Methodology
How YAGO's scores and feedback work
What each tool reads, how the score or feedback is produced, what it cannot tell you, and what to do if a result looks wrong.
Checked against the product on 29 September 2026.
The short version
- YAGO gives feedback on three things: a CV you upload, an interview answer you record, and possible career directions.
- Scores and feedback are YAGO's own assessment, written by AI models that follow instructions YAGO wrote. They are not an employer's hiring decision, and they do not predict whether you will be invited to interview or offered a job.
- AI can be wrong. The same CV can get a different score if you scan it again, and a suggestion can be inaccurate. Check every suggested change against your real experience before you use it.
- The CV scan reads the text of your PDF, not how it looks. The interview feedback reads a transcript of your answer and, if you recorded video, up to six still frames from it.
CV scan
Upload a PDF and YAGO returns a score out of 100, an ATS score and an industry-fit score, your strengths and gaps, and suggested edits.
What it accepts
- PDF files only. The scan does not accept Word files.
- Up to four pages and about 2,800 words. A longer CV is refused, and YAGO recommends two pages.
- A PDF with text you can select. A scanned or image-only PDF has no text to read, so YAGO says it could not read the file and asks for a searchable PDF. It does not read text out of images.
- English or Romanian. These are the languages YAGO supports.
What it reads
YAGO extracts the text from your PDF and sends that text, with the target role if you typed one, to an AI model. The model never sees the file itself, so it does not see fonts, colours, images or spacing.
How the score is produced
The scores are written by the AI model, not calculated by a formula in YAGO's code. The model is instructed to weigh three areas: evidence of impact (45%), fit for the role (30%) and ATS-readable structure (25%). YAGO's code checks that the answer has the expected fields and numbers. It does not recompute the scores.
The instructions also cap the score when a CV has a clear weakness: below 70 if the summary has no concrete number, below 60 if more than half the bullets start with weak verbs such as "Responsible for", and below 55 if it has long paragraphs instead of bullets. These caps are instructions to the model, not rules that the code enforces.
The model is told to judge against a demanding standard: hiring for top technology and large companies. A score reflects that standard, not the requirements of every employer or field.
The ATS score is the model's estimate of how well the CV's structure and wording would be read by automated screening. It is not the output of a real applicant tracking system, and those systems differ.
Will the same CV always get the same score?
Not necessarily. YAGO fixes the model's settings to keep scores steady, but an AI model can still return a different number on another run, and a backup model is used if the first one fails. Treat small differences as noise and look at the suggestions instead.
What the scan does not do
- It does not compare your CV with a specific job description. The only input besides the CV is an optional target role.
- It does not run an employer's applicant tracking system.
- It does not check that what you wrote is true.
- It does not tell you whether you will be invited to interview.
Suggested edits
The suggestions are written by AI. AI suggestions can be inaccurate. Check that every detail matches your experience before applying.
Interview practice
Record an answer to a practice question, up to three minutes. The feedback arrives after you finish recording, not while you speak.
What happens to a recording
- Your speech is turned into text by OpenAI's Whisper.
- An AI model from Anthropic reads the transcript, the question and, if you recorded video, up to six still frames taken from it.
- The model writes the feedback and the scores against a written rubric (version 2.0).
If the frames cannot be taken from a video, the evaluation continues from the transcript alone.
The five dimensions
| Dimension | What it looks at |
|---|---|
| Structure | Whether the answer has a clear situation, task, action and result, in a sensible order. |
| Communication | Pace, filler words, fluency and clarity of delivery. The instructions treat about 130 to 160 words a minute with almost no filler words as the top range. |
| Content | Whether the answer is specific and relevant, says what you did yourself, and gives measurable results. |
| Presence | With video: eye contact and posture, estimated from the still frames. Without video: whether the language sounds assertive and takes ownership rather than hedging. |
| Readiness | Composure: hesitation, long pauses and clusters of filler words, and how quickly you recover. |
The overall score
Each dimension is scored from 0 to 100. The overall score is a weighted average that the instructions set: Structure 20%, Communication 20%, Content 25%, Presence 15% and Readiness 20%. The model writes the numbers and YAGO stores them as returned.
The overall score falls in one of five bands: Standout (94 to 100), Interview-Ready (85 to 93), Strong (65 to 84), Building (50 to 64) and Developing (0 to 49).
What is counted and what is judged
YAGO's code counts the words in your answer and works out a rough pace from the length of the recording. Filler words, pauses, whether each part of a STAR answer is present, and eye contact are the model's judgement from the transcript and the frames. The transcript has no timestamps, so pause lengths are estimates.
Very short answers
An answer of fewer than 30 words is not sent to the model. It is scored 30 on every dimension, in the Developing band, with a note that it was too short to assess.
Language
The feedback is written in the language of the YAGO interface when you upload the recording, English or Romanian. Quotes from your answer stay in the language you spoke.
What it does not do
It does not check whether what you say is accurate, and it does not compare you with other candidates.
Career direction
A short assessment that suggests career directions to test, based on your CV and your answers.
How the result is produced
- You answer 9 to 14 questions. Seven to twelve come from a fixed set of 36 questions about energy, behaviour, trade-offs, constraints and values, picked one at a time to separate the remaining possibilities. Two are open follow-up questions worded by an AI model.
- An AI model first reads your CV to set starting signals about your experience. Your answers then move those signals by fixed amounts. Free-text answers are read by an AI model, and the amount it can move a signal is capped.
- The ranking is done by fixed rules in code, not by AI. Your signals are compared with 14 career families and 48 roles, weighing how well they fit and how much evidence your CV gives.
- An AI model writes the explanation, including "The case against" and "What we cannot know yet".
What the labels mean
| Label | What it means |
|---|---|
| Direct fit | Your CV shows clear evidence and the fit is close. |
| Bridge | A good fit with less evidence so far, or evidence with a looser fit. |
| Stretch | A fit you have little evidence for yet. |
What to keep in mind
The result is a hypothesis to test, not a psychometric test score and not a diagnosis. The ranking is repeatable for the same signals, but the signals themselves come partly from an AI model reading your CV and your free text. You can retake the assessment after 30 days.
Limits that apply to everything
YAGO's Terms of Service say that scores, feedback and recommendations "are generated by automated systems and are provided for guidance only" and "are not guarantees of any outcome, including employment".
AI models from OpenAI and Anthropic process the text, audio or images that YAGO sends them to produce these results. The Privacy Policy names both providers.
None of this compares you with other people, and none of it is an employer's assessment.
If a result looks wrong
- Read the suggestion against your own experience first. You decide what to change.
- If you have scans or attempts left, run it again. A new run can give a different result.
- Tell us. There is no report button on a result page yet. Use "Share your feedback" in the footer of this page, and say which tool, roughly when, and what looked wrong. YAGO does not attach the result to the message, and the form goes to the YAGO team: it is not a formal appeal process.
Your rights over your data, such as access and deletion, are set out in the Privacy Policy.
Versions and dates
| Tool | Version | Note |
|---|---|---|
| Interview practice | Rubric 2.0 | The rubric version is stored with each evaluation. |
| Career direction | Question and role set 2026.09.1 | The version is stored with each run. |
| CV scan | Not versioned | The scoring instructions do not have a version number yet. This page describes them as they were on the date below. |
This page was checked against the product on 29 September 2026. A change to how a score is produced is a reason to update it and this date.
Try it on your own CV
Your first CV scan is free.