Can an AI agent call a recruiter for you?

What did Yelp just launch, and why should a job seeker care?

On 20 October 2026, Yelp announced Check for Me — an AI agent that phones a restaurant, asks how long the wait is for your party size, and texts you the answer. You tap a button on an eligible restaurant's page or ask inside Yelp Assistant, enter your party size and phone number, and the agent handles the call. It's built for walk-in places, where a listed wait time goes stale in twenty minutes. Yelp says it'll roll out over the coming months on iOS in the US, as part of a fall release carrying more than thirty new features. No firm launch date or Android timing has been announced yet.

Yelp isn't first. Google's Ask for Me started life as a Search Labs experiment in January 2025 that could only ring nail salons and auto shops about pricing and availability. At I/O in May 2026, Google said it would expand calling into select categories — home repair, beauty, pet care — and roll it out across the US over the summer. Its own documentation excludes Indiana, Louisiana, Minnesota, Montana and Nebraska, notes that calls are monitored and recorded, and lets businesses opt out of receiving them entirely. Notice the shape every one of these features shares: a single factual question, asked of a business that publishes that fact anyway.

So why does a restaurant wait-time feature belong in a job-search article? Because you're already running your search through an assistant. You paste the job ad into ChatGPT, ask Claude to tighten a bullet, have Gemini draft the follow-up email. The moment a mainstream app starts dialling phones on your behalf, the next thought is obvious: could it chase the recruiter who went quiet eleven days ago? The honest answer is narrower than the hype — and the gap between what an agent can handle on a call and what you still have to do yourself turns out to be the most useful thing in this whole conversation.

Can an AI agent call a recruiter for you?

An AI agent can place a job-search call for you, but only a narrow, factual one — and no mainstream consumer assistant ships a "call my recruiter" button today. What exists is scoped deliberately: Yelp calls restaurants about wait times, Google calls service businesses about price and availability. Point a general voice-agent platform at a recruiter's number and you can technically make the call happen, but you've stepped outside the guardrails those products were designed with, and you now own every word the agent says. That distinction matters far more than the technology, because recruiters judge the call, not the software behind it. Treat it as a tool with a very small blast radius.

Here's where it genuinely earns its keep: confirming a posting is still open, asking a staffing agency's main line which consultant owns a role, checking the correct name and title of a hiring manager, or finding out where a referral should be sent. Each has one correct answer, needs no judgement, and risks nothing on tone. Disclose that it's an AI, every single time. The FCC ruled in February 2024 that AI-generated voices fall under the TCPA, and while a federal AI-disclosure rule was still only a proposal as of September 2026, state law already fills the gap — and several states require consent before you record anyone.

One more thing before you point an agent at anybody's phone: think hard about where it's running. Agentic tools log what they do, and enterprise security teams increasingly treat them as something to monitor. If your agent lives in a corporate account or on a work laptop, your search leaves a trail worth understanding — read up on what your employer can actually see when an AI agent runs your job search before you automate a thing. Use a personal device, a personal number, a personal account. A status-check call traced back to your current employer's phone system is a much bigger problem than a recruiter who never replied.

Which job-search calls can you hand to an agent?

Hand an agent the calls that have one factual answer and no relationship at stake; keep every call where you're being evaluated, persuaded or remembered. That one line does almost all the sorting you need. "Is this requisition still accepting applications?" is a lookup — anyone with a phone and a screen could answer it. "Walk me through why you left your last role" is an audition, and it's the thing you're actually hired on. Most job seekers get this exactly backwards. They automate the conversation and hand-craft the paperwork, when the machine is dramatically better at the paperwork and dramatically worse at the conversation. Flip it, and your week gets more productive immediately.

Timing still follows the old rules, and an agent doesn't buy you extra attempts. The accepted window is seven to fourteen days after you apply, with email as the first channel and a call only once email and LinkedIn have gone nowhere. Keep it to one touch every seven to ten days, stop after two, and if the posting says no phone calls, that includes your robot. With an agency, contact the recruiter who owns the role rather than ringing the employer directly. An agent that dials three times in a week doesn't read as persistent — it reads as a bot nobody switched off, and that's an easy name to filter out.

