Why Your STAR Answers Sound Like Everyone Else's Now

Why Does Every Interviewer Hear the Same Story Now?

Because STAR has a public name and a fixed shape, and a fixed shape is the easiest thing in the world for a model to fill. Paste a job posting into any assistant and ask for a behavioral answer, and what comes back is a Situation, a Task, an Action, and a quantified Result, in that order, in clean prose. Everyone applying to a similar role is pasting a similar posting, so a similar story comes out the other side.

Reaching for a model at all is now ordinary rather than an edge case. In a 2026 Greenhouse survey of 1,700 jobseekers in the UK, Ireland and Germany, 78% said they use AI to tailor their CV or application materials at least some of the time ¹. That figure is self-reported, it covers application materials rather than interview prep, and it is not a population estimate for anywhere. What it does establish is access: the paid session or the editor in the family that used to separate the well-coached candidate from everyone else is now a free, always-open prompt sitting behind almost everyone applying alongside you. Your strongest story and your weakest one arrive sounding equally polished, because polish was never the thing worth measuring.

That convergence tends to cluster by occupation rather than being uniform everywhere. A model reaches for the most-documented version of a task in a given field, so marketing candidates land on a similar campaign-lift story and engineering candidates land on a similar outage-under-pressure story. Recognizing which cluster your field falls into is a useful check on your own prep: if the story you're planning to tell sounds like the obvious one for your role, it probably is, and it's worth picking a less obvious project instead.

What Does Structure Still Buy You?

Structure still matters, just not as the differentiator it used to be. A 2022 re-analysis of the personnel-selection literature put structured interviews at .42 and unstructured ones at .19, roughly doubling what an interview predicts about later job performance ². Those are correlations pooled across many jobs, not a promise about any one interview, and none of the studies involve AI. The gap is still a reason to keep the STAR shape rather than drop it.

What structure buys is a fair comparison across candidates, not a way to stand out among them. An interviewer who scores you on the same four beats as everyone else is running a better process than one improvising a new conversation each time. Your job is to fill that format with something a template can't produce, which is the actual texture of a decision you personally made.

So "generic" means something narrower here than it sounds. A generic answer isn't one that follows Situation, Task, Action, Result; almost every good answer does. It is one where the Action and the Result would be true of a dozen different people in a dozen different companies, because nothing in it required you specifically to have been there. The shape was never the tell. The absence of anything only you would know is.

Keep Three Stories, Not Ten

Ten shallow stories cannot survive a single hard question each. Three deep ones can survive several. Pick the projects you know so well you could answer any angle on them cold, and stop trying to have a pre-written answer for every possible competency on a career-center list.

For each of the three, rehearse four layers most candidates never reach:

• The option you rejected. What else could you have done, and why didn't you? A generated story has no discarded option, because it never considered one.
• The number under the number. If you cite a result, know the baseline it's measured against and the period it covers.
• Who disagreed with you. A named disagreement is checkable and specific in a way a smoothed-over narrative never is.
• What actually went wrong. Every real project has a bad draft or a wrong turn. A generated story skips straight to the clean result.

This is preparation you do once per story, not once per company, which is also why three deep stories cost less time than ten shallow ones.

A model is still useful in this narrower job. Feed it your own retelling of a project and ask for five follow-up questions that would be hard to answer if you hadn't done the work yourself. Answering those out loud, alone, before the call, is a better use of the tool than asking it to write the story in the first place, and it produces exactly the depth a generated answer cannot fake.

Prepare for the Follow-Up, Not the Script

The second question is where the round now carries its information, and what it asks for is the part a script never holds: the option you dropped, the baseline under the number, the person who disagreed. None of this is about anyone spotting that a model touched your prep. What a follow-up finds is not authorship. It's whether anything sits underneath the story.

Spotting is a weak instrument in any case. In a 2021 study, untrained readers told AI-generated text from human writing at chance, and three brief training methods lifted them only to about 55% ³. Those were crowdworkers reading short passages rather than interviewers reading work in their own field, so read it as a limit on casual spotting rather than proof that nobody can ever tell.

What follow-up questions expose whether someone understands the answer they just gave? lays out the pattern from the other side: specify which alternative you ruled out, invert what would have happened if a constraint moved, and say what would have to be true for your own answer to be wrong.

Rehearsing for that pattern looks different from rehearsing a script. Instead of practicing the words of the answer, practice being asked "why that instead of the alternative" about your own project until the response is automatic. Why does every candidate suddenly give the same polished STAR answer? is what employers are told on this exact ground, and it is blunter than the prep sites: piling on more follow-ups makes things worse on its own, because extra turns give a rehearsed story more room, so the advice is to ask for something checkable and to move a round from describing work to doing it. Depth is the thing that survives both.

See how it works >