The Difference Between STAR and BARS in AI Interviews

In the world of structured interviewing, two acronyms often dominate the conversation: STAR and BARS. While most recruiting teams are familiar with asking candidates for a Situation, Task, Action, and Result (STAR), fewer apply Behaviorally Anchored Rating Scales (BARS) to formally evaluate those answers.

As AI increasingly steps into the interview process, understanding the distinction between collecting a story and scoring it has never been more important.

What BARS Actually Does

Think about a customer support role and a competency dimension called “handling an escalation.”

• A low anchor might describe a candidate who passes the complaint to a manager without gathering any details.
• A mid anchor describes someone who resolves the issue but never follows up.
• A high anchor describes someone who resolves it, confirms the fix with the customer, and flags the root cause internally.

In this system, a rating isn’t just an interviewer's subjective opinion about the candidate. It is a match between the candidate's actual answer and a concrete behavioral description written before anyone even met them.

Proper BARS development is more involved than just writing three sentences. It requires collecting real examples of effective and ineffective behavior from people who know the job, grouping them into dimensions, and keeping only the examples that experts independently agree belong there. That agreement step is what separates a scientifically valid scale from a nicely worded one.

STAR Gets the Story, BARS Scores It

These two concepts get mixed up constantly, which makes sense because both show up in the exact same interview.

STAR is about collecting evidence. It pushes the candidate to describe a Situation, the Task they owned, the Action they took, and the Result. This ensures you end up with a specific story instead of a general, hypothetical claim. If the Action part is thin, a good interviewer asks another question until it isn’t.

BARS is what happens next. It is the scoring layer that turns a complete, structured answer into a number you can clearly explain. STAR tells you whether you have usable evidence; anchors tell you how strong that evidence is for this particular role.

Most structured interviews run STAR without anchors and score based on open judgment. This works reasonably well—until you have several interviewers and need their scores to mean the exact same thing. Conversely, anchors without STAR are harder to apply because the answers underneath them are inconsistent. Stacked together, they cover each other’s gaps.

Why This Matters More When AI Does the Scoring

Experienced human interviewers quietly compensate for vague rating scales. They have seen enough candidates to develop an internal benchmark, even if they’ve never written it out.

An AI evaluator has no such implicit benchmark. Every criterion it applies has to exist in writing first, which forces a highly useful kind of honesty upon the hiring team. Plenty of organizations discover during AI setup that their old "scorecard" was really just a list of competency names with numbers (1 to 5) beside them, offering no real guidance.

This structure pays off in the decision-making conversation afterward, too. When a hiring manager asks why a candidate scored poorly on stakeholder management, an anchored scale points directly at a missing behavior instead of just a low number.

There is solid research supporting the broader principle here. Updated estimates published by Sackett et al. in 2022 put the predictive validity of structured interviews at around .42, compared to just .19 for unstructured ones. Anchored scales are a critical piece of that predictive structure.

When It’s Worth Building (and When It Isn’t)

Building robust anchors takes real work, and scales written for a sales role absolutely do not transfer to software engineering.

It usually pays off when you are hiring for the same role repeatedly, when several interviewers need to produce comparable scores, or when the role is critical enough that an inconsistent evaluation comes with high business costs. It is usually unnecessary overhead for a one-off senior executive hire, for roles whose requirements are still being figured out, or for early screening where you are simply confirming basics rather than grading depth.

There is, however, a middle path worth knowing about. Writing two or three clear anchors per dimension—carefully, but skipping the formal academic agreement checks—gets you most of the clarity for a fraction of the effort. That may not be textbook BARS, but a partly anchored scale still vastly outperforms an unanchored one.

What Anchored Scoring Doesn’t Fix

Anchors sharpen the scale, but they don’t settle whether you’re measuring the right thing in the first place.

• Anchors go stale: A scale written for a role as it existed three years ago will quietly mis-score candidates for the role as it exists today.
• Anchors can be written badly: If the top anchor describes seniority or a polished vocabulary instead of actual behavior, you end up rewarding interview skill rather than job competence.
• Consistency isn’t accuracy: A flawed scale applied the exact same way to everyone produces consistent mistakes instead of random ones.
• Review still matters: Anchored output makes human review easier because a reviewer can check the score against the described behavior. That is an argument for reviewing more thoughtfully, not less.

How It Shows Up in Practice

Not every AI interview product uses BARS, and “structured scoring” on a vendor page can mean a lot of different things. If methodology matters to you, ask vendors which framework is applied, whether criteria are defined per role, and whether you can see the score broken down rather than rolled into a single, opaque number.

Coensio’s AI interviews evaluate behavioral answers using a multi-layered approach. STAR is the default framework, meaning the final report shows the candidate's Situation, Task, Action, and Result separately. BARS scoring then places each answer against defined performance levels instead of leaving the rating to open judgment. When a candidate's answer is unclear or incomplete, the interactive interview actively asks a follow-up question aimed at closing that specific gap.

The criteria come ready out of the box, so a team doesn’t have to run a complex BARS development project before their very first interview. Employers who want their own specific evaluation criteria can also customize them. Either way, the criteria are set before the interviews start and applied the exact same way to every candidate. The hiring team then makes the final decision afterward, with deep, objective context in hand.

The Takeaway

Ultimately, BARS is less a technology than a discipline. It forces a team to say out loud what "good performance" looks like in a specific role, clearly enough that a stranger (or an AI) could accurately apply it.

That clarity is worth achieving whether or not software does the actual scoring. AI simply removes the option of skipping it, because a model can only apply criteria somebody wrote down. Recruiting teams that treat this requirement as a chance to sharpen their criteria get a lot more out of their hiring technology than teams that just treat it as another setup screen.

‍