A Structured Interview Prompt That Builds Questions and a Rubric

Most interviews aren't really interviews. They're conversations where one person decides in the first five minutes and spends the next forty-five looking for reasons. Different interviewers ask different questions, score on different scales, and then meet to "compare notes" that don't compare. The candidate who advanced was often just the one who built rapport fastest. That's not a hiring signal. That's a vibe.

A structured interview prompt is the cheapest fix for this. It takes the competencies a role actually requires and generates two things together: the behavioral questions that probe each competency, and a scoring rubric with anchors so every interviewer reads an answer the same way. Same questions, same scale, same definition of a "4." The comparison finally compares.

The keyword is together. A pile of interview questions is easy to find. A rubric with behavioral anchors mapped to those questions is what almost nobody ships, and it's the half that makes the interview structured instead of just scripted.

Why the interview-question lists don't help interviewers

Search "ChatGPT interview prompts" and nearly everything is for the candidate. Novoresume's 45-prompt roundup is explicitly about helping applicants "practice answering tough questions" and "boost your confidence" (novoresume.com). Useful if you're job-hunting. Useless if you're the one designing the loop.

The few recruiter-facing prompts get closer and still fall short. The PMaps behavioral-interview prompt does target HR teams and scores answers on a 1-to-5 scale, but it offers no behavioral anchors — no examples of what a "5" versus a "2" actually looks like for any criterion (pmapstest.com). So the scale is a number with no shared meaning. Two interviewers will score the same answer differently because nothing told them what each level means.

That's the gap. Not "generate interview questions" — that's solved a hundred times over. The gap is competency mapping plus anchored scoring, the parts that turn a question list into a defensible hiring instrument.

What a structured interview prompt produces

• A competency list derived from the actual role, not a generic template
• Behavioral questions mapped to each competency, so every question has a purpose
• A scoring rubric with a written anchor for each level, not a bare 1-to-5 scale
• Follow-up probes for each question, so a thin answer gets a fair second chance
• A consistent kit every interviewer on the panel uses, in the same order
• A scorecard structure that makes the debrief a comparison, not a memory contest
• "Red flag" and "strong signal" notes per competency, so good answers are recognized

Anatomy of the interview prompt

Variables
  {{role_title}}        — the role being hired for
  {{competencies}}      — the 4-6 things the role actually requires
  {{seniority}}         — level, to calibrate the anchors

Prompt
  Role: interview designer building a structured kit for {{role_title}}.
  Task: for each of {{competencies}}, produce questions AND a rubric.
  Rules:
    - Every question must map to one named competency.
    - Every competency gets a 1-5 rubric with a BEHAVIORAL ANCHOR per level
      (what a 1, 3, and 5 answer actually sounds like at {{seniority}}).
    - Questions are behavioral ("Tell me about a time..."), not hypothetical.

Output contract (per competency)
  - Competency name + why it matters for the role
  - 2 behavioral questions
  - 1-5 rubric, each level with a concrete behavioral anchor
  - 2 follow-up probes
  - Red flag / strong signal notes

The behavioral anchor per level is the line the question-list prompts skip. "Rate clarity 1 to 5" is useless across a panel. "A 5 walks through the decision, the trade-off they rejected, and the measured outcome; a 2 describes what happened with no decision visible" — that anchor makes two strangers score the same answer the same way.

Anchors beat scales, every time

Here's the take worth defending: a five-point scale with no anchors is worse than no rubric at all, because it looks rigorous while doing nothing. Everyone scores against their private idea of "good," then the numbers get averaged as if they meant the same thing. They don't. You've added false precision to a gut call.

Anchors fix it by forcing the prompt to define each level in observable behavior. Not "strong communication" but "names the stakeholders, states what they disagreed on, explains how they resolved it." Now the rubric describes what to listen for, the interviewer scores what they heard, and the debrief argues about evidence instead of impressions.

In practice, Claude writes tighter anchors when the prompt asks for the behavior at each level rather than an adjective. Ask for "what a level-3 answer sounds like" and you get a usable anchor. Ask for "describe a level-3 candidate" and you get vague praise. GPT-4o needs the seniority restated near the rubric instruction, or it anchors everything to a generic mid-level bar regardless of the {{seniority}} you set.

Step-by-step: building the kit

1. Name the competencies, not the nice-to-haves

Pick four to six things the role genuinely lives or dies on. "Stakeholder management," "incident judgment," "writing for clarity." Resist the urge to list ten — a ten-competency interview is unscoreable in an hour. This is the input that shapes everything downstream.

2. Set seniority honestly

A level-3 answer for a senior role is a level-5 answer for a junior one. Tell the prompt the actual level so the anchors calibrate. Skip this and the rubric over-rewards juniors and under-rewards seniors.

3. Run it and read the anchors

The questions will look fine on the first pass. The anchors are where to look hard. If a "5" anchor is just "excellent answer," it failed — push the prompt to describe the observable behavior. Re-run with "make each anchor describe what the candidate actually says."

4. Distribute the kit, not just the questions

Hand every interviewer the full kit — questions, anchors, probes — and assign competencies so two people aren't both grilling the same thing. The structure only works if the panel uses it.

5. Debrief against the rubric

In the debrief, every score points to an anchor and an answer. "I gave a 4 because they named the trade-off but not the outcome." That's a conversation you can actually resolve. The kit turned the loop into evidence.

Variables you'll set

Getting started

1. Write the four-to-six competencies that matter for the role.
2. Copy the prompt anatomy above and keep the anchor-per-level rule intact.
3. Run it, then stress-test the anchors — vague anchors mean an unscoreable rubric.
4. Distribute the full kit to the panel and split competencies across interviewers.
5. Score every answer against an anchor, and run the debrief on evidence.
6. For the production kit with anchored rubrics pre-built, use a dedicated pack.

A structured interview is the middle of the hiring loop. It works best when the front and back are just as disciplined — pair it with a fair first pass from the Candidate Screening Agent Pack so the right people reach the interview, and feed the notes into the Discovery Interview Synthesis Prompt Pack to turn raw transcripts into a structured read for the debrief.

The thread through all of this is that structure beats a clever one-off prompt — the same reason the candidate screening prompt ships a scorecard instead of a verdict, and the reason a curated pack outlasts a copied snippet, as the prompt packs vs awesome-prompts comparison spells out. Build the kit once, anchor the rubric, and every interview after that is a fair comparison instead of a coin flip.