How AI interviews are scored

• Only the words of your answers are scored. Your voice, accent, face and grammar never are.
• Each field has four areas. An area gets a score out of 100 only when a quote of your own words supports it.
• For each area, an AI rates three things: the situation, your reasoning, and the result. Those ratings set the score on a fixed scale that is the same for everyone.
• Caps keep each score in line with its evidence. Without a specific case, with only short quotes, or with communication shown only by answering clearly, an area cannot go past a set limit.
• People at the company decide what happens next. The AI does not.

An Aethel interview is scored on what you said, not on how you said it. The published rules put it this way:

So these never count toward a score:

• Your voice, accent or tone. Aethel does not record audio.
• Your face, appearance, background or video. Aethel does not record video, and your camera, if you turn it on, stays on your device.
• Your grammar, word choice, or how fluent your English is.
• How fast you type, or how long you take to answer.
• Your age, gender, race or other personal traits. The interview does not ask about them.
• Your school, past employers or resume, unless you talk about them in an answer.

Grammar is kept out by design. Before the ratings are made, an AI copy-edits your answers into standard English grammar, keeping every fact, and the rating step reads that copy. The copy is not saved, not shown to anyone, and never quoted as your words.

What is scored is four areas for your field, so a caregiver is not judged like an accountant. For example:

• Healthcare support and caregiving: Noticing and responding, Safety and procedures, Communication, and Care team teamwork.
• Retail, food and hospitality: Problem solving, Service and standards, Communication, and Teamwork.
• Finance and accounting: Financial analysis, Accuracy and controls, Communication, and Stakeholder judgment.

The four areas for every field are on the list of fields Aethel interviews for.

Not every reply is an answer. Saying you are ready, making small talk, or asking for the question again is part of a normal conversation, but it is not scored. The published rule:

Asking the interviewer to repeat a question costs you nothing. It just does not count toward that minimum. Each answer is also checked against the question it answers: an answer about something else gets no credit, and the result names that question in a short note. That check is done by an AI, so it can miss.

Every score has to show where it came from. For each area, the AI copies quotes from your answers. Then the published rule applies:

In practice, that means:

• A reworded or made-up quote is rejected, and so is a fragment that is too short.
• A quote that only gives background, such as where you worked, does not count on its own.
• The interviewer's questions never count as evidence.
• For a teamwork area, saying "we" or naming your team is not enough on its own. The quote has to show you working with someone, such as agreeing on a plan or settling a disagreement.

When no quote is left for an area, the area says "Not enough evidence". That is not a low score: it means the interview did not show that area, and the area is left out of the overall score. If speech-to-text mishears a word, the misheard word is never shown as yours. The AI picks another quote, or shows the likely word in brackets.

For each area that has evidence, an AI rates three things in your answers. Each level has a fixed meaning, the same in every field:

Together, the three levels give between 0 and 8 points. The rating step reads your copy-edited answers 14 times, and the area's score is the average of those readings. Each reading also gives the AI's own number for the area. When that number is more than 5 points below the scale, the reading counts as the lower number. A reading never counts above the scale.

To hear the difference between the levels, read what a strong interview answer sounds like, which has made-up weaker and stronger answers for each rating.

The points set the area's score on a fixed scale that is the same for everyone and every field: 8 points is 95, 7 is 91, 6 is 87, 5 is 80, 4 is 76, 3 is 72, 2 is 66, 1 is 60, and 0 is 50.

• 8 points

  Area score

  95

• 7 points

  Area score

  91

• 6 points

  Area score

  87

• 5 points

  Area score

  80

• 4 points

  Area score

  76

• 3 points

  Area score

  72

• 2 points

  Area score

  66

• 1 point

  Area score

  60

• 0 points

  Area score

  50

A made-up example: a specific case with where it was and how big (situation 2), one trade-off weighed (reasoning 2), and a result a supervisor confirmed (result 2) total 6 points, which is 87.

Some limits apply whatever the points add up to. Each one ties a score to the evidence behind it.

• No specific case. Without a specific case, named with where it was and how big, an area is at most 60, whatever the other two ratings are. So a case told without its context (situation 1), with options compared (reasoning 3) and a result followed up over time (result 3), totals 7 points, which would be 91, but the area stays at 60.
• Quote length. An area scores 90 or more only when one of its quotes is complete (it ends at a full stop, comma, colon or bracket in your answer) and at least 20 words long. Otherwise it is at most 89. When its best quote has fewer than 12 words, it is at most 74.
• The top. No area scores above 95.
• Communication. An answer that responds to the question, in a clear order, with a concrete example, counts toward communication, but only up to 80 on its own. A higher communication score needs you to describe explaining, writing or presenting to people.
• Warning signs. In every field except software engineering, the AI is also told what strong work and warning signs look like in that field, such as a caregiver skipping a hygiene step to save time. It is told to score an area low when the answers show a warning sign, however confident they sound.

So an area marked "Not enough evidence" does not pull your overall score down. If only one area has evidence, you see that area's score and no overall score.

The rating is done by an AI, and the published rules are direct about what that means:

Averaging several readings is meant to narrow the spread, but it does not remove it. Speech-to-text can also mishear words, especially with background noise or technical terms, so you can always type instead. This is one reason every score shows its quotes: you, and anyone you share a result with, can check the evidence. If you think a score is wrong, choose "Ask for a review" on your results page. We reply within 2 business days. Replies come from Aethel's AI support assistant.

People make hiring decisions, not the AI. A company that sees your result reads the transcript and decides for itself. Aethel never rejects or advances anyone on its own.

A result describes one interview of about 15 minutes, so it covers only part of what you can do. It is not a certification, license or credential check, and Aethel does not verify degrees, licenses or work history. Aethel has not tested whether results predict job performance. You see your results first, and you choose who else sees them.

Read every rule in full on how the AI interview and scoring work, or open a sample result with a made-up candidate to see scores, quotes and "Not enough evidence" on one page.

Scoring rules differ from one AI interview tool to the next, and this page describes only Aethel's. If a company asks you to take an AI interview, it is fair to ask the company which tool it uses and how that tool works. Useful questions to ask:

• Which tool runs the interview, and where are its scoring rules published?
• What is scored, and what is never scored? For example, is my voice or video analyzed?
• Is the interview recorded, and how long is the recording or transcript kept?
• Can I type instead of speaking, or ask for another accommodation?
• Can I see my own results, and can I ask for a review?
• Does a person read my answers before any decision is made?

Some places have laws about AI in hiring that give candidates a right to a notice or another way to be assessed. See how Aethel supports AI hiring laws.

• How to prepare for an AI interview

  A step-by-step plan for any AI interview: pick three real cases, practice out loud, check your tech, choose accommodations, and know what to do if it breaks.

• What a strong interview answer sounds like

  Made-up before and after examples of interview answers that show how to name a specific case, explain your reasoning, and give a measured result.

• How the AI interview works: the full interview guide

  What to expect on the day, the tech check, accommodations, and what happens after.