Updated
Mock Interview Data: 4,415 AI-Scored Practice Sessions
On this page
Confetto's AI-scored mock interviews show where answers lose points. Across 4,415 practice sessions from 592 applicants, October 3 to November 1, 2026, Content was the weakest of the three rubric parts (74.7% of its points earned). Role-play, ethics, policy and situational questions averaged 69.1 to 74.0, against 78.2 to 79.8 for behavioral, "why this career" and questions about the applicant, and the rubric flagged a missing core point in 30.3% to 41.5% of them, against 9.1% to 15.4%. For applicants with 3 or more sessions, sessions under 150 spoken words scored 6.3 points below the same applicant's average, partly because Confetto's scorer caps very short answers. The scores are Confetto's AI rubric scores, not real interview results. Data through November 1, 2026.
Each finding below lists its result, sample, dates, method and caveat. Every score is Confetto's AI rubric score, out of 100.
1. Content lost the most points
Applicants earned 74.7% of Content points, against 76.4% for Structure and 80.8% for Communication. (4,415 sessions)
| Part | Points earned | Weakest part | Points lost |
|---|---|---|---|
| Content & Completeness50 points | 74.7% | 44.9% | 54.1% |
| Structure & Organization25 points | 76.4% | 38.8% | 25.3% |
| Professional Communication25 points | 80.8% | 6.2% | 20.5% |
- Result
- Content was the weakest part on its own in 44.9% of sessions. Structure was the weakest part on its own in 38.8% and Communication in 6.2%. In the other 10.1%, two parts tied for lowest. Content carries half the points, so it accounts for 54.1% of all points lost.
- Sample
- 4,415 sessions from 592 applicants. Every session in the base set.
- Dates
- October 3 to November 1, 2026. The whole snapshot, all scored on the rubric and scoring model in use since October 3, 2026. The scorer's instructions were revised during the window (see Methods), and Content was the weakest part both before and after the October 18 change (74.4% and 75.3% of its points).
- Method
- Percent of each part's maximum. Each part's score is divided by its maximum (Content 50 points, Structure 25, Communication 25), then averaged over sessions. Points lost is each part's maximum minus its score, as a share of all points lost.
- Caveat
- A rubric's reading, not an interviewer's. These are Confetto's AI rubric scores. They say nothing about how any school or program scores applicants.
2. Scenario questions scored lowest
Role-play stations averaged 69.1 and ethics scenarios 72.7, against 79.8 for questions about the applicant: a gap of 10.7 points. (169 to 945 sessions per type)
| Question type | Sessions | Average | Later sessions |
|---|---|---|---|
| Role-play stationsacting out a conversation | 217 | 69.1 | 69.2 |
| Ethics scenariosa dilemma to reason through | 856 | 72.7 | 73.1 |
| Policy and opinion questionshealth care issues | 290 | 73.5 | 73.7 |
| Situational questionswhat would you do | 326 | 74.0 | 73.9 |
| The professionprofessionalism and communication | 169 | 75.1 | 75.6 |
| Behavioral questionsteamwork, leadership, conflict | 475 | 78.2 | 78.6 |
| Why this careerwhy medicine, PA, nursing | 447 | 79.2 | 79.3 |
| Questions about the applicantyourself, goals, fit | 945 | 79.8 | 80.1 |
- Result
- Role-play, ethics, policy and situational: 69.1 to 74.0. Behavioral, "why this career" and questions about the applicant averaged 78.2 to 79.8, and questions about the profession 75.1. Most of the gap is in Content: role-play sessions earned 66.7% of Content points, against 79.2% on questions about the applicant.
- Sample
- 169 to 945 sessions per type. Types with fewer than 50 sessions (clinical and curveball questions) and sessions on questions with no topic tag are left out.
- Dates
- October 3 to November 1, 2026. The whole snapshot.
- Method
- Topic tags, grouped into types. Each question's topic tags are grouped into types, the first matching group winning, and "why this career" is read from the wording. The overall scores are averaged per type. With each applicant's first session left out, the order holds: 69.2 to 80.1, role-play still lowest. Before and after the October 18 scorer change, the four scenario types stayed below the other three and role-play stayed lowest (69.1 and 69.4).
- Caveat
- Averages, not a ranking of applicants. Applicants choose what to practice, so each type's group is self-selected. The averages do not show how real interviewers weigh each type.
