TLDR
- A STAR answer is easy to rehearse, so the second and third question carry most of a behavioral interview. Tracy AI Interviewer asks them for every applicant, at any hour, and each follow-up depends on the answer before it.
- You write the behavioral questions or let Tracy draft them from your brief. When a story sounds too tidy, replay the exact moment: the transcript sits beside the video, and each assessment point opens the part of the recording it rests on.
- The research on structured interviews is about human-run ones. Use it for the design rules, not as a claim about AI accuracy.
Fact-checked on 4 Nov 2026 against Tracy's own pages and the original research summaries. Sources are listed at the end.
Behavioral interviewing asks people what they did, not what they would do. The common frame is STAR: the situation, the task, the action and the result. Candidates learn it, and a prepared story has all four parts. This page covers what STAR misses, the follow-ups that test a story, and how to run them with Tracy AI Interviewer for every applicant, not only the few a recruiter has time to call.
What STAR misses
A STAR answer gives you a shape. It does not tell you whether the story is the candidate's own. Three gaps to look for:
- Whose action was it? "We" covers a lot of ground. The answer may be a team's work with the candidate on the edge of it.
- What did it cost? Tidy stories end in a result with no trade-off, though real projects usually closed off another choice.
- What would they change? Ask for it: people who did the work usually have something they would redo.
These are our observations about how interviews go, not a measured finding. They are why a first round works better as a conversation than as a form. One-way and two-way video interviews explains why a recording cannot ask the second question.
Follow-ups that test a story
Pick follow-ups that ask for something only the person who did the work would know.
| Gap | Follow-up |
|---|---|
| Whose action | "What did you personally decide?" |
| The trade-off | "What did you give up by doing it that way?" |
| The disagreement | "Who disagreed, and what did they want instead?" |
| The failure | "What went wrong, and when did you notice?" |
| The result | "How would you know it worked? What was the number or the signal?" |
| Hindsight | "What would you do differently now?" |
Tracy's product page shows the same move in an example from its site, which uses invented people: a data engineer says reads were the bottleneck, and Tracy asks, "How much faster were reads, and what did it cost you on writes?"
An example chain
This is an illustration we wrote, not a transcript of a real candidate.
Interviewer: Tell me about a time you handled a difficult stakeholder. Candidate: Our head of sales wanted a feature shipped before it was ready. I talked it through and we reached a compromise. Interviewer: What did you personally say that changed their mind? Candidate: I showed them the support tickets from the last release. Interviewer: What did the compromise cost you? Candidate: We shipped half the feature and lost two weeks of the roadmap. Interviewer: How did you know the half release was the right call?
The chain asks for a source, a cost and a test, and each answer lands on the transcript.
Running it with Tracy
- Set the goal and the questions. Start from the job description or your own behavioral questions. Tracy drafts questions that you can edit or replace.
- Pick the language. Each interview is a two-way video interview in one language, chosen at setup.
- Invite everyone. Send a personal invite, publish an apply page, or let your own system hand over the link. Candidates have nothing to install and start whenever it suits them, so every applicant gets the same follow-ups, not only the few a recruiter can call. People are told about the recording before it starts.
- Let the follow-ups run. Tracy listens for what was said, asks for the missing detail, and answers questions about the process. It waits for the end of a thought, so a candidate who is working out a hard example is not cut off. With CV check on, it also asks 3 to 5 questions tailored to the CV, which tests the stories against what the CV claims.
- Open the result. The assessment for each question comes with its reasoning, and the recording plays beside a transcript timed to it. Response timing and tab switches show up as marks on the recording, for a person to check.
How a first round works with an AI interviewer covers the steps in more detail, and software engineer interview questions shows how follow-ups sit alongside technical questions.
Reading the answers
Write down what a good answer contains before you send the interview: for example, the candidate's own action, a trade-off and a result. Then read each assessment against the transcript. Each point jumps to the moment it came from, so you can check it without replaying the whole interview. AI interview reports goes through what a reviewer should look for.
Tracy does not reject anyone: a person decides.
What the research says
Two research summaries are worth knowing, and both describe human-run interviews.
A 1998 meta-analytic summary in Psychological Bulletin (Schmidt and Hunter) put the average validity of structured interviews for predicting job performance at .51, compared with .38 for unstructured interviews. Validity here is a correlation, not a percentage of correct predictions. The figures are statistically corrected estimates from older studies, and a 2022 re-analysis argues that such corrections for range restriction can overcorrect.
A 2022 re-analysis in the Journal of Applied Psychology (Sackett and colleagues) found structured interviews to be the top-ranked selection procedure for predicting job performance. The authors also report that most of the procedures that ranked high in earlier summaries remain high in rank, with mean validity estimates reduced by .10 to .20 points.
That points to a design rule: ask every applicant the same questions and read the answers against the same evidence. The research is silent on AI interviewers, so it proves nothing about their accuracy.
When to use people instead
- Senior roles where the interview is also a negotiation. A person reads the room and sells the role.
- A final-round conversation with the team. Fit with a team is judged by the team.
- Candidates who have not agreed to be recorded. Tracy records every interview and is not for them.
What it costs
The first 3 results are free. After that it is $7 per credit, charged per started 20 minutes for interviews that produced a result, and a session where nobody answers is never charged. There are no seats. If you want to see the options for the first round before you decide, how to screen candidates when there are too many applicants is the place to start.
Questions
What is a behavioral interview?
Can an AI run a behavioral interview?
Why are follow-ups needed on STAR answers?
Does research support structured interviews?
What does the reviewer get after the interview?
How do I start?
Try it free and run a behavioral interview with follow-ups on your first candidates at no cost, or book a demo to see one worked through.
Sources
All sources retrieved 4 Nov 2026.
- Tracy AI Interviewer, Home page: the follow-up example from the site (invented people), integrity signals, first 3 results free.
- Tracy AI Interviewer, Pricing: free results, $7 per credit, billing, no-answer sessions, no seats.
- Tracy AI Interviewer, Product: questions drafted from a brief, CV check, three ways in, no install, people told about the recording before it starts, one interview language chosen at setup, follow-ups that depend on the answer before it, waiting for the end of a thought, points linked to the recording, transcript beside the video, "Tracy does not reject anyone. A person decides.".
- Frank L. Schmidt and John E. Hunter, The Validity and Utility of Selection Methods in Personnel Psychology, Psychological Bulletin, 1998: structured interview validity .51 versus .38.
- Sackett, Zhang, Berry and Lievens, Revisiting meta-analytic estimates of validity in personnel selection, Journal of Applied Psychology, 2022: structured interviews top-ranked.