Most STAR answers ramble in Situation and lose the Result. A good improvement prompt diagnoses where the story sags, sharpens each segment, and demands a measurable, non-generic Result. These 12 prompts take an answer you already have and make it land in under two minutes.
TL;DR
- Target length is one to two minutes spoken; career coaches advise keeping the whole answer under two minutes, with the Action phase taking roughly half. Situation should be two to three sentences (about 20%), Task one sentence (about 15%), Result one to two sentences with numbers.
- These prompts edit a real story you already lived. Use them to compress, quantify the Result, surface “I” over “we”, and map one story to many questions.
- Tool note (June 2026): ChatGPT (GPT-5.5) is the stricter structure-and-metrics editor; Claude (Sonnet 4.6) writes the more natural spoken cadence. Paste your real prior answers so the rewrite keeps your voice, not the model’s.
- Never let the model invent numbers, dates, or titles. Ask it which numbers you have, then fill them in yourself.
Who this is for
Job seekers refining a story bank, candidates prepping for senior interviews, and anyone whose STAR answers run past three minutes. If you have no stories written yet, start with raw notes first, then bring them here.
When not to use these prompts
Don’t run these on stories you haven’t lived. AI sharpens what you give it; fiction stays fiction, and interviewers probe for detail the moment an answer sounds too clean.
What “good length” looks like
Recruiters consistently advise one to two minutes per behavioural answer, with the bulk of the time spent on what you actually did:
| Segment | Share of answer | Sentences |
|---|---|---|
| Situation | ~20% | 2-3, set the stage only |
| Task | ~15% | 1, your specific responsibility |
| Action | ~50% | 4-6, first-person verbs |
| Result | ~15% | 1-2, with at least one number |
The most common failure is over-explaining the Situation. The prompts below default to this allocation.
Prompt anatomy
Every STAR improvement prompt should carry six elements:
- Role: candidate, hiring manager, or recruiter — name the persona the model plays.
- Context: target role, company, level, your background.
- Goal: one deliverable (a tightened answer, a follow-up list, a tone diff).
- Constraints: word count, banned phrases, must-include facts.
- Tone: confident / curious / measured — pick two anchors.
- Examples: paste one or two of your past answers so the model matches your voice.
A reusable wrapper: tell the model up front Do not invent numbers, names, or dates; if a metric is missing, ask me for it. That one line stops most confabulation.
12 copy-ready prompts
Replace anything in [brackets] with your own text before sending.
1. STAR sag diagnosis
My STAR answer: [paste answer]. Diagnose which segment sags: Situation too long? Task vague? Actions plural ("we did...") not personal ("I did...")? Result generic? Give a per-segment critique plus one specific fix per segment. Do not rewrite yet.2. Tighten to two minutes
Compress this STAR answer to ~250 words / 2 spoken minutes: Situation in 2 sentences, Task in 1, Actions in 4-6 sentences with first-person verbs, Result in 2 sentences with numbers, then a 1-sentence lesson. Keep my facts; cut filler.3. From “we” to “I”
Rewrite this team-led story to surface my personal contribution: (a) what I decided, (b) what I did versus what others did, (c) where I pushed against the group. Stay honest, do not inflate. Flag any claim that needs a number I have not given.4. Result quantification
My Result is vague: [paste result]. List 5 ways to quantify it: percent change, dollars, time saved, error rate, headcount, NPS or satisfaction. Do not invent figures; for each, ask me the exact number you would need.5. Lesson distillation
Add a closing lesson to this STAR: 1 sentence, first person, about what I would do differently or carried forward. Not "I learned the value of teamwork" — something specific to this situation.6. Follow-up question prep
For this STAR answer, predict 3 follow-up questions an interviewer will ask. For each: the question plus a 30-second clarifying reply that answers it directly. Do not respond with another story.7. One story, many questions
My story: [paste story]. List 6 behavioural questions this story could answer (leadership, conflict, failure, growth, influence, decision-making). For each, write the one-sentence opening I would use to frame it for that question.8. Failure story upgrade
My "tell me about a failure" answer: [paste answer]. Improve it so the failure is real (not "I work too hard"), the lesson is specific, and the recovery shows growth without re-spinning the failure as a secret win.9. Add the obstacle (SOAR upgrade)
Rewrite this STAR as SOAR (Situation, Obstacle, Action, Result). Name the one concrete obstacle I overcame and put it before the Action. Keep all facts; the obstacle must come from my text, not be invented.10. Tone calibration for company culture
This interview is at a company with a [direct, low-context / polite, indirect] culture. Recalibrate this STAR answer for that tone without changing any facts. Show a short before/after diff of what changed.11. Staff / Director-level reframe
Rewrite this STAR for a Staff or Director-level interview: emphasise scope, ambiguity, cross-team influence, and trade-offs rather than IC mechanics. Replace "I built X" with "I scoped X with team Y and accepted trade-off Z".12. Pacing and filler removal
Read this STAR as if spoken aloud. Cut filler ("um", "kind of", "basically"), hedges ("maybe a bit"), restarts, and any sentence over 25 words. Return the leaner version, then list every cut so I can restore anything important.Which model for which job
| Task | Better fit (June 2026) | Why |
|---|---|---|
| Enforce STAR structure, push for metrics | ChatGPT (GPT-5.5) | Holds the format and nudges hard for numbers |
| Natural spoken cadence, reflective answers | Claude (Sonnet 4.6) | Smoother conversational phrasing |
| Persistent coaching with your background | Either, via saved context | ChatGPT custom instructions or a Claude Project |
Both Claude Pro ($20/mo) and ChatGPT Plus ($20/mo) are enough for this work; the free tiers handle a few answers but hit limits during a long prep session. Build a Claude Project (or a ChatGPT custom-instruction profile) with your résumé summary, target roles, and target companies pasted in once — then every prompt above inherits that context instead of starting cold.
Common mistakes
- Treating AI output as the final answer: recruiters spot generic cadence in seconds.
- No specific context (company, role, level): the output stays generic.
- Asking the model to “be honest” without giving your real track record: it confabulates.
- One answer reused verbatim for every company: panels compare notes.
- Listing skills without proof: claims without receipts.
- No tone anchor: the answer lands flat.
- Skipping the fact-check: models invent dates, numbers, and titles.
How to push results further
- Paste real examples; your prior STAR stories anchor the output to your voice.
- Ask the model to play interviewer first; weak answers expose themselves.
- Write three drafts and ship the third (first is generic, second over-corrects).
- Time yourself out loud; a two-minute story beats a four-minute one.
- Always read aloud — written and spoken answers feel different.
- Keep a personal story bank and reuse strong stories across questions.
- Run the final answer past someone in the role; peer feedback beats AI feedback.
FAQ
- Can recruiters tell an answer was AI-written?: Yes, when it has no personal detail. Specific names, numbers, and one human moment are the antidote.
- STAR or SOAR?: STAR is the default. Switch to SOAR (prompt 9) when the question is about overcoming a challenge — naming the obstacle is what hiring managers remember.
- How long should the answer be?: One to two minutes spoken; aim under two, with the Action taking about half.
- How many drafts before I’m ready?: Three for important stories, one or two for the rest.
- Should I use AI on interview day?: Only to calm nerves. Don’t rewrite a prepared answer in the final hour.
- How do I keep the tone authentic?: Paste samples of your own real writing into the prompt so the model matches your phrasing.