How to Write a Keynote Speech

The first keynote I wrote was forty-one slides long and had no argument in it. I had a topic, I had things to say about the topic, and I genuinely believed that was the same as having a talk. It was not. I found that out about eleven minutes in, watching a room of two hundred people politely wait for me to get somewhere.

What I know now is that writing a keynote has almost nothing to do with writing. It is a structural problem, and once the structure is right the words are comparatively easy. This is the process I use, in order, and the mistakes it exists to prevent.

Start with one idea

Every keynote that has ever worked on me argued one thing. Not a topic. An argument, with a position that a reasonable person could disagree with. “Community building” is a topic. “Most communities die from consistency problems, not from a lack of ideas” is an argument, and only one of those can be defended for forty minutes.

The test I use is brutal and quick. Say your idea out loud in one sentence, without a preamble. If it takes three sentences, or if it contains the word “and” joining two different claims, you have a topic and you are going to ramble.

Being able to disagree with it matters. If nobody in the room could reasonably hold the opposite view, you are not arguing, you are reciting consensus. That is comfortable, it is safe, and it is why so many conference keynotes are forgotten by lunchtime.

Understand who is in the room

I ask the organiser four questions before I write anything, and the answers change the talk more than anything I could research alone. Who is in the room, what they already agree about, what they are tired of hearing, and what the organiser hopes changes as a result.

That third question is the most valuable and the one nobody volunteers. Every industry has a phrase everyone is exhausted by. Walking into a room and opening with it costs you the first five minutes, and you will not know why the energy dropped.

The fourth question is what stops you writing a talk that is about you. The organiser booked you because they want something to shift. Ask them what, and then build towards it rather than towards your own greatest hits.

If you are still working out how to get in front of these rooms in the first place, I wrote about that separately in how to get speaking engagements.

Write the ending first

This is the single change that improved my talks most. Write the last ninety seconds before you write anything else. Not the summary, the ending: the thing you want the room carrying out of the door.

It works because a keynote is a journey with a destination, and you cannot plan a route without knowing where you are going. Written last, the ending is whatever you happen to have arrived at. Written first, everything before it can be tested against whether it earns that moment.

It also fixes the most common structural failure, which is the talk that ends three times. You feel it as an audience member: the natural conclusion arrives, then another point, then a story, then a thank you. Knowing your ending in advance means you recognise the natural conclusion when you write it.

“A talk without a planned ending does not end. It stops, usually twice, and the room notices both times.”

A structure that works

I am not precious about frameworks and this one has held up across a lot of rooms. Five movements, each doing one job, with the proportions mattering as much as the order.

1. The disruption (2 to 3 minutes). A moment, a story or a fact that makes the room realise the thing they assumed is not quite right.
2. The stakes (3 to 4 minutes). Why this matters to them specifically, not in general. This is where you earn attention for the next half hour.
3. The argument (15 to 20 minutes). Three movements at most, each with one point and one piece of evidence, whether that evidence is data, a story or a demonstration.
4. The turn (3 to 4 minutes). The objection you know they are holding, named out loud and answered honestly. Skipping this is why talks feel naive.
5. The ending (90 seconds). The one you wrote first. Short, concrete, no new material, no summary of what you just said.

The turn is the movement most speakers omit and the one that most changes how a talk is received. Naming the strongest objection yourself, before anyone else can, is the difference between sounding evangelical and sounding credible.

Openings that are not throat-clearing

Most keynotes waste their first two minutes on housekeeping. Thanking the organiser, saying how happy you are to be here, explaining your credentials, telling the room what you are going to tell them. All of it is filler and all of it happens when attention is at its absolute peak.

Start inside a moment instead. A specific scene, a question you cannot answer, a number that does not make sense. Get the room leaning in first, then handle the courtesies twenty seconds later when they have a reason to care who you are.

Your credentials should arrive through the content, not before it. A speaker who demonstrates they know the subject in minute three does not need to have listed their qualifications in minute one, and the room trusts the demonstration considerably more than the list.

Using stories as arguments

Stories in keynotes go wrong in a predictable way: the speaker tells a story they enjoy telling, then reverse-engineers a lesson from it. Audiences can feel the seam. The story was the point and the lesson was the excuse.

Choose the story to serve the argument. Ask what you need the room to believe at this exact moment, then find the shortest true story that produces that belief. If the story does not move the argument forward, it is entertainment, and forty minutes only has room for so much of that.

