What tools support voice responses for language proficiency testing?

Voice responses are scored by a narrow set of platforms: dedicated speaking tests such as Versant by Pearson, Mercer Mettl's SpeechX and SHL's SVAR, plus assessment platforms that score a spoken answer inside a wider test, Testlify among them. The harder question is not which tool can record audio. It is what a recorded answer is allowed to decide about a person.

That second question is where most buying processes go wrong. A speaking score feels objective because a machine produced it, so it quietly becomes the pass or fail gate for roles where spoken English was never the real requirement. Under US law that is where the risk sits, and it is the part no vendor comparison talks about.

TL;DR

• Voice scoring is worth buying when the job genuinely runs on spoken language, and it is a liability when it does not.
• The EEOC's position is narrow: a fluency requirement is defensible only where it is required for the job it is attached to.
• Set a CEFR level per role before you shortlist a vendor. The level is a hiring decision, not a product feature.
• Judge a platform on evidence it will show you (scoring method, human override, item-level data, transcripts), not on the word innovation.
• A screening test is not an accredited exam. Say so to candidates, and never present a screening band as a certificate.
• Budget for reviewer time as well as license cost. The audio nobody listens to is the most common waste in this category.

What is a language proficiency test?

A language proficiency test measures what somebody can do in a language right now, usually across listening, reading, speaking and writing, and reports it as a level rather than a pass or fail. In hiring it is a screening instrument. It estimates working ability for a role, and it does not certify anybody.

Most tests report against the Common European Framework of Reference. The European Commission's CEFR self-assessment grid runs from A1 to C2 and splits ability into five columns: listening, reading, spoken interaction, spoken production and writing. Those last two matter here, because they are different skills. Someone can hold a fluent scripted monologue and still stall in a live back and forth with an annoyed customer.

Keep the vocabulary honest with candidates. A 12-minute automated test gives you a band, not a qualification. Our own language proficiency tests for hiring report on the CEFR scale for that reason: a shared scale makes a shortlist comparable across countries and recruiters.

Which platforms are leading language testing innovation?

No vendor-neutral ranking of this category exists. The dedicated speaking tests most often shortlisted are Versant by Pearson, SpeechX from Mercer Mettl and SVAR from SHL, while generalist assessment platforms including Testlify score spoken answers alongside role skills. Judge any of them on evidence they will show you, not on the word innovation.

Every platform here claims AI scoring, CEFR alignment and fast results, and the validation behind those claims is rarely published anywhere a buyer can read it. We keep a separate per-vendor rundown of tools for testing language skills if you want the shortlist shape.

What separates them is not the model. It is five things you can check in a trial:

• Scoring transparency. Can the vendor show how an audio answer became a number, and can a human reviewer change that number?
• Turn-taking. A real conversation interrupts, clarifies and reacts. A recorded prompt followed by 60 seconds of monologue does not.
• Item-level data. Difficulty and discrimination statistics per question are how you find a question that is quietly failing good candidates.
• Transcripts. Without one, nobody can audit the decision later, and audio alone is unreviewable at volume.
• Published limits. A vendor that names what its scoring cannot see is more trustworthy than one that names nothing.

Only one of those is a model question. Four are product and process questions, which is why the newest engine often loses to the platform that lets a recruiter see and correct the work.

How do you choose language testing software?

Start with the role, not the catalog. Write down the language tasks the job actually contains, pick the CEFR level each task needs, then shortlist tools that can measure those specific skills at that level and show the evidence. Anything else is buying a score you cannot defend.

That sequence is the Testlify Competency-to-Evidence Matrix applied to a single competency. The matrix maps every role to the competencies that matter, then connects each competency to measurable evidence through assessments, simulations, interviews, references and structured feedback. For language, it keeps you honest about which skill you are testing and who makes the call.

Those levels are a starting point, not a standard. Two companies hiring the same job title can land a level apart for good reasons.

Pro tip: set the pass level and write it into the job requisition before the first candidate is invited. A bar chosen after you have seen the scores is a bar chosen around the candidates you already liked, and it is the single easiest thing for a claimant's lawyer to take apart.

What is the best standardised English language assessment for a workforce?

There is no single best one, and a vendor who tells you otherwise is selling. For an existing workforce the question splits in two: a benchmarking exercise across many people wants one short test on one scale, run the same way for everybody, while a promotion or redeployment decision wants a role-specific assessment closer to the actual work. Use the same scale for both so results stay comparable.

Which are the most accurate language proficiency exams?

Accuracy depends on the skill being measured, not the brand. Automated scoring of read-aloud and short spoken answers is the most settled part of the field, because the task is tightly constrained. Open-ended conversation, persuasion and humour are where machine scores and human raters drift apart most, so those are the answers a person should still read or hear. The structured-interview principles the US Office of Personnel Management sets out for structured interviews apply directly: the same predetermined questions in the same order, and the same rating scale and standards for acceptable answers.

Is a job proficiency test the same thing?

No. A job proficiency test measures whether someone can do the work, which may involve language and usually involves more: tools, judgment, procedure, numeracy. A language test measures the language. Buying one when you needed the other is how teams end up rejecting a strong hire over an accent and hiring a fluent speaker who cannot run the process.

Candidate assessment software or a dedicated language test?

Pick a dedicated test when language is the job and the volume is high, and pick a broader platform when language is one of several things you need to measure. A call center screening thousands of applicants for spoken English has a different problem from an agency hiring six bilingual account managers a year.