3. The rubric flagged missing core points most on scenario questions
The rubric flagged at least one missing core point in 20.0% of sessions, from 41.5% of role-play sessions to 9.1% of questions about the applicant. (4,415 sessions)
| Question type | Sessions | Missed a core point |
|---|---|---|
| Role-play stations | 217 | 41.5% |
| Ethics scenarios | 856 | 33.2% |
| Policy and opinion questions | 290 | 30.3% |
| Situational questions | 326 | 32.8% |
| The profession | 169 | 17.8% |
| Behavioral questions | 475 | 15.4% |
| Why this career | 447 | 13.0% |
| Questions about the applicant | 945 | 9.1% |
- Result
- 30.3% to 41.5% on scenario questions. On behavioral, "why this career" and questions about the applicant, the rubric flagged one in 9.1% to 15.4% of sessions. Confetto's rubric expects more parts from a scenario answer: information gathered first, the people and risks weighed, and a decision with reasons.
- Sample
- 4,415 sessions. The same sessions and types as finding 2.
- Dates
- October 3 to November 1, 2026. The whole snapshot.
- Method
- Sessions with a missing-point flag. The rubric lists strengths, things to improve and core points the answer left out. A session counts when it has at least one core point flagged as missing.
- Caveat
- What one rubric expects. Core points are the ones Confetto's rubric looks for, and another interviewer may value different points. The share flagged fell after the October 18 and 20 scorer changes, from 24.3% of sessions before October 18 to 13.7% after, and the four scenario types were flagged most in both.
4. Sessions under 150 spoken words scored lowest
Sessions under 150 spoken words averaged 70.1, against 77.9 to 79.7 for 150 to 600 words. (162 short sessions)
| Spoken words | Sessions | Average | Against own average |
|---|---|---|---|
| Under 150 words | 162 | 70.1 | -6.3 |
| 150 to 300 words | 461 | 77.9 | -0.4 |
| 300 to 450 words | 460 | 79.7 | +1.3 |
| 450 to 600 words | 471 | 78.8 | +0.8 |
| 600 words or more | 2,861 | 76.0 | 0.0 |
- Result
- 6.3 points below the applicant's own average. Against each applicant's own average, sessions under 150 words scored -6.3. From 150 words up, length made little difference (-0.4 to +1.3). Sessions of 600 words or more averaged 76.0, and against the same applicant's average they scored 0.0.
- Sample
- 162 sessions under 150 words. The own-average comparison uses applicants with 3 or more sessions: 143 short sessions from 88 applicants.
- Dates
- October 3 to November 1, 2026. The whole snapshot.
- Method
- Every word the applicant spoke. Words are counted from the speech-to-text transcript across the whole session, the first answer and every follow-up. Every session in the base set has at least 60.
- Caveat
- Partly the scorer's own length rule. Confetto's scorer caps the score of a very short session (at most 68 under 80 words and 78 under 120) and adjusts scores by up to 3 points for answer length, including points off long answers. So part of this gap is that rule. With the 47 sessions sitting at a cap left out, sessions under 150 words still scored -7.0 against the applicant's own average (102 sessions from 66 applicants). A very short session also often skipped a step of the answer.
Methods and caveats
The base set. Every finished mock interview session on Confetto scored from October 3 to November 1, 2026 on the standard three-part rubric (Content & Completeness 50 points, Structure & Organization 25, Professional Communication 25), with a score above 0 and at least 60 spoken words: 4,415 sessions from 592 applicants. A session is one interview question plus the AI interviewer's follow-ups. 69% were on medical school questions, the rest on PA, residency, dental and other health program questions.
What was left out. Internal, admin, test and agent accounts, guest sessions, deleted accounts and accounts of other products. Sessions scored 0, under 60 spoken words, or on another rubric (oral boards, UK clinical exams). Groups under 50 sessions are not shown.
What the scores are. Confetto's AI rubric scores, out of 100. They are not real interview results and say nothing about admissions decisions.
