Hello there 👋

How's it hanging?

Just Jack here again today (and probably for the next few weeks too, while Joe is mid move chaos).

Before I jump into today's email, a quick humble brag.

You are reading a newsletter that is officially one of the best marketing newsletters, according to the freaking Content Marketing Institute.

And, as cheesy as that sounds, that's only possible because legends like you open it, read it and reply to it every week.

So thank you! (And thank you to the Academy. And our parents.)

Now, enough of that. Let's get on with the show.

And if it feels like you can't move for people telling you that "taste is more important than ever" at the moment, today's email is for you.

Let's get into it.

💡 This week's big idea: being able to articulate what you like and don't like is far more important than just having taste. Sick to the back teeth of hearing the word taste right now? Yeah. Me too. It's ev-er-y-where. At the same time, am I going to be a massive hypocrite and add to the conversation around taste? You know it. But that's only because this video with Rick Rubin came up while I was mindlessly scrolling through YouTube and he said something that I thought was pretty interesting. Here it is 👇 ​ If you're not familiar with Rick Rubin, he is the perpetually barefoot co-founder of Def Jam records that has produced records for everyone from Public Enemy to Adele to Johnny Cash to Metallica to Lady Gaga. (That's range, baby.) Dr Dre called him "hands down, the dopest producer ever that anyone would ever want to be, ever". But Rubin can't play any instruments or read music, at all. And according to Rubin, that's all because of his taste. He says: I have no technical ability and I know nothing about music. I know what I like and what I don't like. I'm decisive about what I like and what I don't like. And in the age of AI, that is the thing that we're supposed to be doing. Being decisive. Knowing what we like and don't like. Basically, to be the "human in the loop" guiding AI to do our bidding. The problem is, here's what Ozzy Osbourne said about working with Rick Rubin: "I still don't know what Rubin did. It's, like, 'Yeah, that's good.' 'No, don't do that.' And you go, 'Why?' And he'd say 'Just don't do it." Now, having taste but not being able to articulate it is fine-ish when you're speaking to creative humans that have their own taste and where "no, do better" puts them in a sort of combative, "OK, watch this" creativity mode. It's kinda dickish, but it'll work. But for AI, that's not going to get you anything different. (We've talked about why this is before.) Which means taste isn't really the thing that's going to make a different to your AI output. If anything, really knowing your taste and not being able to articulate it is likely to make working with AI even more painful. But being able to articulate why you do or don't like something? That is what's really going to make a difference. (Note: I just went back and rewrote that sentence to remove the word game-changer because even though I wrote it, it felt like AI had written it. Isn't that weird?) Back on topic... Luckily, being able to articulate your taste? That's a muscle you can train. (And I know, because I had to learn how to do it too.)

If there's one thing I've had to actively learn how to do (and continue to practice) over the past almost two decades of flogging words for money, it's been articulate why I like copy or don't like copy or why something is just plain alright.

Even when I was a green and keen copywriter, I knew my taste.

I knew I loved Patagonia's copy. I knew that almost everything about Brewdog's copy made me do this face👇

And (whisper it), I knew that I didn't love innocent's voice as much as I thought I should love innocent's voice.

But I never really asked myself why.

My thinking around why I didn't like things (or why I did like things) was limited to "that's a bit try-hard" or "ugh, bit twee" or "so good, love it!".

And that wasn't really a problem for a very long time.

In fact, all the time I was doing all the writing myself, it didn't really matter if I could articulate my taste because it was all in my head/gut and making its way onto the page.

But that all came crashing down when I was asked to work on a 300+ page website and verbal identity project for a UK finance company in the late 2010s.

And because the project was so massive (and this was pre-AI), I was responsible for the voice and the messaging and the big picture pages, but they hired some junior copywriters to chip away at the smaller pages and report back to me.

But because I couldn't articulate my taste very well (nor my taste for this brand) and we were up against a looming deadline, I ended up rewriting damn near 50% of the copy the junior copywriters gave me.

And the copy wasn't bad. Not at all. (I just dug back through my old Google Drive to make sure.)

Sure, the bits of regulatory copy and incorrect claims and things that were obviously not right I could flag and explain.

But it was the nitty gritty where it wasn't quite right, but it was 80% of the way there. That's where I struggled.