The split is really about what happens after the score. A dedicated test hands you a band, and you still have to weigh it against the role skills, the work sample and the interview. A platform that holds all of it gives you one candidate record instead of four exports to reconcile.

Where candidate evaluation software fits

Candidate evaluation software is the layer that turns scores into a shortlist: weighted scoring, multiple reviewers, a decision log. It is worth caring about because language decisions are the ones most likely to be challenged later, and a decision log is the thing you will wish you had. If language sits inside that record next to everything else, the reviewer can see a B1 speaker who aced the process test and make a real judgment instead of a reflex.

Do ATS integrations matter for language assessment?

They matter more than the test interface does. A language score that lives in a separate tool gets re-typed, goes stale, or never reaches the hiring manager at all, and a candidate gets asked to retake a test they already passed.

Testlify connects to more than 100 ATS platforms and leaves your ATS as the system of record, which is the right arrangement for any team that already runs one. Candidates can advance to the next ATS stage automatically when they clear the cut-off score you set. Worth knowing before you design the workflow: a rejection made inside Testlify stays inside Testlify and does not write back to the ATS, so the final stage change is still a human action in your system. The ATS integration catalog lists what connects today, and on the self-serve tiers that integration is a paid add-on rather than part of the base plan.

Is it legal to screen candidates on English?

In the US, only where the job requires it. The EEOC's 2016 enforcement guidance on national origin discrimination states that an English fluency or proficiency requirement "is permissible only if required for the effective performance of the position for which it is imposed", and warns that applying uniform fluency requirements across a broad range of dissimilar positions can violate Title VII. One company-wide English bar across every role is the pattern that guidance describes.

Two more lines from the same guidance change how a test should be used. An employment decision can rest on an accent only where the accent "interferes materially with job performance", which is a job-performance question and not a preference question. And rules requiring employees to speak English at all times are presumptively unlawful, so an English-only workplace policy is not a safe companion to an English screening test.

The test itself is a selection procedure like any other. The EEOC's guidance on employment tests and selection procedures puts the burden plainly: an employer must be able to show a procedure is job-related and consistent with business necessity, and the Uniform Guidelines adopted in 1978 set out what test validation means. None of that forbids language testing. It does mean the level, the role, and the reason are written down before you use it, and that a human reviews the edge cases rather than a cut-off doing it silently.

How much does a language test cost to run?

License cost is the smaller half. Per-candidate assessment pricing sits inside a plan, outbound AI phone interviews are billed by the minute (Testlify charges $0.18 per minute), and the real expense is reviewer attention: the audio somebody has to listen to when a score lands near your bar.

Model it on your own funnel first. Take an illustrative support team hiring 20 agents a quarter from 400 applicants: if 15 percent of spoken answers land near the pass level, that is 60 reviews, and at eight minutes each that is a full working day of somebody's time. Bureau of Labor Statistics data puts customer service representatives at 2,666,000 jobs in 2025, a median wage of $44,770 and about 289,500 openings a year, so this is a cost-sensitive funnel where a review day per quarter is a real line item. Budget it and the test earns its keep. Skip it and the score becomes the decision by default.

Testlify

Testlify is an AI-native skills assessment and interviewing platform. Language sits in its test library as one category next to cognitive, role-specific, coding and psychometric tests, and spoken answers are scored inside the same candidate record as everything else.

What it measures

• Spoken answers through voice, video and outbound AI phone interviews, in one-way and conversational formats
• Written production through typed answers, with real-time translation of questions into a candidate's preferred language (quality varies by language pair, and the product says so)
• CEFR-style reporting, with CEFR available as a report type alongside the role and skill breakdowns
• Pronunciation insights and speech authenticity, both of which can be shown or hidden on a shared report
• Role skills in the same sitting, which is the point of measuring language next to the work rather than apart from it

Key features for language screening

• Configurable AI interviewer: avatar, voice, persona, context and custom prompts, with 150+ interview templates
• Attempt and timing control: number of attempts, recording time, preparation time before recording starts, and a five-minute ceiling per video answer
• Automatic multilingual transcripts, which auto-detect the spoken language for recordings of at least 30 seconds
• Auto-scoring with human override, so a reviewer can change a machine score and leave a record of it
• Item-level psychometrics per question (difficulty, discrimination, a quality-risk flag), so a bad question gets found
• Three difficulty levels and 25+ question types, with assessment certificates whose pass threshold you set anywhere from 40 to 100 percent

Limits worth knowing

• Assessments run on Chromium desktop browsers (Chrome or Edge). The mobile app is a capture fallback for audio and video, not a way to sit a whole test on a phone.
• Session recording is not captured on mobile for voice or video questions.
• Testlify does not schedule live human interview panels. Calendar coordination for a panel stays in whatever you use today.
• Reattempting an interview deletes the previous take, so set the attempt policy deliberately.

Best for

Teams under a few hundred people hiring customer-facing or offshore staff, who need language evidence sitting beside role skills in one record rather than a standalone certificate. Need an accredited exam result for an immigration or academic purpose? A screening platform is the wrong instrument. Candidates notice the difference between the two, which matters in a market where candidate experience decides who accepts an offer.

Hire on evidence, not an accent

Pick one role this quarter, write the language tasks and the level into the requisition, then run 20 candidates through a voice answer scored against that level. You will learn more about your real bar in two weeks than a vendor shortlist teaches in two months. Book a Testlify demo to see the voice interview, the transcript and the reviewer override on one candidate record, or try the AI EnglishPro test first.

FAQs