Recruitment chatbots sit at the top of the funnel, where volume is highest and recruiter time is scarcest. Typical use cases:

• Answering candidate questions about the role, salary range, location, and process
• Collecting details such as availability, work authorization, and certifications
• Guiding applicants toward roles that match their skills on a career site
• Booking interview slots and sending reminders
• Re-engaging candidates already sitting in a talent pool
• Conducting a first-stage screening interview and passing results into the ATS

The first five are administrative. The last is an assessment activity, and it carries a different set of obligations.

A scripted chatbot follows a decision tree. It recognizes keywords, serves prewritten replies, and stops when the candidate says something the tree does not anticipate. It works well for status updates and FAQs.

Conversational AI interprets unscripted language, asks a relevant follow-up, and adapts to how the person actually answers. That matters for screening, because a single question rarely produces a complete answer. Without a follow-up, you score a fragment as if it were the whole response.

• Response speed. Candidates get answers immediately instead of waiting on an inbox.
• Availability. Applicants engage outside office hours, which suits shift workers and people already in a job.
• Consistency. Every applicant receives the same questions in the same order.
• Mobile completion. High-volume candidates apply from a phone, and conversational formats fit that behavior better than long forms.
• Recruiter capacity. Administrative screening moves off the recruiter's plate, freeing time for interviews and hiring manager conversations.

Two risks matter most.

The first is explainability. If a chatbot uses a large language model to score or rank candidates, the reasoning it displays is usually generated after the score, not derived from it. That is post-hoc plausibility: a sensible-sounding narrative rather than the actual logic. When a rejected candidate or a works council asks why a decision was made, a plausible story is not a defensible answer.

The second is regulatory. Under the EU AI Act, AI used in recruitment is classified as high risk. Article 13 requires transparency sufficient for a human to interpret the output, and Article 14 requires meaningful human oversight. A screening chatbot that cannot show its scoring criteria leaves the employer, not the vendor, having to justify it.

Ask vendors these questions before shortlisting:

1. Does the tool assess candidates, or only collect information? Be clear which one you are buying.
2. If it assesses, what model produces the score, and is the same logic shown to the recruiter?
3. Are the questions grounded in a structured framework such as STAR, with explicit weightings?
4. Is there a full audit trail linking every score to a specific candidate response?
5. Does the final decision stay with a person, with no automatic rejections?
6. How many languages does it support, and does quality hold across all of them?
7. Does it write results back into your ATS, or create a parallel system to maintain?
8. Do candidates receive feedback, or does the process end in silence?

A vendor who cannot answer question two in plain language is asking you to accept a black box.

Structured, skills-based interviewing is the difference between a chatbot that chats and one that assesses. Hubert takes a two-layer approach: a conversational layer that makes the interview feel human, and a deterministic assessment layer where the same input produces the same output, with full explainability. Recruiters receive scored, auditable shortlists in their ATS across 30+ languages.

The results show up on both sides. Candidates rate the experience 9/10 on average, average completion sits at 96%, and time-to-hire improves by up to 80%. At NSS Group, competency-based screening produced a 50% increase in hires from candidates who would never have passed traditional CV screening. Malmö Stad compressed a recruitment cycle from six weeks to two days.

That last figure is the real argument for taking chatbot selection seriously. A tool that reads CVs faster reproduces the same filters at greater speed. A structured interview asks every applicant to demonstrate competence, which surfaces people a CV would have hidden.

Is a recruitment chatbot the same as an AI interview? No. A chatbot handles conversation and data collection. An AI interview assesses competencies against defined criteria and produces a score a recruiter can review and justify. Some tools do both; many do only the first.

Do candidates dislike being interviewed by a chatbot? Not when the experience is well designed and honest about what is happening. High completion rates and 9/10 average satisfaction are achievable when candidates can answer in their own words, on their own schedule, on their phone.

Can a recruitment chatbot reject candidates automatically? It should not. Under the EU AI Act, high-risk AI systems require meaningful human oversight. Responsible deployments recommend and rank; a recruiter makes the final decision.

Does a recruitment chatbot replace the ATS? No. It should integrate with your existing ATS and write results back into it. A tool that stores candidate records separately creates duplicate work and a compliance gap.