At MetaPret, a verified interpreter is one the MetaPret team has reviewed and approved from résumé and profile. Skill is then measured per request: a test built for that request is scored on three axes, in each direction, and only interpreters who pass are matched.
As of
Author: Paul Dahoon Kim (Founder, MetaPret. Working Korean-English interpreter, ~100 engagements)
Last updated: 2026-10-19
Read time: ~7 minutes
Summary: When an interpreting platform says "verified interpreter," the word usually stands for self-reported experience and a self-paced tutorial. At MetaPret, "verified" means something else. This guide sets out MetaPret's full verification method: the registration review, the test built for each request, the AI monitor, and the three-axis rubric.
Introduction
"Verified interpreter."
The phrase is used across interpreting platforms, and what it means varies widely.
On most platforms, "verified" covers some combination of:
- The platform checks a résumé the interpreter uploaded
- The interpreter completes a self-paced tutorial
- The interpreter passes a hardware and software compatibility check
- (In some cases) a peer referral or an interview
At no point in that list does the interpreter perform a real interpretation task and receive a score. "Verified" does not mean skill was verified. The interpreting market has used the word loosely, and clients have taken it as reliable information.
When MetaPret uses the word "verified," it means something different.
A verified interpreter is one who has passed a scored, request-specific interpretation test, not one whose résumé or certificates were checked. The test is scored on three axes, in each direction, and an interpreter has to pass a new test in the request's field for every request they are matched to.
This guide lays out that difference. It describes MetaPret's verification method at the level of an operating document, not a marketing summary: what is tested, who scores it, which rubric is used, and which interpreters pass and which do not.
I am Paul Dahoon Kim. I have worked as a Korean-English interpreter in Korea across around 100 engagements.
Registration review, then a test for every request
MetaPret's verification has two parts: a profile review done once at registration, and a test taken fresh each time a request comes in. Skill is measured in the second part.
Registration review
When: When an interpreter registers with MetaPret. Registration is open to anyone.
What is reviewed: The résumé and the profile photo. The MetaPret team reviews them and approves the interpreter. The grade (Associate, Junior, or Senior) is set manually by the MetaPret team at that approval.
What this review does not do: It does not measure interpreting skill. Being approved at registration does not mean an interpreter will be matched to any request.
The test for each request
When: Every time a client submits a request.
Who takes it: Every interpreter who wants to be a matching candidate for that request.
What it tests: Interpreting ability in the field and format of that request, in both directions of the language pair.
How it works:
- The client states the field of the request (medical, IR, M&A, legal, advertising pitch, general business, factory tour, and so on).
- MetaPret's system builds a test for that field and format.
- Approved interpreters who apply to the request take the test.
- An interpreter becomes a matching candidate for that request after passing the test.
How it is scored:
- The recording is scored on three axes, in each direction: content fidelity, target-language fluency, and delivery. No one scores their own test.
- An AI monitor cross-checks the scoring and raises an alert if scoring patterns drift over time.
Why this matters:
Interpreting skill varies by field. An interpreter who is strong on IR calls may be weak on the vocabulary of a medical conference. An interpreter who is strong in M&A negotiation may be unfamiliar with the labor-union context on a factory floor. A résumé and general experience cannot guarantee accuracy in a given request.
So each test is built to the weight of the request. A medical test looks at drug dosage instructions, patient consent procedures, and diagnostic vocabulary. An M&A test looks at negotiation modality, financial terms, and legal procedure. They are different tests.
The cumulative effect:
The more request tests an interpreter passes, the more measured evidence builds up, rather than self-reported experience.
This accumulated data is what lets MetaPret's matching become more accurate over time.
The scoring rubric: three dimensions
Request tests are scored on this rubric, in each direction. Points come off for each error from a fixed table; the levels below show what strong and weak renditions sound like on each dimension.
Dimension 1: Content fidelity
Did the interpreter carry over the meaning and logical structure of the source without omissions?
Excellent: Every concept rendered accurately. No modality drift. Logical structure preserved, no important detail missing.
Good: Minor omissions, but the core meaning and structure come through.
Fair: Shifts in meaning that an attentive listener would catch.
Weak: Meaning drift that a listener could miss.
Poor: Significant loss of meaning.
