How to Stop Saying Um, Uh and Like
Published Updated 8 min read
Short answer
Stop using filler words by replacing each one with a closed-mouth pause rather than trying to delete it. An um is a placeholder produced while the sentence is still being planned, so silence has to take its place. Under 3 fillers per minute reads as normal speech; above 8 they become the content.
You reduce filler words by getting comfortable with silence, not by trying to delete the fillers. Every um is a placeholder your mouth produces while your brain is still assembling the sentence, so removing it without replacing it leaves nothing to fill the gap and the um comes straight back. The method that works is to swap each filler for a deliberate closed-mouth pause, one situation at a time, and to track the rate rather than judge the feeling. Under 3 fillers per minute reads as ordinary speech. Above 8 is where listeners start hearing the fillers instead of the point.
Why you say um in the first place
Filler words are not a bad habit in the way biting your nails is a bad habit. They are a signal that speech production has outrun sentence planning. You committed to starting a sentence before you had finished deciding how it ends, and the um buys the fraction of a second you need to catch up.
This is why fillers cluster in predictable places rather than scattering evenly. They appear before numbers, before names you are not certain of, at the start of an answer to a question you were not expecting, and at the seam between two ideas where you have to choose which one goes next. Look for those four positions in your own recording and you will find most of your fillers sitting in them.
It also explains why fillers get worse under pressure and better when you are describing something you have explained fifty times. The vocabulary is identical. The planning load is not.
A filler and a thinking pause are the same moment
Both occupy the same instant. Only one costs you anything. A thinking pause is silence while you decide what comes next; a filler is noise while you decide what comes next. The listener uses that moment identically either way, which is the part most people get wrong: they believe silence will read as being lost, when in practice it reads as being deliberate.
Recordings make this obvious in a way live conversation never does. A one-second silence feels enormous from the inside and passes almost unnoticed from the outside. The reason it feels enormous is that you are the only person present who knows you had not finished planning.
So the target is not fewer words. It is the same moment, spent silently. That reframing moves more speakers than any amount of trying to catch the um mid-flight.
The replace-with-silence method
Record sixty seconds answering a question you have not rehearsed, and count your fillers. Do not try to fix anything on this pass. You need a number, and you need it from unedited speech.
Listen back and note where each filler landed. You are looking for position, not count: before a number, at the start of an answer, at the seam between ideas. Most people find two positions account for the majority.
Record again on the same prompt with one instruction: when you feel the filler coming, close your mouth. Closing the mouth matters mechanically - um and uh are both produced with the jaw open and the vocal folds already running, and a closed mouth cannot make either sound.
Expect the second recording to feel slow and stilted. It will not sound that way. Compare the two on playback rather than from memory, because memory reports how it felt to produce and playback reports what a listener received.
Repeat daily for a week on unrehearsed prompts. Most people roughly halve their rate in that time and then plateau. The plateau is the point where the remaining fillers are structural, and need the pace and structure work rather than more of this drill.
Five passes. The whole method is one substitution, repeated until it stops requiring attention.
- Record sixty seconds answering a question you have not rehearsed, and count your fillers. Fix nothing on this pass - you need a number from unedited speech.
- Listen back and note where each filler landed, not how many there were. Look for the four positions: before a number, before a name, at the start of an answer, at the seam between two ideas.
- Record the same prompt again with one instruction: when you feel the filler coming, close your mouth. A closed mouth cannot produce um or uh - both need the jaw open and the folds already running.
- Compare the two recordings on playback rather than from memory. The second will feel slow to produce and will not sound slow to a listener.
- Repeat daily for a week on prompts you have not seen. Most people roughly halve their rate, then plateau where the remaining fillers are structural.
Numbers to aim for
Under 3 fillers per minute is normal speech. Nobody notices, including you on playback. There is no benefit to chasing zero, and speakers who reach zero usually get there by over-rehearsing, which costs more in warmth than the fillers cost in credibility.
3 to 8 per minute is where a critical listener registers something without being able to name it. Most nervous speakers sit here.
Above 8 per minute the fillers become the content. In a pitch or an interview this is the level at which people report afterwards that the speaker seemed unprepared, even when the substance was strong.
Track rate rather than count, because rate survives comparison across recordings of different lengths. Twelve fillers in ninety seconds and eight in sixty are the same problem; the raw counts suggest otherwise.
| Fillers per minute | What a listener registers | What to do |
|---|---|---|
| Under 3 | Nothing. This is ordinary speech | Leave it alone - chasing zero costs warmth |
| 3 to 8 | Something is off, without being able to name it | The closed-mouth drill, daily, for two weeks |
| Above 8 | The fillers instead of the point; reads as unprepared | Fix pace first, then the drill |
Mistakes that keep the rate high
Substituting a different filler. Replacing um with so, right or actually does nothing, because the listener registers the hesitation rather than the specific syllable. If your um count drops while your so count rises, you have moved the problem rather than solved it.
Practising on rehearsed material. Reading a prepared paragraph aloud produces almost no fillers regardless of skill, which makes it useless as practice and misleading as measurement. Use questions you have not seen.
Speeding up to escape. Many speakers accelerate to reach the end of a sentence before the planning gap opens. That trades fillers for a pace problem and usually makes comprehension worse.
Working on fillers and pace in the same week. Both are timing habits and they interfere with each other. Fix whichever number is further from target, then move to the other.
A two-week drill
Days 1 to 3: one unrehearsed minute daily, no correction attempted. You are establishing a baseline and getting past the discomfort of hearing your own voice, which takes about three sessions for most people.
Days 4 to 10: the same daily minute, with the closed-mouth substitution as the only instruction. Note the rate each day. It will not fall smoothly, and days after poor sleep will be visibly worse, which is worth seeing rather than hiding.
Days 11 to 14: raise the pressure. Answer something you would genuinely rather not, or set a timer forcing you to start within two seconds of reading the prompt. The rate will rise. The point is to find how much of the gain survives stress, because that is the only part that will be there in the meeting.
Questions
Why do people say um?
Because speech production runs ahead of sentence planning. You start a sentence before you have finished deciding how it ends, and the um fills the gap while you catch up. It is a timing artefact rather than a vocabulary problem, which is why people with large vocabularies say um just as often under pressure.
Are filler words actually bad?
In ordinary conversation, no. Under about 3 per minute they are invisible and arguably make speech sound more natural. They start costing you above roughly 8 per minute, and they cost most in high-stakes settings where a listener is already deciding whether you know your material.
How long does it take to reduce filler words?
Most people roughly halve their rate within two weeks of daily one-minute recordings, because the substitution is a motor habit rather than a skill. Going from halved to consistently under 3 per minute under pressure takes longer, usually six to eight weeks.
How do I measure my filler words?
Record yourself answering an unrehearsed question for sixty seconds, transcribe it, and count um, uh, like, you know, so, actually and basically. Divide by the length in minutes for a rate you can compare across recordings. FounderVoice does the transcription, the count and the timestamps automatically.
Does slowing down reduce filler words?
Usually yes, because a slower pace lets sentence planning keep up with production. But it is a side effect rather than a method, and speakers who slow down without addressing the discomfort with silence often just spread the same number of fillers across more time.