AI Recruiting Chatbots: What Works, What Candidates Hate

Somebody just pitched you a recruiting chatbot, and the deck had a number in it you can't quite believe. That's a reasonable place to start, because the category is genuinely split down the middle.

Two recent, credible surveys looked at how candidates feel about AI in hiring and came back with what look like opposite answers. One found candidates walking away in droves. The other found 77% of them happy with the experience. Both are right. They measured two different things wearing the same "AI" label.

This piece separates the two, shows what the named-client results actually say once you label them honestly, and gives you a straight checklist for telling the difference before you sign anything.

What an AI Recruiting Chatbot Actually Does

Strip away the vendor language and a recruiting chatbot does three jobs: it answers questions a candidate would otherwise email or call about, it screens against a short list of knockout criteria, and it schedules the interview once both sides are interested. Everything else — voice agents, SMS threads, web widgets — is a delivery channel for one of those three jobs, not a fourth job.

Text, SMS, or voice

Web chat suits your careers page. SMS suits high-volume, hourly-wage roles where candidates are far more likely to read a text than open an email. Voice agents are the newest of the three and the most divisive — Bullhorn's data below rates them surprisingly well, but they also sit closest to the "evaluative interview" line this whole piece is about. The channel matters less than what happens on it.

Chatbot vs. conversational AI vs. agentic AI

A chatbot follows a script with some language-model flexibility bolted on for open-ended replies. Conversational AI is the broader category either sits inside — natural language over text or voice, instead of a form or an IVR tree. Agentic AI goes further: it doesn't just reply, it takes multi-step action on its own — sourcing a shortlist, sending outreach, following up without a human triggering each step. Most of what's sold as a "recruiting chatbot" today is the first category with the second category's marketing.

The Results People Actually Report

Chipotle is the number every vendor deck borrows: time to hire dropped from 12 days to 4 — a 75% cut — after rolling out a conversational hiring platform, with application completion climbing from 50% to 85%. That figure comes from the vendor's case study, and it's worth noting the exact wording Chipotle's own newsroom used when announcing the rollout: the technology was "expected to reduce the amount of time it takes to hire...by as much as 75%." The corroboration is real. The "as much as" is real too.

That's the enterprise, high-volume-hourly end of the category, and it's a useful proof point. But it's not the evidence that matters most if you're running an agency desk. This is:

Here's a wider slice of what's been published, sorted by how much you should trust it. This is the part almost no vendor page does: grading its own evidence.

"Vendor-published" means the number comes from the platform selling the tool, with no disclosed independent methodology. That doesn't make it false. It means treat it as a best case, not an average.

The Candidate Side — Where This Goes Wrong

Every stat above is about the employer's side of the screen. Flip it around and the picture changes fast.

That population detail matters more than it looks. These are US figures, not a global average — worth knowing if you're benchmarking against a different market.

The trust numbers underneath tell the same story from a different angle. A Gartner survey found only 26% of candidates trust AI to evaluate them fairly, and 25% say they trust an employer less once they know AI is doing the evaluating. Back in the Greenhouse data, 46% of candidates want the option to request a human interview, and 39% want a clear explanation of what the AI is actually measuring. Neither ask is unreasonable. Neither is common practice yet.

The Distinction That Decides Everything

Here's the thing almost every AI-and-hiring hot take skips: those two datasets aren't contradicting each other. Greenhouse measured AI conducting evaluative interviews — often undisclosed, often silent afterward. A separate Bullhorn survey measured AI inside recruiter-mediated agency workflows — answering, scheduling, closing the communication gap. They're not the same category wearing different clothes. They're genuinely different jobs, and candidates are telling you, with real consistency, which one they'll accept.

One honest caveat: these two surveys are roughly a year apart — Bullhorn fielded June–July 2025, Greenhouse published April 2026 — on a variable that's moving fast, so some of the gap is timing. But the use cases are different enough that timing doesn't explain it away.

There's a bridge stat that ties the two together: 32% of candidates in Bullhorn's survey cite "receiving responses to all applications" as AI's single biggest benefit. Candidates aren't anti-AI. They're anti-silence. A chatbot that closes the application black hole is solving the actual problem. A chatbot pretending to be a judge is creating a new one.

Why Speed Matters More Than Almost Anything Else

16% of candidates never got acknowledgment that their application was received. Three-quarters have stopped working with a recruiter because the process took too long or communication was poor. Candidate satisfaction with recruiter speed and responsiveness fell nearly 20% year over year, according to Bullhorn's GRID 2025 Talent Trends Report. That's not an AI problem. It's the problem AI chatbots happen to be well suited to fix.

There's an older, frequently misapplied study behind the "speed to lead" argument, and it deserves its caveats stated plainly rather than smuggled in as a recruiting stat. A Harvard Business Review analysis of 1.25 million sales leads across 42 companies found firms contacting a lead within an hour were nearly 7x more likely to qualify it than those waiting an hour longer, and over 60x more likely than those waiting a full day. That study is from 2011, and it's about sales leads, not recruiting. It's directional, not proof — but it lines up with what Ettain Group reports inside actual recruiting: a 46% response rate in under 10 minutes.

Speed isn't the whole game. But of everything a chatbot can plausibly fix in a recruiting process, it's the one with the least controversial evidence behind it.

The Compliance Layer

This is a short section on purpose — it's genuinely a rabbit hole and this isn't the place to go all the way down it. Two rules come up constantly:

• NYC Local Law 144

  Applies to employers and employment agencies. Requires a bias audit within one year, public posting of the results, and 10 business days' notice before an automated employment decision tool is used. In force since July 5, 2023.

• Illinois AI Video Interview Act

  Requires pre-interview notice, an explanation of how the AI works, and consent. Requires deletion within 30 days on request. In force since 2020.

57% of candidates in the Greenhouse research think AI-use disclosure should be a legal requirement everywhere. In a growing number of places, it already is.

This isn't legal advice — confirm current requirements for your jurisdiction before deploying anything that scores or ranks candidates.

How to Evaluate a Recruiting Chatbot

Ten questions that separate a chatbot that helps from one that quietly costs you candidates.

Where Chatbots Shouldn't Be Used

Final rejections deserve a human voice, even a form-letter one. Anything that scores or ranks candidates without a human reviewing the result shouldn't ship, regardless of what the vendor calls it. Complex or sensitive candidate situations — a counteroffer, a visa question, a compensation negotiation — are exactly where a script breaks and a recruiter's judgment is the entire product. And client-facing communication is a relationship, not a workflow step; automate that one carefully or not at all.

FAQ: AI Recruiting Chatbots