Most speakers think the problem with their keynote is the content. It's rarely the content. In this episode John breaks down why writing your keynote with AI is a bad idea, even as the tools keep improving, starting with a story about a speaker he coached at TSL whose entirely AI-written talk was, in John's words, drivel.

John walks through the planning-stage traps that happen before a word gets written: AI's sycophancy (it will tell you your topic is great and bookable when it has no actual market data to base that on), and its habit of fabricating statistics, quotes, and studies with total confidence and zero hedging. He then moves into the writing-stage tells that show up once a talk is drafted, particularly the overuse of rhetorical devices deployed because they pattern-match to "sounds like a keynote" rather than because the moment earned them.

The throughline: AI can't lose you a booking, so it never needs to hedge. Everything it writes for you sounds finished. That's the tell, because nothing true costs that little to say.

This is part one of a two-part series. Part two covers what AI genuinely can't do for your keynote: write your stories, or build a framework that means anything.

FAQ

Is it a bad idea to write a keynote with AI?

Yes, according to John Ball, because AI produces confidently mediocre talks. It cannot access personal stories, emotional experience, or genuine insight into what makes a specific keynote work for a specific audience, and it will validate a topic's commercial viability with false confidence.

What is AI sycophancy in the context of keynote writing?

AI sycophancy refers to the tendency of AI tools to validate whatever topic a speaker brings to them, including telling the speaker their topic is strong and highly bookable, without having any actual data on what talks are currently being booked or what the market wants.

What are common signs that a keynote was written by AI?

Common tells include overuse of rhetorical devices such as the rule of three, tricolon, and "not this, but that" constructions used because they pattern-match to keynote style rather than because the moment calls for them, along with generic phrasing such as variations on "a wolf in sheep's clothing," and bland, generic stories that show no sign of having actually happened.

Can AI write good stories for a keynote?

No. AI cannot generate a speaker's personal stories, emotional experiences, or unique perspective, since it has no lived experience of its own. When prompted for a story, it tends to produce generic content or invent one, according to John Ball.

Does AI fabricate information in keynote research?

Yes. AI tools can invent statistics, quotations, and studies and present them with full confidence and no indication of uncertainty, which has caused real-world problems in other professions, including cases of lawyers citing non-existent legal statutes generated by AI.

CHAPTERS

00:00 AI Keynote Red Flags

01:09 Welcome And Host Intro

01:35 The AI Written Keynote Story

03:56 Speechwriters Versus AI

06:01 Why AI Flatters You

09:51 Ethics And Smart Use

11:44 Hallucinations And False Certainty

14:14 Rhetorical Padding Tells

15:58 Outlining Tool And Real Stories

17:26 Confidence Without Stakes

18:20 Next Episodes And Wrap Up

Visit https://strategic-speaker.scoreapp.com to take the 2-minute Strategic Speaking Business Audit and find out what's blocking you from getting more bookings, re-bookings, referrals and bigger fees. There's a special surprise gift for everyone who completes the quiz.

Want to get coached for free on the show? Fill in the form https://forms.gle/mo4xYkEiCjqtz9yP6, and if we think your challenge could help others, we'll invite you on.

For speaking enquiries or to connect with me, you can email [email protected] or find me on LinkedIn

You can find all our clips, episodes and more on the Present Influence YouTube channel: https://www.youtube.com/@PresentInfluence

Thanks for listening. Rating the show 5* on Spotify helps their algo recommend the show, so please take a moment to follow the show and leave a rating.

Mentioned in this episode:

Do you want to join the Known, Booked & Paid Blueprint September Case Study group?

This September I'll be working with a group of 5 speakers who are already speaking but not getting the level of bookings and speaker fees that they really want. We'll work on the 3 key areas that will allow your speaking business to flourish. Since it's the founding group, there's an extra special deal on the price for joining that will never be repeated. Go to https://presentinfluence.com/kbpb for full details.

If you've ever written a keynote talk or any kind of talk,

2

you've maybe either used AI already

or you've thought about doing it.

3

And there are specific giveaways that can

show when a talk has been written a little

4

bit too much with help from AI, especially

in a keynote, and there are traps that

5

happen before a word's even been written.

6

Many speakers are outsourcing

their own judgment.

7

Not just speakers, I mean many…

8

all of us all over the place are

outsourcing our own critical judgment

9

to AI, thinking it must be correct.

