Copywriting secrets from Harry Dry

Harry Dry | The Marketing Meetup

Abstract

A copywriter gives a conference talk of just under twenty minutes, opening on a slideshow of advertisements photographed across one day of travelling London's underground — fifty-seven seen, one of them liked — and takes that day as the state of the craft. The argument is that abstract copy is not a writing failure but a symptom of having nothing to say, so the repair sits upstream of the words. The first half is diagnosis; the second is offered as a little solution in two parts, having something to say and then saying it well.

• Writing that feels hard is evidence it is being done properly, because difficulty is the cost of thinking.
• A line can be tested for borrowed language in three ways that need no taste: an exact-phrase search, a parody read, and a competitor's signature.
• The problem in one word is abstraction, and a language model detects it far better than it repairs it.
• A category-generic line leaves the best copywriter in the world nothing to convert, because nothing has been decided yet.
• Positioning fits in two columns: what the thing is up against, and why anyone would pick this one.
• Most copy is matter of opinion where it needs to be matter of fact, and the conversion is usually a substitution.

Everything below is the talk in the order it was given, from the fifty-seven advertisements through the three tests to the swim-school exercise and the cold tap. For anyone building a personal brand the relocation is the useful part: a bio that reads like everybody else's is not a wording failure, and every test offered here is cheap enough to run alone, on work already published.

Chapter summaries

01:03 - Every selling surface is one discipline, so the craft beneath them does not vary

Homepages, storytelling, positioning, offers, ads, hooks, social media and sales are not adjacent disciplines — they are all the same one. The list is read out as eight separate surfaces and then collapsed into one word, and no case is made for the merger beyond the collapse itself.

"Homepages, storytelling, positioning, offers, ads, hooks, social media, sales."

— 01:03

What the move asserts is that the craft judged on every one of those surfaces is the same craft: if the test is always whether a person is moved to act, the format underneath the test does not change what the writing has to do. The second opening move is about how the work feels. A children's book character is quoted banging his head against a wall in the hope that it will produce a story.

"Then Toad began to bang his head against the wall. Why are you banging your head against the wall? asked Frog. I hope that if I bang my head against the wall hard enough, it will help me to think of a story, said Toad."

— 01:46

The response offered is that the banging is the job — difficulty is the cost of thinking, and the easy version is the one assembled out of phrases that already exist. It is offered as encouragement rather than as a theory of composition.

02:22 - Advertising's default output says nothing concrete enough for a reader to check

The default output of the advertising industry is copy that says nothing specific enough to be checked. The evidence is one day of house-hunting across London's underground, photographed along the way.

"Some pictures to show you the problem with advertising or copywriting or marketing."

— 02:26

The haul runs: live life to the max; be a force for good; where your money works; unleash the power of pay; essential skills to excel; a 360 view of your business; beat your competitors in less time; and a chocolate bar whose whole promise is the phrase and more.

"Destination nuclear: be a force for good."

— 02:44

The mechanism is not laziness but safety. A claim with no concrete referent cannot be disagreed with, which is exactly why it gets chosen — and one of the lines is challenged on precisely that point, what is meant by essential skills and what is meant by excel. Where every line in a category makes the same uncheckable promise, the expected registers as nothing. It is presented as one person's single day of sampling, not as a survey.

04:55 - The one memorable ad refused a benefit its whole category already shares

The one ad worth liking in a day's sampling worked by refusing a claim its category treats as a benefit.

"There's one ad I like that day. I saw 57."

— 04:52

Fifty-seven advertisements seen, one of them liked. The exception works subtractively: it names a claim the category treats as a selling point and disqualifies it.

"Designed to clarify that not tested on animals is already industry standard, not a product benefit."

— 04:59

Naming a benefit as an industry standard removes it from the persuasion and leaves only what is actually differentiating — which is why the line is memorable while the other fifty-six are not. What the product itself does never gets said. The position the diagnosis is made from is marked out loud.

"I feel like it's very easy to stand up on a stage and pontificate and criticise. It's much harder to do."

— 05:06

05:16 - Bad writing is words nobody chose, gummed together in an order somebody else set

Bad writing is not badly chosen words — it is words not chosen at all, gummed together in an order somebody else already set. The diagnosis is handed to a passage written in 1946, read out in full.

