Hello there 👋

How's it going?

Apologies we're landing in your inbox a day late. Yesterday was buuusy and as we went to do one last read-through of the newsletter before we hit send... we realised that our brains were mush.

And with today's email being a topic that can so easily fall into "woah, hot take" territory, we wanted to make sure we looked it over with fresh eyes.

Right now, we're in the very early stages of updating a book that Jack wrote a few years ago on how to DIY your messaging & brand voice.

And in that book, there is this smart-arse remark that Jack was very smug about when he was writing👇

8 years later and that doesn't quite ring so true, does it?

Because the not-quite-human copy is everywhere.

But before you roll your eyes, close the tab and say "ugh, another AI bad article from copywriters, how revolutionary"... this isn't that.

Pinky promise.

Let's get into it 👇

Our process? Think of your copy like you're a director of a film...

In the 1960s, some French cinema buffs said there were two types of film director:

The first type of director was called a metteur‑en‑scène or "one who puts on the stage". These directors were highly proficient and could deliver a technically impressive, perfectly enjoyable film that lacks any authorial voice or clear point of view.

Nowadays, think almost any director of your average Marvel film. (Exceptions apply, of course.)

The second type of director was called an auteur. These are filmmakers whose vision and voice find their way into every frame of the film. The kind of directors where the way they shoot a film is so personal and specific to them that nobody else would direct the film in the same way.

Think Martin Scorsese, Greta Gerwig, Wes Anderson...

And when it comes to your copy, even in the best scenarios, AI copywriting is a metteur‑en‑scène.

It learns from billions of bits of copy on the internet then stitches them together flawlessly. It knows copywriting formulas and how to write hooks and all that stuff that makes copy technically work.

And yet, the end result always feels a little hollow and generic, no matter the prompt.

Why? Because it's the element of the unexpected that separates OK copy from good copy. It's that strong sense of deeply-held opinions that makes voices stand out from the noise.​

And that kind of copy can only come from flashes of inspiration and crazy "this might just work" ideas. AI can only remix what's happened before.

The scary thing is, AI isn't just taking the pep out of your copy's step. It's adjusting everyone's idea of what good copy is.

Have you read the New Yorker article about how AI is homogenizing the way we think?

It's a banger. Here's the TLDR 👇

Essentially, Cornell carried out a study recently where they observed as AI helped people write essays.

This is what they found:

Isn't that super interesting?

If you really dig into that, you'll notice that the real issue isn't "AI copy is bad" at all. It's that we often don't feel confident enough in our own writing to stand up for the things that make it pop.

AI copy is what it is precisely because it convinces us to cut and edit those phrases and ideas that make us stand out.

In the same article, Max Kreminski (from Midjourney) says that this loss of identity is a real pattern he sees all the time.

He says that most AI users start out full of creative ideas that they want to execute (AKA, using it as a tool to serve their vision) and very quickly shift into “curationist mode” where they're no longer creating, they're choosing the option that feels closest to what they had in mind.

Think of how you use AI.

Are your standards for the output as high as they would be for your own work? Or that you'd expect from an in-house writer or freelancers? Probably not, right?

We're the same.

In our experiments with AI copy, we found that our standards for "this is good to go" shift from needing to be word perfect for our own writing to being "that's pretty good actually" for AI.

And when we ask it to analyse things we've written, it's always the bits we love best that it tries to cut.

And, if you let it, those successive compromises have a snowball effect.

Over time, your voice and your identity starts to atrophy as AI pulls your copy back to safety of a model that writes the next word based on what's statistically most likely to come next, not what should come next.

Quick note: we think it's worth adding that we know AI didn't create this problem at all. It has just exacerbated it.
​
We've seen misalignment between managers/founders and copywriters lead to exactly the same end result: technically fine but hollow, kinda-lifeless copy. All the bits that made it popped nixed and sanitized before they hit publish.
​
Those copywriting formula/templates you can buy do the same thing too.
​
They're all about sanding down and removing the risky, challenging, creative bits to make your copy as safe and palatable as possible.

And when we know that people love brands that are meaningful, different and salient (AKA stand for something, stand out and easy to remember) and can even sell up to 70% more, it's easy to see that this approach to copy is at odds with what customers actually want to see.

In other words, safe copy is actually risky. And risky copy is safe.

Confusing, right?

Lean into your weird, messy, not-even-close-to-technically-perfect copy.

When we look back over our own work or through our swipe files, our absolute favourite bits of copy aren’t built on a perfectly placed Power of Three.

They're not the headlines we've written when we've stuck to a framework or ticked every single box.

In fact, they're always the bits of copy that feel like they tumbled out of someone’s brain mid‑conversation.

There's weird asides. Off-kilter analogies. Unfiltered, in-your-face bits of copy that make us say “Oh wow… they actually said that?”

(If we preferred technically perfect copy, we'd be called Do Words Well.)

Essentially, we love the kind of copy that brings us back to this quote by Rory Sutherland:

We love them rogue variables.

And they don't have to be big, OOH campaigns or big, impactful h1s, either

They can be as simple as this 👇

That "not your average coke dealers" is the kind of bold, punchy bit of copy that only Karma Drinks could write.

It says everything about them in 5 words.

And it's all the better for being bold and brave.

And the science is on the side of the weirdos and rule breakers, too.

