Legitimate agencies are paid by the employer, never by the candidate, so any request to fund equipment, training, certification, a visa or a "background check" is fatal — as is the overpayment trick, where an inflated cheque or transfer lands and you're asked to wire back the difference. It genuinely helps to understand how recruiters get paid and whose side they're on, because the incentive model predicts the behaviour. After that, watch sequencing and theatre: ID numbers and bank details requested before any written offer, "interviews" held entirely in chat with cameras off, a face that glitches when the interviewer turns their head, and relentless pressure to sign today.

What can an AI assistant actually verify about a recruiter?

An AI assistant is genuinely good at the mechanical half of a fake recruiter check and genuinely incapable of the human half. Paste a message in and it will pull out every domain, link and phone number, flag lookalike spellings and homoglyph characters, compare the wording against how the company describes its own roles elsewhere, read email headers you supply, and tell you whether a public careers page lists that requisition. What it cannot do is confirm that a human being is who they claim to be. It can't see inside an account, can't date a profile's creation, can't hear that a voice has been cloned, and can't verify anyone's employment. Structure, it handles. Identity, it guesses.

So hand it the mechanical work and keep the judgement for yourself. Three prompts earn their keep. "List every domain, link and phone number in this message and tell me which do not belong to the company's primary domain." "Here are the raw email headers — does the sending domain match the display name, and does it pass authentication?" "Draft a short email to the address published on the company's own careers page asking whether this requisition and this recruiter are real." Expect gaps, too. Most assistants can't fetch LinkedIn, AllJobs or Indeed pages at all, so you'll be pasting text in by hand, and a breezy "this looks legitimate" is a guess wearing a verdict's clothes.

There's a nastier wrinkle now that so many searches run through assistants and agents. Postings and recruiter messages are increasingly written to persuade the machine rather than the person — stuffed with the keywords an agent screens for and the reassurances it's primed to accept. Several scams aimed squarely at AI job-search agents work precisely by getting the agent, not you, to wave the contact through. So never let the assistant's conclusion be the last word, and never let it auto-apply or auto-send your documents. Use it to gather evidence, then make the call yourself and say it out loud in one word: verified, or not verified. Nothing leaves your drive on "probably".

Why shouldn't you send a full CV before those checks pass?

Because a full CV is a ready-made identity kit, which is exactly why fraudsters ask for it so early. Your legal name, mobile number, home address, dates that reveal your age, a decade of employers, plus two referees with their titles and direct lines — that bundle is enough to open accounts, answer knowledge-based security questions, attempt a SIM swap, or social-engineer your old manager with a story that lands because every detail in it is true. The money trail backs this up: reported US losses to job scams climbed from roughly $90 million in 2020 to about $501 million in 2024, and the fourth quarter of 2025 alone generated $150.4 million in losses across some 25,000 reports.

Run disclosure in two stages instead. Stage one, which goes to anyone who clears the domain and careers-page gates, is a lean CV: name, city, a dedicated job-search email address, one phone number, your roles, achievements and skills. Stage two — national ID number, bank details, date of birth, full street address, passport scan, referees' contact details — waits for a written offer from a verified company domain. Market conventions help you here. Israeli CVs carry no photo and no ID number, so a "recruiter" asking for a תעודת זהות number at first contact is already off-script. In India, Aadhaar and PAN belong in payroll onboarding, never in a file you email to someone you met on Monday.

There's a second harm worth naming, and it isn't credit fraud. Harvested CVs get recycled as borrowed identities, with fake candidates using real work histories to interview for remote roles — the pattern laid out in the North Korean remote-job infiltration cases, where eleven governments warned in July 2026 about identity reuse and live deepfake video at scale. Keeping your CV lean protects you on both fronts and costs you nothing in hiring terms, because none of the data you're holding back influences a screening decision anyway. Before you send that lean version anywhere, it's worth running a free CV analysis to see exactly what a parser and a recruiter extract from your file.

How do you run this check in under ten minutes?

The whole routine fits inside ten minutes, and it runs in a fixed order so you never get sentimental about an offer that looks good. Minute one: read the sender's domain character by character, right to left. Minutes two to four: open the employer's careers page yourself — typed into the browser, never clicked from the message — and find the requisition. Minute five: confirm the recruiter is named on the company's own site, or has a profile with history that predates the outreach. Minutes six to eight: ring the company's published main number and ask whether both the role and the person are real. Minutes nine and ten: reread the message for money requests, data requests and manufactured urgency.

Here's where I'll be blunt. If you're still screening on typos and clumsy phrasing, you're calibrated to 2019. Generative text made polish free, so grammar now tells you nothing — and neither does a slick careers portal, a verified-looking badge, a branded onboarding site or a convincing face on a video call, all of which are cheap to fake in 2026. Judge structure instead: domains, requisition numbers, phone numbers you looked up yourself, and the direction in which money and documents are travelling. And accept the trade honestly. I'd rather you lose one real opportunity to excess caution than hand your identity to the one scam in fifty polished enough to pass a casual glance.

Finally, keep a record. A simple tracker listing the company, the recruiter's name, the exact sending domain, the date, what you sent and which verification steps you completed turns a vague unease into evidence — and it makes it obvious when the same fake "talent partner" resurfaces three weeks later under a new company name. Report confirmed fakes to the platform and to your national consumer-protection or cybercrime reporting service, because that's the data that gets the infrastructure taken down. One last thing: if an assistant is running your search, be deliberate about which inbox and file permissions you hand it, since an agent with full mailbox access widens the damage from a single bad click.

Frequently asked questions