10

and it, and it's producing

confidently mediocre talks that

11

will end up dying on stage.

12

So I wanna tell you a little story about

somebody who I worked with, who wrote

13

their whole talk with AI and how that

went for them, and a little bit about the

14

planning stage and some of the things to

15

look out for, and things like the

sycophancy and fabrications, and

16

really what the writing stage tells,

like the rhetorical padding that

17

you will often see and maybe not

even pay that much attention to.

18

I've been seeing it repeatedly in client

work, and you'll see it and hear it in

19

other people's talks and probably on their

LinkedIn posts and other places as well.

20

So welcome to Professional Speaking,

the show for professional speakers

21

who are serious about speaking and

becoming known, booked, and paid.

22

My name is John Ball, professional

speaker, speaker business coach,

23

stand-up comedian, and sci-fi nerd, and

I'm here as your guide on the journey

24

to a successful speaking business.

25

Now, if you've tuned into the show

before, you maybe already know that I

26

used to be a coach for a very well-known

speaker business coaching company.

27

And in that time, I can remember working

with a young guy who was very eager

28

to get on stage and talk, and one of

the things that he decided to do as

29

a bit of a shortcut because he was so

wanting to have his keynote done and

30

be selling it as quickly as possible.

31

I loved his enthusiasm,

but there was a problem.

32

He just put a prompt into AI and

said, "Write me a talk about this."

33

And the AI did exactly

that, and it was terrible.

34

Now, this was a few years ago, so I'm

willing to grant that this was the earlier

35

days of AI tools when things would be very

obviously written by AI and very terrible.

36

And I admit that the tools have been

getting better, and they probably

37

will continue to get better and

better, maybe at some point to a

38

level where it will be impossible to

distinguish whether something has been

39

individually written or written by AI.

40

But we are not there at the

moment, and so there are things

41

that you need to watch out for.

42

And I'm aware that that means that this

episode could be very dated if you are

43

listening to this a long time after, like

a year or more after it's been recorded.

44

A lot of what I'm saying might not

actually be quite so relevant anymore.

45

But there are some pa- some parts

of this that I feel will always

46

be relevant to this as well.

47

So in all honesty, AI

didn't invent bad keynotes.

48

They certainly existed beforehand as well.

49

But it did help to remove the one

person who used to be able to tell

50

you whether they were any good

or not before you got on stage.

51

Now, this young guy who had written

his keynote all with AI, it was drivel.

52

It really was drivel.

53

This was, as I say, a few years back.

54

AI really wasn't quite so good.

55

Not, it was not even so good in the

collaboration parts as it is now.

56

But I remember that there's

also one of my other…

57

One of my friends was

coaching him a bit as well.

58

We had a little chat about it

because we had both seen the keynote.

59

Now, this was in private.

60

No, no public conversation.

61

But we'd both seen it, and we, like,

we, we were laughing a bit, but also

62

mainly because it never got seen by

or heard by the public because it was

63

terrible, and we both fed back to him

that it was absolutely terrible, and

64

he needed to write something himself.

65

And there are a lot of traps that

we can easily fall into if we are

66

using AI a little bit too much.

67

Now, you might be thinking I want to write

this talk myself, and here's one thing I

68

think with this, like having talks that

are written for you by someone else, it

69

does happen, and there are people who

will pay other people, speechwriters, as a

70

great profession, to write talks for them.

71

And I don't think speechwriters should be

feeling too threatened just at the moment

72

because even though somebody can just go,

"Well, AI write my speech for me," you

73

know it's not gonna be fantastic, right?

74

I think you might want to make some

adaptations and not just deliver

75

whatever AI gives you, but there are

undoubtedly gonna be people who will

76

just take whatever gets churned out

and throw that out into the world.

77

So you're probably going to

encounter that in different places.

78

But what happens in good speech writing

relationships is the writer will get to

79

know as much as they can about the person

that they're writing for, so that they can

80

write in their style, in the way that they

speak, write for them, and maybe even try

81

and call some of their stories in as well.

82

So if they can, they'll have meetings with

the person they're writing the talks for.

83

In the situations where that

doesn't always get to happen,

84

it's not gonna go quite so well.

85

The rapport might not feel quite so

strong, that something's gonna feel

86

a little off in the talk, and it is

maybe gonna feel like the person who

87

delivered the talk didn't actually write

it themselves because of those things.