"As I have tried to show, modern writing at its worst does not consist in picking out words for the sake of their meaning and inventing images in order to make the meaning clearer."

— 05:21

The claim in the passage is not that the words are poorly chosen. It is that they were not chosen at all.

"It consists in gumming together long strips of words which have already been set in order by somebody else."

— 05:36

Pre-set strips of language require no thought, and the passage is explicit that this is the attraction rather than an accident.

"The attraction of writing this way is that it's easy."

— 05:39

The scope is bounded in the passage's own words — writing at its worst, not writing in general. That boundary matters for what follows, because the three tests offered next are tests for this specific failure and not general measures of quality. It also explains why the fix is upstream of the sentence: a phrase somebody else ordered can be replaced only once there is something of your own to put there. The habit is recognisable enough that a bio assembled from borrowed phrases reads as competent and lands as nothing.

05:58 - An exact-phrase search counts the companies that wrote a line first

Borrowed language is mechanically detectable: put the line in quotes in a search engine and count the companies that wrote it first.

"The first one is, how do you stop gumming together words set in order by somebody else? Well, it's a very obvious answer. You can search them on Google with curly quotes."

— 05:58

The test converts a private hunch about staleness into a visible number.

"If you put something in curly quotes in Google, it's a direct match. So you can see how many people have wrote the same thing before."

— 06:07

Run on a payments provider's own site line, the count comes back in the hundreds.

"There's genuinely hundreds of companies who are describing their payment provider as take your payments to the next level. The words have been killed. They've lost all their meaning."

— 06:21

Nothing about the test requires taste, experience or a second opinion: the search box supplies the verdict, and the only input is a line already written. It is the cheapest of the three tests offered and the only one that returns a number rather than a judgement.

06:33 - A parody voice leaves only content, so a line with nothing in it collapses

Copy that can be parodied is copy with nothing in it; sincerity and fact are parody-proof.

"Read aloud what you write in a phony American accent."

— 06:33

The credit is given immediately, and the source is only half-remembered.

"That's not original. Unfortunately, it's in the copy book and I forgot who first said it."

— 06:36

A parody voice strips tone out of a line and leaves only its content, so a line whose content is empty collapses into a joke while a factual one is untouched by the reading. Demonstrated live on two of the day's advertisements, the vague one becomes an announcer's punchline and the honest one simply survives. The verdict on why is given in three words.

"It's honest. You can't parody. You can't cliche honesty."

— 07:13

07:17 - An ad a competitor could sign transfers preference to nobody in the category

If a competitor could put their name to the ad, it is not your ad. The third test is attributed rather than invented.

"Never write an ad a competitor can sign."

— 07:21

A claim available to everyone in a category transfers preference to nobody in it, so the test is performed by rewriting the ad with a rival's name underneath and seeing whether anything breaks. One university's promise to equip you with the essential skills to excel survives the swap unchanged, which is the verdict. Against it, a claim no rival can sign.

"Nobody's perfect. Number one in England for overall student satisfaction."

— 07:40

The counter-example is a run of years, each of them checkable.

"Plymouth can't do that ad. Neither can Durham."

— 07:51

That is the whole test: a line a competitor could put their name to is not doing the one job a positioning line exists to do. It is the same instinct behind a differentiation audit run against the field rather than against your own taste.

07:54 - Abstraction is a hiding place, because thinking in concrete terms is hard

The problem, in one word, is abstraction — retreating from concrete reality because thinking is hard.

"I think in one word the problem I'm getting at here is abstraction."

— 07:54

The mechanism named is avoidance rather than incompetence.

"So we hide away. We pull back away from concrete reality."

— 08:04

Abstract words are a hiding place: they let a company appear to have said something while committing to nothing that could later be checked or held against it. The room is handed a printed card of such words as a reading pass — one spotted on your own homepage is a shot of pickle juice, five in a row is the whole bottle.

"This is what most websites look like. It's prefabricated words tacked together like sections of a prefabricated hen house."

— 08:24

The image is borrowed from the same 1946 passage, and it is doing the same work: prefabrication is a description of the process, not of the taste of the person who ran it.

08:30 - A language model detects abstraction well and repairs it badly

A language model is a poor fixer of abstraction and a decent detector of it.

"I don't think AI is particularly good at solving this problem, but it's not bad at spotting it."