And I hadn't practiced how to identify the specifics of what I liked/didn't like or how to articulate it in a way that was actionable and useful but I didn't want to do a Rick Rubin-style "nah, don't do that" either.

So it was quicker to just rewrite it all myself.

☝️Sound familiar?

Almost every single person we've spoken to about using AI to write copy recently has said something like "sometimes it's so bad and off-brand that I just end up writing it myself rather than spending hours asking for new drafts".

That's why (post-RewriteGate, at least) I've made it part of my weekly habit to grab a coffee and spend time just looking at copy and asking "why do I like this?" and "why does this work?"

Likewise, if I see copy I can't stand or that doesn't quite land for me, I do the same thing. "Why didn't I like this?" or "What would it take to fix this up until I did like it?"

In fact, that's the whole reason our swipe file is set up to have sub-pages that dug into the why of it all.

We wanted to make sure every entry has our initial notes on why we loved it (AKA our gut taste) and then a much deeper dive into why we think it's clever, the techniques it's using, why we think it's clever... (AKA, how we train ourselves to articulate taste).

Let's do a little exercise together to build your taste muscles

Last week, Joe sent me this bit of copy by HelloFresh that got a lot of reactions online.

We're going to feature it in a swipe file email soon, but today, I'm not going to comment on it at all.

Just take a look and see what you think.

Now you've seen it, ask yourself three super-simple questions:

1. Do you like it? Listen to your gut reaction here. No thinking or intellectualising it. Yes? No? Meh?
   ​
2. Then, ask why you feel like that? What do you like about it specifically? What don't you like? If it's a big no, why is that?
   ​
   Dig deeper too. What line or phrase gets a big tick? What would you change and why?
   ​
   If you don't like it, is there a version of this you would like? Is it the execution or the idea?
   ​
   Likewise, if it's a yes, is it a yes because you feel like you should like it? (Like me and innocent's voice.)
   ​
   Sometimes I find that I can really like the sentiment of a bit of copy or I can love the idea itself and the bravery to do it, but something about the execution leaves me a bit colder than it should.
   ​
   ☝️It's those kind of revelations are so good for figuring out the specifics and the granular nuances of your taste.
   ​
3. Finally, ask whether your brand would say this?
   ​
   ​There's a big difference between "I like this" and "our brand would say this".
   ​
   If your brand wouldn't say this, why not? Is it the humour? The innuendo? The tone?
   ​
   Or is it something a bit bigger than that? Is it that you're not this familiar with your audience?
   ​
   Taste as a personal preference is great if you're writing to promote yourself. But if you're writing for a brand, your taste has to filter through that brand too.
   ​
   Joe and I are both absolute suckers for pop culture references in copy, but it's definitely not right for every client. When we made a "by the power of Grayskull" joke in copy Discord? Perfect. If I'd added it to the copy I wrote for OVO Energy? Not so much.

I know this exercise feels a little basic and hand-hold-y.

But when you work through those three questions, you're not just learning to spot the details you like/don't like, you're also developing the vocabulary and language you need to explain your taste to your colleagues and AI.

Which means when anyone (robots or human) writes something that isn't on-brand at all, you'll be able to articulate why you don't like it a little better.

Instead of "that's not quite right, can you rewrite it and make it friendlier and more in our voice?", you'll find yourself saying "This is getting closer, but you're making a joke that doesn't need to be there. Remove that joke and instead, focus on creating a connection with the reader."

☝️ And that is the kind of detailed and nuanced feedback that AI can work with to write better copy.

Like with everything else, AI copy has made quantity much easier but quality much harder.

There's this quote I like about being creative and taste from Ira Glass, host of This American Life:

And as somebody who wrote a lot of bad copy before I wrote any half decent copy, that's how things were for me.

But things are a lot muddier now.

The people writing copy now aren't just copywriters.

In fact, everyone I've spoken to that's been tearing their hair out over AI copy has been a founder or a marketer or a designer that has had "copywriting" and "AI wrangling" sneak into their job description in the last few years.

And, completely understandably, that means that they don't have those 10,000 hours of writing, reading, studying and really obsessing over copy wired into their gut.

And, as if to make things even harder, AI is giving them copy that's technically serviceable.

It has a structure, argument, a voice (even if that voice is Generic AI voice), your USPs, it builds to a point...

And that makes the job of going from OK to good so much harder.

Let's say we wanted to learn to draw.