• ​Risky, distinctive copy out‑performs safe presence by 300%​
• A Stanford study found that uncommon wording increased brand recall by 22 %.​
• The Von Restorff Effect tells us that things which don’t “fit” are more likely to be remembered.
• And Nielsen Norman’s 4,000-person tone study showed that a conversational, slightly messy tone outperformed corporate polish on every UX metric, from trust to delight.

But we're not saying "don't use AI to write copy". Not at all, actually.

We know that AI is fast becoming a "use it or get left behind" tool.

The problem is when we let AI go from a copy assistant (AKA, doing the messy first draft) into a full-time copywriter. (Or, worse still, our brand voice guardian.)

Because your customers don’t want technical perfection. They want the distinctive, sometimes a bit messy, sometimes not perfect but always real copy that comes from people.

Let AI tidy things up, sure. Let it help with the heavy lifting. Let it take some scattered ideas and turn them into a messy first draft.

(After all, editing is easier than writing from scratch.)

But don’t let it erase your brand's fingerprints in the process.

And don't let it transition itself from a tool to the guardian of your voice.

Because it's that special sauce, that unique give-a-shit-ness... that's the most valuable part of your brand's copy.​

You just have to remember to fight the robots so you don't lose it.

Here's how 👇

Step 1: really, really define your brand voice.

In the days of skinny jeans and Mr Brightside, a brand voice guide wasn't much more than a collection of a few tone adjectives.

Our voice is funny, sarcastic, optimistic, fun-loving...

And for a while — when most brands didn't take voice seriously — that was enough.

But now, everyone has a voice. And everyone is fighting for the same space in a customer's brain.

Not to mention, the Norman Nielsen group found that AI is crap at producing on-brand copy when you just give it tone descriptors.​

Instead, both for human and robot writers, you need a brand voice that has things like:

• This, not that guidelines: “Confident, not smug.” “Witty, not whimsical.”
• Cadence rules: We vary our sentence lengths and we're not afraid to start sentences with a conjunction.
• Unique style beats: We use brackets to make conversational asides to our customers. (Like this, hey!)
• Goldilocks examples: here's our brand voice going too far, here's it not loud enough and here's it juuuust right.

(We went into more detail on how to write good brand voice guidelines a few weeks ago!)

Step 2: Give your AI assistant that full brand voice PDF

Here's a pro tip: switch ChatGPT to o3 mode and upload your brand voice guidelines before you ask it to do anything.

Then, ask ChatGPT to explain your brand voice and messaging back to you.

Once it's done that, make sure you're 100% aligned. If it's made errors or not quite getting it, explain where it's missing the mark.

Then, switch to whichever mode you like to use to write.

Now, ask it to use that understanding to give you three copywriting examples that are a perfect match for your brand voice.

(Even better, ask for a product description for one of your products, a welcome email and a new h1 for your website. It does better with specific tasks.)

At this stage, you usually end up with some pretty decent copy that's much closer to your voice than you get from prompts.

But, as always, the copy is always a bit Tom-Hanks-eyes-in-Polar-Express.

That's where you come in.

Step 3: retrain your brain to go from curator to guardian of tone

As you read your copy, remind yourself that the choice isn't "does this get the job done?" or "does this tick all the boxes?"

It's "could only our brand say this?".

☝️That's how you stop that subtle, slow erosion of your voice.

And when you look at copy through that lens, you start to see things differently. You feel like Neo in The Matrix.

You'll notice weird analogies that you'd never say in real life. (AI is baaaad at analogies. Because the best ones come when we spot common links between what we're writing about and our lived experience. Like film studies and AI in this email, AI wouldn't make that connection.)

You'll notice generic ways of describing your brand.

You'll notice that the copy doesn't feel concrete and specific to what you're selling.

Because beneath all of the surface level decoration, AI can never know what you know.

It'll never know the smell of your products. It can never pick up on the subtle ways your customers speak about your brand. It'll never feel the enthusiasm for your products that you feel.

And that's what you need to channel into your copy. That only-we-can-say-this-ness is a game-changer.

Step 4: Forget that teacher over your shoulder and get weird with it

Whether that teacher over your shoulder is Cornell's AI model or Mrs Rhodes in Year 5 telling you that "you have no talent for words. Stop trying to be imaginative and just fill in the blanks"...

(Note to self: book therapy appointment to unpack Year 5.)

Everyone has that little voice that says their writing needs to play safer. To be grammatically correct. To be "professional".

But the best copy comes when you ignore that voice and get weird with it.

And if you're struggling, a good metric is this: would anybody be put off by this copy? Is there anyone that is going to read this and say "urgh, don't like that, unfollow"?

Because if not, you're probably playing it safe. Go bigger.

Because, remember, polarisation isn't a bad thing: all the data shows that putting people off your brand is the key to growing faster.

👋 If you want to figure out and define your brand voice and turn it into docs that get anybody (human or otherwise) writing in your voice, we'd love to help.

We've spent the last 10 years helping ecomm, lifestyle and B2C brands define their voice and document it for their team (and now, robots).

We'd love to help you reclaim your weird next.

​Nab a spot to chat on our calendar here!​

We'll see you next week for more ways to do words gooder.

Peace and love ✌️ Jack and Joe co-founders, co-brothers and co-pywriters at Do Words Good​