— 08:30

The split follows from what the machine actually does. Detection is pattern-matching against a corpus, which a language model does well; replacement requires having something to say, which it does not have. The prompt offered is a detector and nothing more.

"Read the following copy. Can you highlight every word or sentence which contains abstract verbs, abstract nouns, abstract adjectives, abstract adverbs, and cliches?"

— 08:43

The expected output is a long list of things to change, and even that is hedged.

"If you're not at the top of your game, you will get back a long list of things to change."

— 08:56

Both halves of the claim are qualified in the same breath — not particularly good at solving it, not bad at spotting it — which is a narrower and more usable claim than either the enthusiasm or the dismissal that usually surrounds the question. The limit is structural rather than temporary: the part a model cannot do is the part that requires a commitment.

09:10 - Imitation feels safe because it removes both the risk and the labour

Companies abstract for two reasons — most have little to say, and they are not good at saying it — and imitation is the safe response to both.

"I think firstly, most companies have little to say. And secondly, they're not too good at saying it."

— 09:17

Both reasons point at the same response.

"So what do they do? They imitate. It's easy to imitate. We feel very safe when we imitate."

— 09:22

Copying an existing phrasing removes the risk of being wrong and the labour of being specific at the same time, which is why it is the default rather than the exception. It is offered as a reading of why the day's fifty-seven advertisements looked the way they did rather than as a finding, and the hedge is in the sentence itself.

09:38 - A category-generic line leaves the best copywriter in the world nothing to convert

A category-generic positioning line gives the best copywriter in the world nothing to work with. The worked example is a category-generic homepage: switch to Robly, email marketing for everybody. Asked from the floor what would improve it, the room supplies the missing question — who is it for — and the answer confirms it.

Words sell by converting a reason to choose. Where no reason has been established there is nothing for the words to convert, and skill at the sentence level cannot manufacture one.

"They got nothing to say in the first place. It's email marketing for everybody."

— 10:10

The count makes the problem concrete.

"There's 365 email marketing companies. Mailchimp is the biggest. They're also email marketing for everybody."

— 10:15

The largest company in the category makes the same claim, which leaves the smaller one with a line that argues for the category and not for itself — the position a whole sub-category occupies at once.

10:24 - Sellable words are downstream of nine questions about the business itself

Copywriting is not writing words, and it is not even writing words that sell; the sellable words are downstream of nine questions about the business.

"When we think about copywriting, it's a big mistake to think it's just writing words, or write words which sell."

— 10:26

The correction is aimed at a specific request — the meeting that opens by asking for a line — and it relocates the work underneath the sentence. Nine questions are listed as what sits below the iceberg.

"What are you competing against? Why you? Who's it for? When's it for? What is it? What are you not? Why did you start the product in the first place? What's your point of view? Why do you exist?"

— 10:42

Each question forces a commitment, and a commitment is the raw material a line can be cut from; the set is summarised as story and positioning. The ease claimed at the end is worth reading precisely.

"When you have these elements in place, summarised by perhaps story and positioning, it's really easy to write words which sell."

— 10:53

The ease is claimed for the writing, not for the answering. Working out what you are competing against and why anyone should pick you is the expensive part, and it is the part that is usually skipped.

11:05 - Positioning is relational, so the claim only means something against its alternative

Two of the nine questions carry most of the weight, and they lay out as two columns: the alternative on the left and the reason to choose on the right.

"I like to lay this out very simply with left and right."

— 11:05

Two of the nine questions carry the weight, and the layout is the method.

"What am I competing against on the left side, and why you on the right hand side."

— 11:13

The right-hand claim only means something against the left-hand alternative, which is why the columns are filled in that order. A bakery is worked through in one line.

"Well, it's baked in bakeries. Bread in bakeries takes better. So you have, real bread isn't made in factories, it's baked in bakeries."

— 11:28

The alternative is factory-made bread; the reason is that bread from a bakery tastes better; the line that comes out — real bread isn't made in factories, it's baked in bakeries — is the two columns pushed together with nothing added.

11:36 - Where a whole local category says one sentence, the sentence is the opportunity

Where a whole local category says one sentence, the sentence itself is the opportunity. The exercise handed to the room is a local category with eight competitors in it.

"In Southampton there is eight swimming schools for children, all very closely proximate."