We'd start doodling stick men. Then we'd do crappy looking people with wonky proportions. Then we'd draw stuff that looks alright if you squint and tilt your head. Then we'd get into oh, man, that's pretty good! territory. Then (if we stick with it) we'd end up at the stage where people say "oh wow".

And every time we took a step up that ladder, we'd do that by being able to spot what we did wrong in our last sketch and fixing on the next go around.

It's learning 101. Be bad. Improve a bit. Be a bit less bad. Improve a bit more. And so on.

Basically, this 👇

But with AI, you don't have to go through that ego death voyage through the Valley of Sucking.

You get given something that's already 70% there and then you have to be able to spot and articulate the nuances and issues that are wrong with it in order for it to be better.

That's super tricky.

It's like you've skipped all those lessons on learning to draw, but you're now being asked to judge some GCSE-standard art and write down how it could be improved.

You know it's not perfect. You know it's not professional quality yet. But it's pretty damn good. And so the changes it needs to get better are probably tweaks and nuances and things that aren't immediately obvious.

AKA, things you need to have a tonne of first-hand experience or taste in to spot.

Which means with AI, you're using a tool that's supposed to mean you don't need a trained eye to use it, but you also kinda need a trained eye to spot the things that aren't right and fix them.

In other words, it's a Catch 22 situation where to get AI to write better copy you need to know how to write better copy at which point you don't need AI to write copy...

It's a knot that's tough to unpick.

Which is why we're such big proponents of training those taste muscles

We know we've said before that best way to get something decent out of AI is to have proper voice documentation and copywriting skill documents and a system in place.

That's still true.

But it is a lift that you probably don't have time to DIY.

So the second easiest way to get something decent out of AI is to just have a strong opinion, articulate it clearly and never, ever budge.

Remember, AI will always want to pull you to that beige middle ground. It will always want to sand the edges off. It will always remove the stuff that's unique and interesting.

​This is called Galton's Law of Mediocrity. A recent study found that when AI is asked to create ads, it's the creative stuff (metaphors, emotions, visual cues) that gets the chop first, while factual stuff survives longest.

And even if you ask it to add that stuff back in, it can't recreate distinctiveness or emotional connection.

Those are the things you have to fight tooth and nail for. To train yourself to spot the stuff that's disappearing (or not appearing at all) in the AI copy and to know your taste well enough to know what needs adding back.

To be able to say "this is functionally fine, but it feels soulless, how can we fix it?"

Because fighting back and sticking to your guns takes you from AI copy like this:

To AI copy like this:

That was with just a few rounds of iteration with feedback like "Don't lead with exaggerated pain points and fear, just lead with frustration" and "Make it feel more like a friend talking to a friend about a shared experience, not a brand selling paintbrushes".

Which is why training that taste muscle makes all the difference.

Rewriting your AI copy for hours on end is no different to me rewriting that copy for all those 300-some pages.

It works, but you'll burn out fast.

Once you learn to spot and articulate what you're after and why, you'll be able to get AI to get you 90% there and then tidy the rest.

And the good news is that once you start making a mental note of your taste, your brain won't stop.

(Ask me how I know.)

So next time a bit of copy catches your eye, good or bad — whether it's an ad, a social media post or something we share in this newsletter — don't stop at your gut reaction. Dig a bit deeper. Run through those three questions in your head again.

Eventually, you end up with a little mental library of things you like/don't like that you can spot from a mile away.

And if you really want to speed run figuring out your taste, here's a pro tip: go and check out our swipe file, find a bit of copy you don't like in there and then read why we liked it (or at least respected the craft) and figure out where you disagree with us.

In my experience, there's nothing like a good bit of contrarian thinking to really sharpen your understanding of your own taste.

(Being forced to figure out why I didn't love innocent was arguable far more useful and rewarding than explaining why I didn't like Brewdog's copy.)

And remember, we're always up for talking shop and popping the hood of bits of copy or messaging. So if you find some stuff that makes you feel something (either good or bad) and you're not sure why, hit reply and we'll tell you what we think to kick things off.

Until next week, take it easy ✌️

Jack and Joe co-founders, co-brothers & co-pywriters @ Do Words Good​

PS. If you need a hand figuring out how to document your voice so AI can use it or pinning down your brand's voice and messaging, hit reply and we'll chat.