88

Now, there are certain circumstances

where that might be acceptable, and

89

in the political world you're gonna

encounter that more probably than

90

in the professional speaking world.

91

It is arguable that politics is a form

of professional speaking in some ways,

92

but it's not where you're getting paid

specifically because you are a speaker,

93

it's where speaking is just part of that

role for most people So in the times where

94

you are being paid to be on stage as a

speaker, if you come across as insincere

95

or if you come out, if you come out

with garbage that just doesn't really

96

work or doesn't capture people, doesn't

pull them in, it's not gonna serve you

97

and it's not gonna serve your audience.

98

So I've even experimented, I've

even experimented with having AI

99

help me to write some comedy bits.

100

And I have to say, whilst it comes up

with some interesting ideas, and it

101

sometimes gets very confident about

something, "Oh, that's gonna kill.

102

That's gonna be hilarious.

103

There is gonna be an uproar," it's

very rarely right about that because

104

it doesn't have its own personality.

105

It's not a person.

106

It's a thing.

107

It's a collection of knowledge that's

been scraped off of different parts

108

of the internet and put into one place

to give you, to feed you answers.

109

And we are generally all too

willing to personify it and see it

110

almost as a friend, an invisible

friend that just gives us all the

111

answers that we need for things.

112

And so AI will tell you

that your topic's great.

113

It will tell you that you will

have no trouble booking it.

114

ChatGPT will give you 97% confidence

that you have a great topic and

115

that you're gonna do very well

with it, and it doesn't know.

116

It doesn't have the information about

what kind of talks people are buying.

117

It doesn't have the information about

what's actually going on in the industry.

118

It doesn't have the information about

what actually even makes a great

119

keynote talk because most keynote

talks are not in the internet.

120

They're not available for it to scrape.

121

Even though some of them may

occasionally get transcribed,

122

for the most part they don't.

123

Like, some of them may get video even

used in-house, but they don't really

124

make it onto public channels where

they're likely to be indexed and put

125

into available to, for AI tools to scrape

and to, to reference for creating talks.

126

And that means that what knowledge it

does have, it's pulling from maybe some of

127

the great speeches of all time, political

speeches that have been transcribed

128

maybe comedy sketches and other talks

that it has transcriptions for that

129

are available for it, and TED Talks.

130

TED Talks, TEDx Talks, stuff

that is readily available.

131

But does that really give you what you

need as a professional keynote speaker?

132

Not so much.

133

And you're not gonna be wanting to

deliver, a mountain top speech to

134

inspire the whole world,… You're

probably not gonna be wanting to

135

just do a TED Talk as well, although

it would be a great experience to

136

have if you get to do TED or TEDx.

137

But, but AI is going to confidently

tell you that your talk is great based

138

on what it has, not actual knowledge.

139

AI is a bunch of data that has been pulled

together to give you the best answer that

140

it can give you, but it doesn't know.

141

And even what it does know,

even if it's helpful, even if

142

it's useful Is not personal.

143

AI can never have your level of

stories, your personal experiences,

144

your emotional experiences, because

it doesn't have its own emotions.

145

It can't have your insights into

things that actually happen to you.

146

It can't have your unique perspective

on a particular, on a particular

147

situation or thing, because it just

doesn't really have that capability.

148

It's just trying to pull out what it

thinks is the best answer or the consensus

149

or what it thinks you most want to hear.

150

But it can't give you

your own personal insight.

151

And this is part of the problem with

putting too much of our thinking

152

into AI tools and getting back

whatever we get back and trusting it.

153

And we have a tendency to trust

that AI probably knows more than

154

us, and in some sense it does.

155

But there's a sense in which it really

doesn't as well, because there are just

156

things that it cannot know, it cannot

access, and it cannot help you deliver.

157

And that's partly good

news for keynote speakers.

158

Now, I do know that there are people

who are completely washing their hands

159

of AI, want nothing to do with it.

160

The data center stuff is very bad.

161

There's lots of ethical

reasons not to use it.

162

Privacy and stealing people's work

and all these kinds of things.

163

I'm like 100% agree.

164

And I will say if that's where

you're at and you don't want

165

to use AI at all, good for you.

166

It is gonna slow you down, but what

you may come out with will probably

167

be better and more authentic than

what somebody using AI to help them

168

write their keynote is going to be.