— 11:50

The figure is corrected mid-sentence.

All eight advertise the same thing, and that sameness is the whole opportunity.

"Now they will say the same thing. Swimming lessons Southampton. Welcome to Southampton based quality lessons, affordable prices."

— 11:58

Undifferentiated supply means any specific claim becomes the only specific claim available in the category, so the cost of being first is low and the return is the entire field. The clue given before the room is set to work is to start from the left column, which is where the failures already are — the same move as reading a market by what everyone in it does merely adequately.

13:00 - Inverting each failure the category shares produces claims that are already specific

The position is built by inverting each thing the category does badly, item by item.

"How I would approach this is we have busy classes. So what about if our swim school is just three kids to one class?"

— 13:02

The left column is already a list of complaints — busy classes, supervised play, no clear goal — and each one negated produces a claim that is specific because the failure was specific. Supervised play inverts to actually teaching children to swim.

"There's no clear goal. What about if the swim school is positioned as ditching armbands within six weeks?"

— 13:14

The three inversions compress into one line.

"You end up with Nitro Swim School. Others do lessons. We teach swimming."

— 13:21

Nothing in the result was invented. Every claim in it is the negation of something the category already does badly, which is why the method survives being handed to a room with one minute to run it. The gap is not hidden; it is the list of things the category is already failing at.

13:25 - Cleverness is compensatory, filling the space where the substance is missing

Once there is something to say, the writer's job is to get out of the way; cleverness is what fills the space where substance is missing.

"When you have something to say, you can get out of the way. You don't have to be clever."

— 13:25

Technique is compensatory: it gets reached for when the underlying claim cannot carry the line by itself, so cleverness in a piece of copy is usually evidence of an unsettled position rather than of craft. The instruction to the room is direct.

"You should all be a little bit less technical."

— 13:32

It is the shortest passage in the talk and the one that makes the structure of the argument explicit — the first half was diagnosis and the tests, the second half is having something to say, and the writing itself is what happens last and least.

13:53 - The oldest usable test asks whether a line would help a salesperson sell

The oldest usable test of any advertising decision is whether it would help a salesperson sell in person. The source is named, dated and paginated.

"A book by Claude Hopkins in 1926, page 22 actually. The chapter is called just salesmanship."

— 13:53

The question it opens with is offered as a general-purpose answer rather than a rule about wording.

"There is one simple and right way to answer many advertising questions."

— 14:04

The test is one question, asked twice.

"Ask yourself, would this help a salesman sell the goods? Would it help me sell if I met the buyer in person?"

— 14:09

A salesperson gets one reaction and cannot hide behind abstraction, so their language is already optimised for a real buyer rather than for a document. It is presented as one of three ideas that helped most, with the credit given plainly — none of the three are his own.

14:22 - The front door removes every line that only works inside a document

The front door is the working form of the salesmanship question: the product in your hand, about to knock, and whatever you say when it opens is the copy.

"Whenever you're writing something, you want to imagine yourself with the product in your hand walking to a front door."

— 14:24

Imagining a real person opening a real door removes every line that only works inside a specification. The worked example is a satellite internet kit, first in the voice of the document.

"You're not going to say, hey, I got this new Wi-Fi, it's effortless setup and it's got instant connectivity for your household."

— 14:39

Then in the voice of somebody holding the box.

"Pointed at the sky, plug it in. Directions work in either order."

— 14:54

The second version is not shorter because short is better. It is shorter because each clause is an instruction a person could follow while standing on their own doorstep, and the reassurance is carried by the last clause rather than asserted by an adjective.

15:00 - Demonstration outranks description, because an event cannot be met with a counter-opinion

Where a claim can be demonstrated, demonstration outranks description.

"If you're not a salesperson, you might say, oh, it's 40 millimetres thin."

— 15:06

A specification is an assertion the audience has to accept. A demonstration is an event it witnesses, and the difference is that the second cannot be met with a counter-opinion.

"You have Steve Jobs, who puts the MacBook Air into a yellow manila envelope to show how thin it is."

— 15:11

The envelope does the arithmetic that the number was asking the audience to do.

The rule this sets up runs through the rest of the talk: where a claim can be shown rather than stated, showing it removes the step at which the audience has to decide whether to believe you. That is why an act with a cost attached reads as evidence.

