Judged on what you say. Never on how you sound or look.
SeekIQ scores answers against a rubric the hiring team sets before anyone applies. A person decides by default, and any automation a team switches on is shown to candidates and can go to a person. This page explains how, step by step.
What we assess. Job-relevant skills, agreed before the first application.
Before a role opens, the hiring team chooses the skills the job needs and writes a rubric for each one. Every candidate for that role is measured against the same rubric.
01Skills
What the job needs, such as calming an upset customer or planning a shift. Each skill has a written rubric from 1 to 5.
02Evidence
Every score quotes the words it is based on, so a reviewer can check it against the recording.
03Consistency
Candidates get equivalent questions that share one rubric, so nobody gets an easier test.
04Enough to judge
When an answer gives too little to go on, the score says so instead of guessing.
What we never assess. Some signals are noise dressed up as insight. We leave them out by design.
None of these reach a score. Personal details are removed before an answer is scored, and rubrics describe job skills only.
- AccentSylheti, Lagos or Glasgow, every answer is judged on what was said.
- AppearanceScores come from the words of an answer. No photo or video goes into a score.
- EmotionNo “confidence” or “enthusiasm” detection. Nervous is not a trait.
- Language choiceAnswering in your own language instead of English never costs a point, unless the role truly needs English.
- Voice pitch and paceA deep voice, a soft voice or a slow speaker gets the same score for the same answer.
- AgeAge is never a rubric item and is never scored.
- GenderGender is never a rubric item. Names are removed before an answer is scored.
- Religion and ethnicityBeliefs, caste and ethnicity are never rubric items and are never scored.
You choose who decides. By default a person on your team decides. Autopilot stays off until you switch it on, job by job.
- A person by defaultEvery stage starts manual: only a member of the hiring team advances or rejects a candidate there.
- A reason every timeA rejection always carries a reason, whether a person made it or a rule the team switched on, and every changed score records why.
- Automation you can see and stopIf a team switches on autopilot for a job, each automated decision names its rule, the candidate is told and can ask a person to review it, and one switch stops it. Where the law limits automated rejection (the EU, the UK and New York City) it stays off until an owner confirms it.
How a score is made. Six steps, in plain words.
This is what happens to an interview answer, from the question to the decision.
- 01
The rubric comes first
Before the role opens, the hiring team picks the skills that matter and writes down what weak, solid and strong answers look like.
- 02
Everyone gets a fair question
You hear a question written for one of those skills. Another candidate may hear different wording, but both answers are measured the same way.
- 03
Your words are written down
Your spoken answer is turned into text. You can correct it, and the original recording is kept beside your correction until it is deleted, 30 days after it was made by default.
- 04
Personal details come out
Your name, phone number and other personal details are removed before the answer is scored.
- 05
The answer is compared with the rubric
SeekIQ's interviewer suggests a score from 1 to 5 for each skill and quotes the words it relied on. If there is not enough to judge, it says so.
- 06
A person decides, or the team's rules
A recruiter reads the evidence, can listen to your answer and makes the decision. If they change a score, they write down why.
How we test for bias. Before any change goes live.
We keep a test set of answers, recorded with consent by people of different genders, ages, regions and accents, in every language we support.
Every new rubric template, prompt version or speech model runs through that set first. The same answer should get the same score whoever says it and in whichever language. If it does not, the change does not ship.
Job descriptions and rubrics are also checked for biased wording, such as age limits or gendered terms, and the person writing them is asked to fix it.
A quarterly audit. Our commitment, with every output on the record.
Pass rates, compared
Each quarter we will compare pass rates at every stage by gender, region and language, look into any gap and publish what we find.
Every AI output is logged
The audit log keeps each score and summary for at least six months, with the model version and prompt version that produced it.
Open to review
Enterprise customers can export the full audit log for their own reviews.
The first quarterly report will be published here after a full quarter of live interviews.