Scorer changes during the window. The rubric's three parts and their weights and the scoring model stayed the same from October 3, 2026. The scorer's instructions changed several times. October 18: answer length judged per answer instead of per session (before, 1 to 3 points came off any session over 750 words, after, off answers averaging over 150 words), content bands to tell strong answers from adequate ones, ethical positions judged by their reasoning, and premises the interviewer invented no longer counted against the answer. October 19: each question type's answer shape added to the scorer's guide. October 20: the content limit reworked to fix unfairly low scores, with no change of scale. October 26: named interview red flags (such as a résumé-style "tell me about yourself") cap Content on questions about the applicant. October 30: no praise for red-flag content, clinical findings in role-play count for safety, an integrity cap for falsifying records or gaming the grader, and tips no longer written as labels. November 1: transcript fixes and a rude aside costing points in its own part rather than through the integrity cap. The length cap on very short sessions was in place the whole time. Findings 1 to 3 hold on both sides of the October 18 change: 2,585 sessions scored before it and 1,803 after.
How the numbers were checked. A read-only pull of production data on November 3, 2026, computed twice by two separately written programs that matched on every number, after a second read of every scored session through another query to confirm the pull was complete.
Limits. Applicants choose how much and what to practice, so groups are self-selected. Question types come from topic tags. Only totals and averages are reported, never a person.
How to cite
Confetto, Mock interview data, November 2026, confetto.ai/data. Data from October 3 to November 1, 2026, published November 3, 2026. Licensed under CC BY 4.0 (creativecommons.org/licenses/by/4.0): free to reuse with credit to Confetto and a link to this page.
Please describe the numbers as coming from Confetto's AI-scored mock interviews. More of our practice data: answer length and retries.
Quotable lines
To cite: Confetto, Mock interview data, November 2026, confetto.ai/data. License: CC BY 4.0, free to reuse with credit and a link.
- Across 4,415 of Confetto's AI-scored mock interview sessions from 592 applicants, October 3 to November 1, 2026, applicants earned 74.7% of Content points, against 76.4% for Structure and 80.8% for Communication. Content was the weakest part on its own in 44.9% of sessions and accounts for 54.1% of all points lost.
- In Confetto's AI-scored mock interviews (October 3 to November 1, 2026), role-play stations averaged 69.1 out of 100 and ethics scenarios 72.7, against 79.8 for questions about the applicant. 169 to 945 sessions per type. With each applicant's first session left out, role-play is still lowest.
- Confetto's AI rubric flagged a missing core point in 41.5% of role-play sessions and 33.2% of ethics sessions, against 9.1% on questions about the applicant (4,415 sessions, October 3 to November 1, 2026). 20.0% of all sessions had at least one missing core point.
- In Confetto's AI-scored mock interviews (October 3 to November 1, 2026), sessions under 150 spoken words scored 6.3 points below the same applicant's average, while longer sessions scored within 1.3 points of it. 143 short sessions from 88 applicants with 3 or more sessions. Confetto's scorer caps very short sessions, so part of the gap is that rule.
Frequently asked questions
Which part of an interview answer loses the most points?
Content, in Confetto's AI-scored mock interviews. Across 4,415 sessions, applicants earned 74.7% of Content points, against 76.4% for Structure and 80.8% for Communication. Content carries half the points, so it accounts for 54.1% of all points lost.
Which interview questions are hardest?
Scenario questions, in our data. Role-play stations averaged 69.1, ethics 72.7, and policy and situational questions 73.5 and 74.0, against 78.2 to 79.8 for behavioral, "why this career" and questions about the applicant. The order holds with each applicant's first session left out, and before and after the October 18 scorer change.
How often do applicants miss a key point?
Confetto's AI rubric flagged at least one missing core point in 20.0% of sessions: 41.5% of role-play sessions and 33.2% of ethics sessions, against 9.1% on questions about the applicant.
Do short interview answers score lower?
Very short ones did. Sessions under 150 spoken words scored -6.3 against the same applicant's average. From 150 words up, length made little difference (-0.4 to +1.3). Part of the gap is the scorer's own rule, which caps very short sessions, and short sessions often skipped a step.
Are these real interview results?
No. They are Confetto's AI rubric scores from mock interviews practiced on Confetto. They are not real interview results and say nothing about admissions decisions.
How should I cite this data?
Confetto, Mock interview data, November 2026, confetto.ai/data. Please describe the numbers as coming from Confetto's AI-scored mock interviews, October 3 to November 1, 2026. The data is licensed under CC BY 4.0, so you can reuse it with credit to Confetto and a link to confetto.ai/data.
Does the data include personal information?
No. The page reports totals and averages only, for groups of 50 sessions or more. Internal, test and deleted accounts are left out, and no person or single session can be identified.