15:22 - Most copy is matter of opinion where it needs to be matter of fact

Most copy is matter of opinion when it needs to be matter of fact. Seven car-advertising lines are put up and sorted live.

"Two of them are matter of fact, five a matter of opinion."

— 15:59

The two facts are a count of men at a named factory with one job, and a comparison of leg room against a limousine. The five opinions are built on love of the drive, uncompromising power and a symphony of high performance. The mechanism is that an opinion invites the reader's counter-opinion, while a fact gives them nothing to argue with and something to picture. The same contrast is run on two newspaper front pages after a goalless international draw: one paper runs the manager's sacking and the team not clicking, the other runs a count.

"682 passes against Slovenia. One pass between Kane and Bellingham."

— 16:33

The conversion is usually a substitution rather than a rewrite, because each vague word conceals a specific one the company already knows.

"Instead of a leading chef, it could be a New York City chef. Instead of packed with chicken, you could say the number one ingredient is fresh chicken."

— 16:58

A dog-food pack reading tasty ingredients, leading chef, packed with chicken becomes checkable one word at a time, and supplying the specific word costs nothing. The rule as it is quoted from the copywriting anthology is garbled in the recording, while every worked example runs one way only: opinion into fact.

17:17 - The strongest conversion is an act the seller would perform, not a better word

The strongest conversion of opinion into fact is not a better word but an act the seller would perform. Asked from the floor for a fact to replace tasty ingredients, the room offers fresh. The stronger answer is not a word at all.

An act is unfalsifiable evidence of the claim and cannot be restated as an opinion, which is what separates it from even the best substitution.

"All ingredients you would eat. That's fact. I will eat the chicken."

— 17:23

The seller eating the product is the same move as the laptop in the envelope: the claim stops being something the audience is asked to accept and becomes something it has watched happen. It is also the cheapest version of the move, because it requires no production and no budget — only that the claim be true.

17:28 - Perception is comparative, so writing is a search for juxtaposition

All writing is a search for juxtaposition.

"All writing is just a search for juxtaposition."

— 17:36

The line is cited from an advertising book and the reaction to it is anticipated.

"Now this sounds weird and technical, but honestly I've made a career from this line."

— 17:37

Perception is comparative: a value is only legible against the value beside it, which makes the pairing rather than the phrase the unit of the work. Two levels are distinguished — juxtaposing words is simple, juxtaposing ideas is described as more technical. At word level, the move is a large number set against a small one.

"Marketers, you will spend 22,000 hours of your career writing. Spend two learning how to do it well."

— 17:51

The gap between the two figures is the argument; neither figure makes it alone. It is the shell behind a great many reusable opening lines.

17:58 - A single figure has no cold tap beside it and cannot be felt

A single impressive figure has no cold tap beside it and therefore cannot be felt. The brief handed over was a single satisfaction figure for a corporate card.

The reason the missing pairing matters is put physically.

"Why do I care about juxtaposition? It's because we only know the hot water because of the cold water. We only know it's sunny today because it's been freezing for two weeks."

— 18:21

A figure with no counter-figure beside it has nothing to be measured against, so the reader has no way to feel whether it is large. The rewrite opens on the failures — eight businesses tried the card and decided it was not for them — and only then supplies the number it was hiding.

"That's how many businesses tried our beloved corporate card in 2024 and decided it was the perfect fit."

— 18:52

Leading on the cold number makes the hot one measurable rather than asserted, and builds a small story out of two counts. The same figure is made legible by the number placed next to it, not by being made bigger.

19:05 - An idea without tension is not yet a line, because it names nothing displaced

An idea with no tension in it is not yet a line; the fix is to give it a point A and a point B. The last example is turned on the speaker's own work, an essay finished and due to publish the next day.

"Going to publish it tomorrow. And my idea was that big ideas come from conviction."

— 19:08

Stated that way the thesis supplies only a conclusion.

"There's no hot tap. There's no cold tap."

— 19:14

Naming what the idea displaces is what supplies the cold tap, so the rewrite puts the alternative in front of the claim.

"So I wrote, big ideas have much less to do with creativity than you think and much more to do with conviction."

— 19:16

Point A and point B, in one sentence, on the writer's own argument. The talk then closes by declining to close: there is no big ending, stated plainly and followed by thanks. A wrap-up is a stop signal, and naming the absence of one keeps the ending in momentum.

