Assess communication skills by listening for two things

Two of our test candidates spoke English at a C1 level, at 154 and 142 words a minute, with almost no filler words. Our transcript-based scorer gave them 5 and 4 out of 10 on communication. On paper their answers looked flat. Anyone who heard the recordings knew they could talk. That gap is the whole problem with trying to assess communication skills from text, and it is why we stopped trusting either half on its own.

The opposite failure happened too. On 23 April 2026 a tester hesitated on purpose through an entire interview, and our new audio scorer still gave the delivery 7.72 out of 10. Five bugs were behind it, including a pace penalty that never fired and a transcription setting that collapsed repeated words. After the fix, the same interview scored 3.26. One scorer read only the words and missed fluent speakers. The other heard only the sound and missed a struggling one.

If your first-round screen used to be a recruiter phone call, communication was the thing that call was really measuring. This post covers how to keep measuring it once the call is gone. It is part of a series on how one voice interview replaces the recruiter phone screen.

How do you assess communication skills in candidates?

Assess communication skills by asking every candidate the same questions that require explaining something to a listener, then scoring two halves separately: the content (did they answer the question, clearly and in order?) and the delivery (were they easy to follow out loud?). Write the scoring anchors down before the first interview, and keep the recording for borderline calls.

Two halves, scored apart, is the core of it. Most screening collapses them into one gut feel, and a gut feel is where accent bias and confident-but-empty answers slip through.

Why the recruiter phone call measured communication badly

Communication matters to employers more than almost anything else on a job post. LinkedIn ranked it the number one in-demand skill on its 2024 list. Yet the usual test is a 10-15 minute recruiter call with no written criteria.

SHRM’s guide to running effective phone screens lists what a good recruiter listens for: whether the candidate can carry a conversation easily or sounds rehearsed, and whether they give direct, relevant answers. The same guide warns recruiters not to judge on likability. That warning exists because likability is what an unstructured call rewards. The candidate who chats easily about cricket for two minutes gets a better note than the one who answered every question precisely but quietly.

The call also mixes the two halves. A strong accent makes a recruiter rate content lower, and a warm voice makes a vague answer sound better than it was. Separating them is the first fix.

What “communication skills” means in a screen

For a first-round screen, communication breaks into four things you can actually hear:

1. Answers the question asked. The candidate listened. Asked about a missed deadline, they talk about a missed deadline, not their strengths.
2. Puts the point first. A clear answer leads with the conclusion, then gives one example. A weak one circles for a minute before arriving.
3. Is easy to follow out loud. Steady pace, few false starts, words a listener can catch. This is clarity, not accent.
4. Adjusts to the listener. Asked to explain something to a customer, they drop the jargon they would use with a colleague.

The first two belong to the content half and can be judged from a transcript. The third is the delivery half and needs the audio. The fourth needs the right question, which is why the question design matters as much as the scoring.

A 0-3 rubric you can use tomorrow

Score each half separately, on every candidate, with these anchors written down before you start:

Two questions that produce real evidence for both halves:

• “A customer’s refund is a week late and they are angry. Explain to them, as you would on the call, what happened and what happens next.”
• “Tell me about a time you had to explain something complicated to someone who did not know the subject. What did you say first?”

Avoid questions that only test confidence, like “Why should we hire you?” They reward rehearsed speeches, which tells you about preparation, not communication.

The three traps that make communication scores unfair

Accent is not clarity. A listener’s familiarity with an accent changes how easy it feels to follow, so a scorer, human or AI, can mark down a perfectly clear speaker. Before you rely on any delivery score, listen to a sample of the low ones and ask whether a real customer would have struggled. We wrote more about this in AI screening and non-native English speakers.

Fluent is not substantive. A smooth answer with nothing in it should score 1 on content, however good it sounds. Candidates now read prepared answers off a second screen, and those sound fluent and polished. Watch for answers that sound like an essay, and see how rehearsed answers show up in a voice interview.

A fluency level is not a job score. A CEFR level describes general language proficiency (the Council of Europe’s level descriptions are written for language learning, not hiring). A B2 speaker can be excellent on a support line, and a C1 speaker can still fail to answer the question. Use the level as one input. Our explainer on what CEFR scoring measures goes further.

When should communication decide the outcome?

Decide this per role before screening starts, not after you have heard a candidate you like.

• Speaking is the job (support, inside sales, counselling, telecalling): set a written floor. Below it, the candidate does not move forward, whatever the rest of the score says.
• Speaking is part of the job (operations, analyst, most graduate roles): weight communication as one part of the score, perhaps a fifth of it, and never let it reject on its own.
• Speaking is incidental (back-office, some technical roles): score it, note it, and let the hiring manager weigh it.

Write the choice into the job’s screening setup so the same rule applies to the 1st and the 400th candidate.

How HireQwik scores communication

HireQwik runs the first-round screen as a 10-20 minute voice interview the candidate takes in a browser, at a time they book. Communication is scored after the call from the full transcript, like every other part of the interview, with the candidate’s own words quoted as evidence.

On accounts with speech assessment switched on, a second score comes from the audio: pace, hesitation, pronunciation and a CEFR fluency level, measured on the candidate’s own audio track. Pace between 130 and 180 words a minute carries no penalty, and neither do fewer than 3 hesitations per 100 words. That delivery score is averaged 50/50 with the transcript score for the communication dimension only, so it can move communication up or down, while technical and experience scores stay transcript-only. Both halves are stored and shown next to the result, so a recruiter can see exactly which half drove the number. If the audio is poor, the blend switches off and the score falls back to the transcript. Under 30 seconds of candidate speech, no speech analysis runs at all.

A job can set a communication floor, which is 4.0 out of 10 by default, for the roles where speaking is the job. The AI recommends; HR decides who moves forward. If you want to see what a candidate sees and what the scorecard shows, take the live AI interview demo, and read how to review an AI communication score before you trust one.