169

So that said, I do use it.

170

I do use it, um, to bounce

some ideas around, but I don't

171

trust everything that it says.

172

If AI says your talk's going to be

booked out in no time and everyone

173

is gonna be in high demand, and

then you go into the marketplace and

174

you find that's not the case, then

it's gonna be very disappointing.

175

But that's a hard way to find out.

176

And that means y- y- you've trusted

your whole research, your whole

177

investment into AI that isn't gonna

be able to give you perhaps some of

178

the insight that, that a coach might.

179

That if you work with a real person who

has insight and knowledge of the speaking

180

industry and what people book and what

works and what doesn't, and what, what

181

a keynote needs to look like and be like

to be effective and to get booked, and

182

it just doesn't have all this information

and, and it won't because these things

183

are not being codified into the industry.

184

And there's so many different opinions

and ideas about this that you're gonna get

185

different senses of where things are at.

186

You're only ever gonna get a sense of it.

187

You're never really gonna have

full insight that you might get

188

from somebody who's actually in the

industry and able to give you valid

189

feedback as to what-- whether what

you're doing is gonna work or not.

190

AI will fabricate information and

will give you false certainty,

191

invented statistics, even invented

quotations, um, studies, and have

192

no equivocation about this as well.

193

It would just say, "Oh, this, this is it.

194

This is the study referred to."

195

And even if you use Gemini, which

is generally better for that kind

196

of research, you still have…

197

I think you still have to

double-check, and I hope that you do.

198

Um, it can be a great research

aid to help you back stuff up in

199

your talk that needs backing up.

200

You've probably heard of these sorts

of things, but there's been cases where

201

lawyers have put cases into courts and

listed statutes that have been relevant

202

to them, and then it's turned out that

they've used AI and those statutes didn't

203

even exist, but they never checked, and of

course, that causes all sorts of problems.

204

And it's not just in those environments,

it's pretty much everywhere.

205

The stuff that hasn't actually been

checked and isn't even real has

206

been delivered as though it was.

207

There's been leaders in the business

world, in politics, in pretty much

208

every area who have just delivered

their talks straight off of ChatGPT

209

because they haven't hired a speech

writer or they just haven't had time

210

and just put some, put some prompts

in and read out whatever they got.

211

A real coach will give you some pushback

before you even really get to that

212

point of having created your whole

keynote based on a fallacy or based on

213

something that might not actually work.

214

And hopefully you'll get the enough

pushback to know whether you're…

215

whether what you're doing is commercially

viable, whether it's gonna resonate

216

with the rooms that you go into, and

whether your take is different enough

217

and viable and needed within the

organizations that you want to go and

218

speak to AI just can't tell you that.

219

It can give you its best guess, um,

but it won't be able to give you the

220

actual real pushback as well, and it

just won't be able to challenge you

221

in the ways of thinking that a real

conversation with a real person will.

222

And the actual personal friction that

you can have with somebody when you

223

are being challenged, that pushes

us to raise our level of thinking.

224

AI doesn't really push us to

raise our thinking at all.

225

It actually encourages us to switch

that off and to switch off our

226

skepticism, and that is detrimental.

227

But it keeps you coming back to the

AI tool, which of course is what

228

the AI companies want you to do, and

maybe that's exactly why it does it.

229

Uh, in the writing stage, you've probably

already seen this with social media posts

230

and the likes, and I've noticed this.

231

One of the reasons why I've been

particularly cutting back on having AI

232

write some of my social media posts or

even co-write, because I find that it

233

just uses a lot of rhetorical devices.

234

And whilst I love rhetorical devices

in talks and I think rhetoric is a

235

very interesting area that most of

us don't fully understand, but things

236

like the rule of three, tricolon or

the not this but that, um, the…

237

these things get deployed because they

pattern match to sou-sounds like a keynote

238

rather than because the moment actually

earned them or because it was the right

239

rhetorical device to use in the moment.

240

So you'll see that things get

padded out with rhetorical devices.

241

The one that I've been seeing a lot

recently on posts, and I know it's

242

come up in some of mine as well,

because I keep thinking, "Oh, it

243

makes me wince every time I see it."

244

Uh, and this will be outdated probably

if you're listening to this six

245

months, 12 months on from recording.

246

But the one that bothers me the most

right now is AI keeps throwing back this

247

thing of, "Is this dressed as that?"