Personal Branding Lessons

The talk is aimed at companies with homepages, but every test in it runs on one person's bio, landing page or profile without modification — and each one is cheap enough to run alone, in an afternoon, on work that is already live.

Search the drafted line in curly quotes before it ships

Put the sentence inside quote marks in a search engine, which forces a direct match, and count how many people wrote it first. A payments company's promise to take your payments to the next level comes back on hundreds of sites; a personal tagline about helping brands tell their story will come back on thousands. The count is the whole verdict — there is nothing to interpret and no taste involved. Anything above a handful means the words have already been used up, and whatever they once meant has been spent by the people who got there first. 05:58

Read the draft aloud in a parody voice

Read the line out in an exaggerated announcer's accent and listen to what happens to it. A line built on tone collapses into a joke, because the parody strips the tone out and leaves only content — and there was no content underneath. A line built on a fact survives the reading unchanged, because a fact does not have a register to remove. The test takes ten seconds and is the only one in the talk that needs no second device. 06:33

Put a competitor's name under the finished ad

Rewrite the line with somebody else's name at the bottom and see whether anything breaks. A university promising to equip you with the essential skills to excel survives the swap intact, which is the failure: a claim the whole category can make transfers preference to nobody inside it. The same test applied to a five-year run of a checkable ranking breaks immediately, because no rival can sign it. For a personal brand the swap is easier still — put a peer's name on your own about page and see whether a single sentence objects. 07:17

Answer the below-the-iceberg questions before asking anyone for a line

The request that opens a meeting by asking for a line is the request that cannot be filled. Words sell by converting a reason to choose, so where no reason has been established there is nothing for the words to convert, and the best writer in the world is handed the same nothing. The nine questions underneath are the work: what you are competing against, why you, who it is for, when it is for, what it is, what you are not, why you started, what your point of view is, and why you exist. 09:38

Fill the up-against column before the why-you column

Draw two columns and write the alternative on the left before writing the reason to choose on the right. The order is the method rather than a preference: positioning is relational, and the right-hand claim only means anything measured against the left-hand one. A bakery's whole line falls out of the pairing — factory-made bread on the left, baked in bakeries on the right — with nothing added between them. Starting on the right produces an adjective; starting on the left produces an argument. 11:05

Invert each thing the category does badly

List what the category does badly, item by item, then negate each entry and read what is left. Eight swim schools inside one small radius all advertise lessons, quality and affordable prices; the left column reads busy classes, supervised play, no clear goal. Inverted, those become three children to a class, actually teaching them to swim, and armbands gone within six weeks. Nothing in the result was invented — every claim is the negation of a specific failure, which is why the method survives being handed to a room with one minute to run it. 13:00

Walk the line to a front door before publishing it

Imagine the product in your hand, walking up to a door, about to knock, and ask what you would actually say when it opens. It is the working form of a question from 1926 — would this help a salesman sell the goods, and would it help me sell if I met the buyer in person. A satellite internet kit described as effortless setup with instant connectivity becomes, in the salesperson's version, put it on the ground, point it at the sky, plug it in, and the directions work in either order. The door removes every line that only works inside a document. 13:53

Convert each matter-of-opinion word into a matter of fact

Sort the piece into two piles, opinion and fact, then work the opinion pile one word at a time. An opinion invites the reader's counter-opinion; a fact gives them nothing to argue with and something to picture. A dog-food pack reading leading chef and packed with chicken becomes a New York City chef and the number one ingredient is fresh chicken — the same claim, made checkable, at no cost. Most of the time the conversion is a substitution rather than a rewrite, because the specific word was already known and simply not used. 15:22

Put the cold number in front of the hot one

Find the counter-figure to any impressive number and lead with the smaller one. A satisfaction rate standing alone has nothing beside it to be measured against, so a reader has no way to feel whether it is large; we only know the water is hot because of the cold water. Rewritten, eight businesses tried the card and decided it was not for them, against twelve thousand and fifty-nine who decided it was the perfect fit. Same underlying figure, two counts, and a small story built out of the gap between them. 17:58

Questions

Each answer ends at the moment in the recording where it is given.

How do I tell if my copy is made of phrases everybody else already uses?