The most common failure modes:
- Compressing Korean modality (rendering "검토해 보겠습니다" as "we will give it serious consideration": a polite refusal shifts into a serious review)
- Blurring numerical ranges (rendering "between 20% and 25%" as "around 20 to 25 percent": a closed range shifts into a vague one)
- Misreading Japanese indirect refusal (rendering "難しいですね" as "we're finding it difficult": a polite refusal shifts into room to negotiate)
Dimension 2: Target-language fluency
Is the rendering concise, with accurate equivalent expressions and correct usage that sounds natural in the target language?
Excellent: Sounds like natural speech from a native speaker of the target language. Accurate equivalents, concise, correct usage.
Good: Mostly natural, with minor awkwardness.
Fair: Noticeable literal rendering or translationese.
Weak: Unnatural usage and phrasing that the listener has to smooth over.
Poor: Unnatural as target-language speech, or frequent usage errors.
Dimension 3: Delivery
Excellent: Pace, tone, and prosody match the source. Easy to listen to.
Good: Minor pacing issues, otherwise clean.
Fair: Some hesitation or fillers that affect the flow.
Weak: Distracting delivery issues.
Poor: Hard to follow as audio.
Who scores:
A professional evaluator confirms the scores on all three dimensions. Self-scoring is not allowed.
The role of the AI monitor:
The AI monitor does not replace the evaluators' scoring. It analyzes scoring metadata: consistency, drift in scoring over time, and statistical anomalies in score distributions. It uses data to catch patterns that human evaluators can miss.
After passing: how interpreters appear
MetaPret selects the interpreter from the candidates whose test for that request was confirmed. The client does not compare candidates.
What the client sees:
- Grade (Associate, Junior, Senior)
- Language pair
- City and working region
- The confirmed interpreter's profile, with why their experience and background fit that request (individual scores are not shared)
What the client does not see:
- Evaluators' notes
- Registration review notes
Interpreters receive written feedback on their test.
What "verified" does not mean
For credibility, here is what MetaPret does not promise with the word "verified."
- That every engagement will be perfect. Verification raises the odds; it does not guarantee 100%. That is why, if a match falls short of expectations, MetaPret works out a resolution with the client.
- A one-time certification. The request test is taken anew for every request. It is not a static credential.
- That the interpreter is strong in every field. An interpreter who passed the test in one field can fail it in another.
- A replacement for external credentials such as AIIC or KSCI. External credentials are used as a supporting signal. They do not replace MetaPret's verification.
How clients can check
When the match is confirmed, the client receives:
- The confirmed interpreter's profile: grade, language pair and city
- Why that interpreter's experience and background fit the request
- A short field introduction written by the interpreter (optional)
Individual scores are not shared. If the client needs more before the engagement, a pre-briefing call with the interpreter can be arranged.
Conclusion
The phrase "verified interpreter" has been used lightly in the market. At MetaPret, the word stands for an entire operating system.
A registration review and a test for every request. A three-axis rubric plus an AI monitor. Re-verification for each request. All of it is published methodology.
This guide is not a marketing summary. It describes the system MetaPret runs every day.
Submit a request. Matching with a verified interpreter starts at metapret.net/request.
FAQ (FAQPage schema)
Q: Once an interpreter passes a test, are they done?
A: No. To become a matching candidate, an interpreter takes a new test in the request's field for every request. That applies equally to high-stakes fields such as medical, legal, and M&A.
Q: Who are the evaluators, and are they vetted too?
A: A professional evaluator confirms the score. The evaluator's screen is blind: no applicant name, region or tier is shown. Self-scoring is not allowed, and no interpreter ever scores their own test.
Q: Can clients take part in the evaluation directly?
A: Client feedback after the engagement is recorded separately from test scores.
Q: Are AIIC members exempt from the test?
A: No. AIIC membership is a signal considered in the registration review, but it does not replace the request test. AIIC members take the test for every request too.
Q: Can clients see test scores?
A: No individual scores are shared with the client. MetaPret selects from the candidates whose test was confirmed, and the client receives the confirmed interpreter's profile and why their experience fits the request. Interpreters receive written feedback on their test.
About the author
Paul Dahoon Kim (김다훈)
Founder & CEO of MetaPret. He has worked as a Korean-English interpreter across around 100 engagements. He takes on interpreting engagements himself every quarter to keep first-hand experience of how the verification system works in the field.
Author page · Contact: cs@metapret.net