Keep them short and specific. Named details, real dialogue, one scene rather than a chronology. A ninety second story with three concrete details lands harder than a five minute one covering an entire year.

Use your own failures more than your successes. Success stories position you above the room. Failure stories put you inside it, and a room that feels you have made their mistake listens differently to what you learned from it.

Slides, and whether you need them

I write the talk completely before opening any slide software. Every time I have done it the other way round, I have ended up with a deck narrated rather than a talk delivered, and the difference is obvious from the third row.

Ask whether you need slides at all. Some of the strongest keynotes I have seen used none. If you do use them, they should carry things speech cannot: an image, a number, a comparison, a demonstration. Not your sentences in bullet form.

One idea per slide, and be willing to leave the screen black. A blank screen pulls attention back to you, which is exactly what you want during your most important thirty seconds. Most speakers never use it because it feels like something has gone wrong.

Practical insurance: know your talk well enough to deliver it if the technology fails. It will eventually, and the speakers who handle it calmly are the ones whose talk never depended on the screen.

Rehearsing properly

Reading your talk silently is not rehearsal. It is reassurance. The only rehearsal that tells you anything is out loud, standing, timed, ideally to another person or a camera you are willing to watch back.

Rehearse the opening and the ending more than the middle. Those are the two moments the room remembers, and they are the two you most want to deliver without looking at anything.

On the day

Arrive early enough to stand on the stage before anyone is in the room. Look at where the audience will be, find where the clock is, work out where you will walk. Ten minutes of this removes most of the small surprises that throw people in the first minute.

Talk to a few attendees before you go on. Ask what brought them. You will almost always hear something you can reference, and referencing it early tells the room you are speaking to them rather than delivering a talk you happen to be delivering here.

Then let go of the script. A keynote delivered word-perfect from memory sounds recited. Knowing your structure cold and finding the words in the moment sounds like thinking, and rooms respond to thinking.

Adapting one talk to different rooms

You do not need a new keynote for every booking, and pretending otherwise is how speakers exhaust themselves. One well-built argument can serve many rooms provided the examples change.

Keep the structure and the argument fixed. Swap the stories, the industry references and the stakes section, which is the part that must be specific to the people in front of you. That is usually two hours of work rather than two weeks.

Be honest with organisers about what is customised. A talk adapted with genuine research into their sector is worth more than a generic one, and it is also what justifies a higher fee, which I covered in how much to charge for speaking engagements.

If you are building towards a signature talk that opens larger doors, the same principles apply at a different scale, and I set out that route in becoming a TEDx speaker. Professional bodies such as the National Speakers Association are also useful for finding people who will critique a draft honestly.

Nerves, and what actually helps

I still get nervous, and after a few hundred talks I have stopped treating that as a problem to solve. The nerves are the body preparing, and the speakers I know who claim to feel nothing are usually the ones delivering a talk they have stopped caring about.

What helps is preparation of the specific kind. Knowing your first ninety seconds cold, so that the worst moment is on autopilot. Knowing your structure well enough that losing your place is recoverable. Having delivered the ending out loud enough times that it arrives without effort when you are tired.

What does not help, in my experience, is the standard advice. Picturing the audience in their underwear is nonsense. Memorising the whole script increases anxiety because now there is something to forget. And arriving early to a quiet room to run it once more usually makes it worse, not better.

The one physical thing that reliably works for me is talking to people beforehand. Ten minutes of ordinary conversation with attendees puts my voice in a normal register and turns an anonymous audience into a room containing people I have met.

Handling questions

Question time is where a good keynote can unravel, usually because the speaker treats it as an appendix rather than as part of the talk. It is not the same skill as delivering, and it is worth preparing for separately.

Repeat the question before answering. Half the room did not hear it, the recording certainly did not, and repeating it buys you five seconds to think. It also lets you gently reframe a rambling question into an answerable one.

Answer briefly. The most common error is treating each question as a prompt for another three minute talk. Short answers leave room for more questions, and a session with six questions feels far more engaged than one with two.

Say you do not know when you do not. It is disarming, it is more credible than a hedge, and audiences remember it favourably. The speakers who lose a room in question time are almost always the ones who tried to answer something outside their knowledge.

Prepare for the hostile one. Every subject has an objection somebody will raise sharply, and if you covered it in the turn you can simply refer back. If you did not, acknowledge the point honestly rather than defending, which is what actually deflates it.