Verify the person before you let anything dial them. Scam postings exist to harvest precisely the details an eager candidate volunteers on a call, and a voice agent has no instinct for a question it shouldn't answer. Run the sender and the posting through a proper eight-point check that separates a real recruiter from a scam first. Then cap what the agent may say: no national ID or social security number, no date of birth, no current salary, no bank details, no document copies. Give it a script, one question, and permission to answer anything else with "I'll have the candidate follow up by email."

Why do AI agents fail on screening calls and follow-ups?

AI agents fail on screening calls because a screening call isn't an information transfer — it's an assessment of you, delivered in your voice. A recruiter on a fifteen-minute screen is testing whether you can explain a messy project clearly, whether your dates line up with your CV, and how you handle a question you didn't prepare for. An agent answering on your behalf either recites a script, which lands flat and strange, or improvises about your career, which is how fabrications end up on the record. Neither outcome survives the next round, when a human asks the same question and gets a noticeably different answer.

Then there's the loop nobody asked for. Yelp already sells a voice product to restaurants, so its new feature can put one AI agent on hold waiting for another — press coverage of the launch flagged exactly that risk. Hiring got there first. Recruiter-side voice agents now call applicants within minutes of submission, score the screen and write the disposition straight into the ATS; vendors like Paradox and Phenom are pushing toward running the entire top of funnel, and LinkedIn's Hiring Assistant went generally available in September 2025. If your agent takes their agent's call, nobody in that exchange learns anything about whether you can do the job.

The reputational cost is the part that's hard to undo. Recruitics' report, released on 20 October 2026, found more than one in four talent acquisition leaders at large US employers would block AI agents from applying on a candidate's behalf. That's the mood you'd be calling into. Here's my contrarian take: the phone has quietly become a differentiator precisely because everybody automated it. When a recruiter's inbox is stuffed with near-identical AI-written notes, a ninety-second voicemail in a real human voice — naming the role, giving one specific reason you want it — stands out more now than it did in 2019. Don't delegate your last remaining advantage.

What should you automate instead of the phone call?

Automate the screen you can't talk your way past: the parse. Long before any phone rings, your CV has to survive text extraction, keyword matching and a recruiter's few-second scan — and unlike a conversation, that's a mechanical problem with a mechanical fix. So automate your CV check, your tracker and your job alerts, and keep the calls, the thank-you notes and the negotiation human. The arithmetic is simple. Candidate-side automation raised application volume, volume forced employers to filter harder, and harder filters punish broken formatting and keyword gaps long before anyone forms an opinion about your personality or your phone manner.

The volume numbers explain the squeeze. LinkedIn now takes roughly 11,000 applications a minute, up around 45% in a single year. Greenhouse, hosting about 175,000 live jobs, averages 254 applicants per opening, with applications per recruiter up 412%. Recruitics found recruiters lose close to nine hours a week to applications that turn out unqualified, duplicated or misrepresented. In that environment a two-column template that scrambles in extraction isn't a cosmetic flaw — it's the entire outcome. Run your CV through a free analysis and you'll get a score out of 100, five category scores including ATS compatibility, a layout read, and — paste the job description in — the skills that posting wants and your CV never mentions.

Build the rest of your automation around things that don't need a personality. A tracker holding company, role, date applied, source and next action, so your day-seven and day-fourteen follow-ups fire on schedule instead of whenever you remember. Saved searches and alerts, so you see postings inside the first forty-eight hours while the pool is still small. A reusable answer bank for notice period, salary expectation and reason for leaving, so you stay consistent across every screen. A CV tailored per application rather than one file sprayed at forty roles. None of that needs a voice agent, and all of it moves the number that matters: interviews per ten applications.

Frequently asked questions