How to Tell a Real AI Screening Call From a Job Scam

Why Doesn't the Voice Tell You Anything Any More?

Because a machine calling you about a job is now a product vendors sell for first-round screening, not a tell. Alex, formerly Apriora, describes its product as an always-on first-round interviewer that runs a personalised interview by phone call, video or text, and its own page claims 1,000,000 candidates interviewed ¹. That counter is the vendor's, cumulative and unaudited.

A text-message version is sold too. Paradox says its assistant Olivia starts screening a candidate against the job's requirements as soon as a conversation opens, then books an interview for anyone who qualifies ².

Candidates who do not qualify are dispositioned or pointed at other roles, and the application and screening can start over SMS or WhatsApp with no account to create ². That is vendor marketing describing vendor software, and it does not show you the questions or the pass rules.

At least one large employer has named its chat-based system in public. Chipotle announced on 22 October 2024 a Paradox platform fronted by a virtual assistant it called Ava Cado, phased across more than 3,500 restaurants ³. In the company's own words, Ava chats with candidates, collects basic information, schedules interviews and sends offers to the people managers select ³.

A synthetic voice is now evidence of nothing in either direction.

Check What the Call Asks For

The separating question is what the caller wants from you, because a screening call and a scam want different things. A first-round screen asks about the role, your work history, your notice period and when you can speak to a person. It has no use for your bank account, a photograph of your driving licence, or a payment for equipment.

• Ordinary: the role, your experience, your availability, and a time to talk to somebody named.
• Not ordinary: a Social Security number, bank or payment details, or an ID photo, before any offer exists.
• Not ordinary: money you are asked to pay for a laptop, training or a background check, or money sent to you to pass on.
• Worth stopping over: a caller who will not name the company, or who offers only a mobile number and a free email address as the way back.

That list is practical guidance rather than a measured finding. No source in this article counts how often a fake recruiter asks for bank details. What the list does is separate two kinds of request that legitimate hiring does not mix.

Urgency is the other behaviour worth reading. A screening call is arranging a conversation, so it can be arranged again tomorrow, and hanging up to call back costs a real employer nothing.

Pressure to finish before you look anything up is itself the ask.

Not remembering the application is a weaker signal than it feels. People apply in volume, and some use tools that apply on their behalf. Employers do place screening calls to candidates with no memory of applying, and the employer-side guidance on candidates who don't remember applying treats it as a consent gap rather than as fraud.

Failing to place the company name is a reason to check. It is not a verdict.

Verify Through a Channel the Caller Didn't Give You

Verification means going back to the employer by a route the caller had no hand in choosing. Look up the company's careers page yourself, or reply to the recruiter address on the application you actually sent. Ask whether a screening call went out for that role, and say when you can take it. A real employer can wait fifteen minutes for that.

Anything the caller hands you as proof of the caller is not proof.

What you say to a hiring bot can be stored and exposed, which is the practical reason to give a screening call only what a screening call needs. In 2025 the researchers Ian Carroll and Sam Curry reported that McHire, the Paradox-built application platform McDonald's uses, accepted 123456 as both username and password on an owner administration screen ⁴.

A flaw in an internal interface then let a logged-in account read applicants' contact details and chat histories ⁴. The researchers put the exposed dataset at more than 64 million applicants, a figure they inferred from sequential applicant IDs rather than counted ⁴.

Their own timeline has the credentials disabled the same evening and the vendor confirming a fix the next day ⁴. This is a researchers' write-up, not a regulator's finding, and it says nothing about how the chatbot screened anyone.

The same caution covers anything asking for a password. What you hand over when a bot applies from your account goes through the credentials question in full, and no screening call needs a password.

What If It Really Is the Employer's AI?

Then it is a first-round screen, and the useful questions are about the process rather than the caller. Ask which company the call is for, whether it is being recorded, and who reads the transcript afterwards. Those answers cost an employer nothing, and a real screener can give them or route you to somebody who can.

In the European Union there is a rule on whether the caller has to say it is an AI at all. Article 50(1) of the AI Act requires providers to design systems that interact with people so the person is informed they are dealing with an AI, unless that is obvious ⁵. That duty applies from 2 August 2026 ⁵.

The duty sits on the company that built the system rather than on the employer using it. It is a duty to say what you are talking to, and nothing past that: no human alternative, no account of what happens to your answers, no right to opt out.

No US rule this article can point to reaches the voice itself, and when an AI recruiter has to say what it is sets out where the state notice laws land instead.

The transcript question has a partial answer at one vendor, in marketing. Alex's candidate page describes an AI that asks follow-up questions based on your answers, and its testimonials say those answers then go to a human recruiter who reaches out if there is interest ⁶. That page is built from selected testimonials, so read it as candidates' accounts of the flow rather than a guarantee.

Ask anyway, and write down the answer you get.

See how it works >