248

Or, "Is this wearing a

coat that's like that?"

249

And it is of this, that, well, it's,

there's something in disguise, but

250

it seems to always phrase it in this

rhetorical way of like a wolf in sheep's

251

clothing, and I see that come up in so

many different posts, and I've seen it

252

come up in people's talk content as well.

253

And for me, it's a, it's often a signifier

there's like probably used AI for this.

254

I can't be 100% sure, but I'm

pretty sure And I don't really

255

want to see that because it does

mean that you, you've actually been

256

getting AI to write your talks.

257

So I created a tool to help my clients

write their talks, but the tool

258

is only there to help them outline

their talks with their own content.

259

It's not allowed, well specifically

under instruction, not to create

260

content for them, but only to pull

the content out of them to help

261

them outline and structure a talk.

262

So even if they put the instructions

in to create this part or write

263

me this bit, it's not gonna do it.

264

It's gonna say, "No, this is

just to outline your talk."

265

And even though I do encourage

my clients not to use AI so much

266

in their talking matching to, to

get very hands-on, writing is…

267

I, I think writing stuff out is

great, but I also find that I used to

268

think, "Oh, you know, better to write

stuff out, it feels more personal."

269

But I actually get more creative

personally when I'm typing stuff in now.

270

So I'll open up a document and

I'll just start typing away,

271

and the ideas start to flow.

272

And if I've got a structure

to put them into, then that's

273

much, much easier to do that.

274

And I, I don't see this so much, but

it does come up, like we get these

275

sort of bland, generic stories that

never, probably never really happened.

276

And AI will create stories if it

feels that something needs that.

277

But I think telling generic stories in

general, like you put your, your real

278

story content into it, it doesn't really

know how to do great storytelling, so

279

it's gonna bring it back to you in a

way that isn't particularly helpful.

280

And I'm gonna cover some more

of this in my next episode

281

on this in a few weeks' time.

282

A common thread against every trap in

working with AI to create your keynotes

283

is having confidence without stakes.

284

AI can't lose a booking for you, so

it never really needs to hedge on

285

something like something's on the line.

286

Uh, it can just tell you very confidently

tell you that this is right, or what

287

you've done is great, or you have…

288

you've created the most phenomenal

keynote that's ever been written.

289

And you're thinking, "Wow, my,

this must be really amazing.

290

I'm gonna blow people's minds with it."

291

And it has no idea what market

response to that is going to be like.

292

Everything, everything AI writes

for your keynote sounds finished.

293

But that's the tell.

294

Nothing that's actually true

costs that little to say.

295

There has to be stakes.

296

When you put something out into the

marketplace, stakes are on the line, and

297

false confidence that's been given to

you by an AI tool is not gonna save you.

298

So in the next episode, I wanna

talk a little bit more about what AI

299

genuinely can't do, which is stories.

300

Also your frameworks and we're gonna get

a bit more into that, and I've got a great

301

episode coming up towards the end of the

month on frameworks with Susanna Reay.

302

Susanna Reay is a good friend of

mine, and she has an amazing book

303

on framework thinking that I can't

wait to bring you that episode.

304

Her book is phenomenal, and we'll

talk all about that very soon.

305

But next week I'm gonna be bringing you

an interview, a returning guest, uh,

306

Jack of Mike Verret, a marketing expert,

who's gonna be talking about how you can

307

really develop the relationships that

you need to have in the industry and

308

talk effectively about what you do to

be able to help you position yourself.

309

So if you like all of those

positioning conversations, we have

310

a bit of that discussion with Mike.

311

Tune in for that next week.

312

Not one to miss.

313

We'll be back talking more about AI

in a couple of weeks, then followed

314

by Susanna Reay's ep- amazing episode,

which I a hundred percent urge

315

you not to miss, a- and more great

stuff coming after that as well.

316

With all of that, if you're writing a

talk right now and you need a bit of

317

help, get the help you need, but don't,

oh, don't trust that AI's gonna be able

318

to deliver for, deliver it for you.

319

If you wanna have a conversation

about that, reach out to me.

320

All my contact information is in the

show notes, and I'm happy to, happy

321

to have a conversation with you to

see if there's any way I might be able

322

to help or direct you in doing that.

323

But for now, until the next episode, go

and do something worth talking about.

324

I'll see you next time