Delivering a keynote online

Online keynotes are a different medium wearing the same name, and the mistake is delivering the in-person talk through a webcam. Everything is compressed: attention spans, tolerance for build-up, and the amount of energy that survives the connection.

Cut the length by a third and cut the build-up hardest. The disruption and the stakes need to arrive faster, because there is no room energy carrying the audience through a slow opening and a browser tab is one click away.

Ask the organiser whether cameras will be on. A talk to a wall of black rectangles requires you to generate all the energy yourself, and knowing that in advance lets you prepare for it rather than discovering it in minute two.

Managing length and the clock

Almost every keynote I have watched run long did so for the same reason: the speaker had more material than time and decided to say all of it faster. Rooms detect this immediately. The pace lifts, the pauses disappear, and the ending arrives breathless.

Build for eighty percent of your slot. A forty minute booking should be a thirty-two minute talk. That margin absorbs the late start, the longer introduction, the technical hiccup and the two extra beats you will take when a moment lands well.

Know your section timings, not just your total. If you know the argument should be finishing at minute twenty-four, you can adjust in real time rather than discovering at minute thirty-five that you have three points left. That single piece of preparation has saved me more often than anything else.

Decide in advance what you will cut. There should be one story and one supporting point that can be dropped without breaking the structure. Choosing them under pressure produces bad decisions; choosing them in rehearsal means you cut cleanly and nobody notices.

Turning one keynote into a year of material

A keynote is the most expensive piece of thinking you will do all year, and most speakers use it once. That is an enormous waste, because the structure that works on stage adapts to almost every other format with modest effort.

The argument becomes an article. Each of the three movements becomes a shorter piece. The stories become social posts. The turn, where you name and answer the strongest objection, is frequently the single best piece of writing in the whole thing and works well on its own.

Record it whenever the organiser allows. Even a phone on a tripod at the back gives you something usable, and a clip of you delivering a real point to a real audience is worth more for future bookings than any professionally produced showreel, which I get into in how to get speaking engagements.

Then let the talk evolve. The version I deliver now shares an argument with the first one and almost nothing else, because each room taught me which parts landed. A keynote that has not changed after ten deliveries is a keynote that stopped listening.

Getting feedback that is worth having

Most feedback on talks is useless because most people are polite. “That was great” tells you nothing, and asking “what did you think?” almost guarantees you get exactly that.

Ask narrower questions. What is the one thing you remember. Where did you stop paying attention. What did you disagree with. Each of those produces a specific answer, and the second one in particular is the most useful question I know for improving a talk.

Pick your critics deliberately. One person who knows the subject, one who does not, and one who will actually be blunt with you. Three responses of that kind are worth more than fifty survey scores, which mostly measure whether people enjoyed the morning.

Then watch the recording, all the way through, once. It is unpleasant and it is the fastest improvement available. You will see the verbal tic you did not know about, the section where the room went flat, and the two minutes you could cut without losing anything.

Frequently asked questions

How long should a keynote be?

Most run between 30 and 45 minutes. What matters more than the length is that the structure fits it. A 20 minute argument stretched to 45 is the most common reason a keynote feels slow.

How many ideas should a keynote contain?

One. Supporting points are fine, but a talk arguing several things gives the room nothing to carry out. If you cannot state your argument in a single sentence, the talk is not ready to be written.

Should I write the whole thing out word for word?

Write it out to think it through, then work from structure rather than script. Delivering a memorised script sounds recited, while knowing your structure cold lets you find the words live, which sounds like thinking.

Do I need slides for a keynote?

No. Some of the best keynotes use none. If you do use them, they should carry what speech cannot, such as an image, a number or a comparison, rather than repeating your sentences as bullets.

What is the best way to open?

Inside a specific moment: a scene, an unanswerable question, or a number that does not make sense. Save the thanks and credentials for later, once the room has a reason to care who is speaking.

How much should I rehearse?

Out loud, standing, timed, several times, with disproportionate attention on the opening and the ending. Silent reading builds confidence without revealing the problems that only appear when you speak.

How do I handle the objection I know the audience has?

Name it yourself, before they can, and answer it honestly. That movement is what separates a credible talk from an evangelical one, and it is the section most speakers leave out.

Can I reuse the same keynote for different events?

Yes, and you should. Keep the structure and argument fixed while swapping the stories, examples and the stakes section for each audience. That is a few hours of adaptation rather than a new talk.