How to Stop Saying Um in a University Interview

Why fillers happen

“Um” is a placeholder. It holds your turn in a conversation while your brain finishes deciding what the next few words are. In a chat with a friend that is fine, because the friend waits. On a recording there is no turn to hold, so the placeholder does nothing except go into the transcript.

The reason it shows up so much in interview answers is specific: you are thinking out loud with no plan for the sentence. You know the story, roughly, but you are choosing the words and the order in real time, and every fork in the road (which detail next? how does this sentence end?) produces a filler while you choose. Nerves make it worse by making you start talking before you have chosen anything, because silence feels dangerous. The prep timer makes it worse again, because thirty seconds is enough to panic in and not enough to plan in unless you already know what planning looks like.

So the filler count is a measurement of how many decisions you were making mid-sentence. Reduce the decisions and the count drops. That is the whole method.

The fix is structure, not suppression

Suppression means trying to catch each “um” on the way out. It fails for the same reason “don’t think about an elephant” fails, and it costs attention you needed for the answer. Structure means deciding, before you open your mouth, what the four or five sentences are going to be about, so that the only decisions left while speaking are word choices, which are fast.

Concretely: a story answer is a situation, a thing you did, a result, and a line about what changed. That is the STAR shape, and it is worth learning for this reason alone: if you always know the next beat, you never have to decide it mid-sentence. In a thirty-second prep window you can pick the story, the one number in it, and the last sentence. Those three choices remove most of the forks. Students who do this do not sound rehearsed; they sound like someone who knows what they are about to say, which is a different thing.

The second half of structure is the full stop. Fillers cluster at the ends of sentences that never ended, the ones that trail into “and, um, so yeah.” Decide to end sentences. Say the last word, stop, breathe, start the next one. The silence you are afraid of is about half a second long and reads as composure.

The second version is shorter, contains a number, and has no fillers, and the speaker did not suppress a single one. There was nothing to suppress, because each sentence was decided before it started.

A four-step drill, ten minutes a day

1. Record one answer, cold. Pick a question you have not prepared, give yourself thirty seconds of prep and ninety seconds to answer, on camera, and do not stop or restart. This is your baseline, and it will be worse than you think. That is the point.
2. Count, and mark where. Play it back or read the transcript and count every um, uh, like, so, basically, kind of, and you know. Write the number down. Then look at where they cluster. Almost always it is the opening (before you have decided how to start) and the transitions between beats (before you have decided what comes next).
3. Plan the beats, then record again. Same question. This time, use the prep to write four words on a sticky note: the situation, the action, the number, the last line. Record. Count again. The number usually halves on the second take, and it halves because of the four words, not because you tried harder.
4. New question, same method. The next day, a fresh question you have not seen, thirty seconds, four words, one take, count. Track the count across days on one line of paper. The drill is not about the question; it is about getting the four-word plan to happen automatically inside a prep window.

A week of this, ten minutes a day, moves most students from a count in the twenties to single digits on a ninety-second answer. The count matters less than the trend, and the trend matters less than the moment, usually around day four, when you catch yourself planning the last sentence during the prep timer without deciding to. There is no shortcut through the week. Most advice on fillers, this page included, is about what to do and how; what you cannot avoid is the reps, because the four-word plan only becomes automatic by being done, and each recording comes out a little cleaner than the one before. Do the ten minutes.

Why reading your own transcript is the fastest way to see it

You cannot hear your own fillers. Your brain edits them out in real time, the same way it edits out your own breathing, which is why students are surprised by playback and then surprised again by how much they missed on playback. A transcript defeats the editor. On the page, every “um” is a word, sitting between two other words, and you can see the sentence it was standing in for and the decision you were making at the time. Read one transcript of yourself and the fillers become visible in your speech for good, in a way that a hundred playbacks never quite manage.

This is exactly what AcceptMe’s interview tool does. It asks you a question you have not seen, on the prep and answer clock you set, records one take, and gives you the transcript with the coach’s notes on what you said and how it was structured. The fillers are there to count without counting, and the notes tell you which beat you were missing when they happened. Two runs are free, which is enough to do the baseline and the second take of the drill above and see the number move.

How much filler is actually fine

More than you think. Natural speech, from people who are good at it, runs at something like one filler every twenty or thirty words, and nobody notices. A ninety-second answer of around 200 words with five or six fillers in it sounds like a person talking. The problem starts when fillers arrive in clusters, three in a sentence, or when they replace the ends of sentences, because those are the moments where the listener notices you had not decided what to say.

So the target is not zero. Zero sounds like a script, and reviewers can hear a script. The target is no clusters and no trailing ends: a filler here and there inside a sentence that then finishes cleanly. If your count is under about eight on a ninety-second answer and none of them are in a row, stop working on fillers and work on the content.

What reviewers really score

The person who watches your Kira recording is scoring against a competency rubric: did you name a real situation, say what you did, and show a result or a change in you. Delivery matters only to the extent that it lets them hear that. A clean answer with no story in it scores low; a slightly messy answer with a real situation, a real action and a number scores well. Nobody has a “filler words” row on their rubric, and an applicant who says “um” four times while telling a specific, honest story with a result is in far better shape than one who says it never while describing how much they learned.

That is worth holding onto, because the fear of fillers produces worse answers than the fillers do. Students who are monitoring their own speech for “um” are not monitoring whether the story has reached its result yet, and the clock runs out on a story with no ending. Build the answer, end the sentences, and let the odd “um” through. The reviewer is listening for what you did.

Where to go from here

The structure that makes fillers unnecessary is the STAR method, and the Kira Talent practice guide explains the prep timer, the answer timer and what the person scoring the recording is actually looking for. If the fillers are worst on the opener, the tell-me-about-yourself guide gives you a sixty-second shape to have ready before the first question lands, and the setup guide covers the camera and lighting so that is not what you are thinking about either. Then do the baseline: AcceptMe’s interview tool will give you a fresh question, your program’s clock, one take, and the transcript, and the count on the page is the number you are going to bring down over the next week.