Put the line inside quote marks in a search engine, which forces a direct match, and count how many companies wrote it before you. A payments provider's promise to take your payments to the next level came back on what is described as genuinely hundreds of sites. The verdict offered is blunt: at that point the words have been killed and have lost all their meaning. The test needs nothing but a search box and the line you already wrote. 05:58

Why does my homepage sound exactly like my competitors' homepages?

Because the claims on it are ones any competitor could sign. One university's promise of essential skills to excel transfers word for word to a rival with only the name changed, which means it moves nobody. The counter-example is a claim built from something unrepeatable — five consecutive years ranked number one in England for overall student satisfaction. A claim available to the whole category transfers preference to nobody inside it. 07:17

Can AI fix my marketing copy?

The claim made is hedged in both directions: not particularly good at solving the problem, not bad at spotting it. The prompt given pastes the copy in and asks for every abstract verb, noun, adjective and adverb, plus every cliché, to be highlighted — and the long list of things to change comes back if you are not at the top of your game. Detection is pattern-matching. Replacement needs something to say, which the machine has not got. 08:30

Why can't a good copywriter improve my website?

Because the fault is upstream of the words. A homepage reading email marketing for everybody could be handed to the best copywriter in the world and still not move, since there is nothing there to convert into a reason to choose. Words sell by converting a reason; where none has been established, there is nothing for them to work on. The fix is the nine questions underneath the line, not a better writer. 09:38

What do I need to work out before I write a single line?

Nine things: what you are competing against, why you, who it is for, when it is for, what it is, what you are not, why you started, what your point of view is, and why you exist. Summarised as story and positioning. With those in place, writing words which sell is described as really easy — ease claimed for the writing, not for the answering. Start with who it is for. 10:24

How do I position a business when every competitor says the same thing?

Treat the sameness as the opening. Eight children's swim schools inside a 500-metre radius all advertised lessons, quality and affordable prices, which left every specific claim unclaimed. Start from the left column — what the category does badly — then invert each item: busy classes become three children per class, supervised play becomes actual teaching, no clear goal becomes armbands ditched within six weeks. Rate the position before committing to it. 11:05

What is the simplest test of whether a marketing line is any good?

Ask whether it would help a salesperson sell the goods, and whether it would help you sell if you met the buyer in person — a question from page 22 of a 1926 advertising book. The working version is a front door: the product in your hand, about to knock, and whatever you would say when it opens is the copy. It killed effortless setup and instant connectivity for a home Wi-Fi kit. 13:53

How do I turn vague claims into something people actually believe?

By sorting every line into matter of opinion and matter of fact, then converting the opinions one word at a time. Leading chef becomes a New York City chef. Packed with chicken becomes the number one ingredient is fresh chicken. The strongest version is not a better word but an act — picking the dog food up and eating it. An opinion invites a counter-opinion; a checkable fact does not. 15:22

Why do the numbers in my marketing copy fall flat?

Because a single figure has nothing beside it. A brief of 99% of companies satisfied in 2024 was identified as having no juxtaposition in it at all — we only know the water is hot because of the cold water. The rewrite led with the cold number: eight businesses tried the card and decided it was not for them, against 12,059 who decided it was the perfect fit. Same underlying figure, two counts, one story. 17:58

What actually makes an idea feel big?

Tension between two points rather than a conclusion on its own. Big ideas come from conviction has no hot tap and no cold tap, so it states without landing. Rewritten as big ideas having much less to do with creativity than you think and much more to do with conviction, it carries a point A and a point B. The example is his own finished essay, offered as a working rewrite rather than a tested result. 19:05

Sources

Copywriting secrets from Harry Dry

The sections above follow the episode's own order. Each timestamp is the point in the recording where the idea is discussed, so the post reads as a map of the conversation rather than a rearrangement of it.

Quotations come from the episode transcript with spoken filler ("okay", "right", "you know", "um", "actually", "kind of", "sort of") and false starts removed. No words are added: square brackets mark anything inserted for sense, and an ellipsis marks any cut. British spellings are restored, and speech-to-text errors were repaired against context only where the intended word is unambiguous. Where a line appears both in a cold-open teaser and in its place in the conversation, the timestamp cites the second — the point at which it is actually said.

Quotes are stamped with the moment they are said rather than with a speaker's name.