AI and copywriting

Can AI write my copy?

Mostly, yes. It will be grammatical, structured and perfectly adequate. That is a more interesting answer than it sounds, and it changes what you should be paying a copywriter for.

I use AI in my own work. Pretending otherwise would be both dishonest and, given what I teach, slightly ridiculous. So here is the useful version of the answer rather than the defensive one.

What AI genuinely does well

• Structure. Ask for a landing page and you will get a sensible shape: headline, problem, solution, proof, call to action.
• Volume and variations. Twenty subject lines in ten seconds. For ad copy, where you want angles to test, that is genuinely useful.
• First drafts of things you already understand. If you know what the page must argue, a model will get you to a draft faster than a blank screen will.
• Summarising research you have already gathered. Point it at fifty customer reviews and it will find the patterns quickly.
• Tidying and tightening. It is a good editor of mechanical problems.

If your situation is on that list, you may not need to hire anyone. I would rather say so than sell you something you can do yourself.

What it cannot do, and why

A model is trained on what has been published. It is very good at producing the average of everything written about your topic. The problem is that the thing which persuades your particular buyer is almost never on the public internet.

The argument that closes a sale usually comes from somewhere private: the objection your sales team hears on every call, the reason three customers asked for refunds last quarter, the phrase a client used in an interview that you would never have written yourself. None of that is in the training data, because none of it was ever published.

So a model will give you something that reads well and argues nothing specific. A template produces a page that reads fine and persuades nobody in particular.

The convergence problem

Here is the part worth thinking about if you are weighing this up properly.

You and your three closest competitors are using the same handful of models, trained on the same public web, prompted in roughly the same way, about the same market. The outputs converge. Everyone’s copy starts to sound like everyone else’s copy, because it is drawn from the same pool and averaged the same way.

Which means the only durable source of difference left is the inputs nobody else has: your customers, your sales calls, your failures, your actual numbers. Research stopped being the unglamorous part of copywriting and became the entire competitive position.

That is the quiet irony of cheap writing. It made the research more valuable, not less.

What you should be paying for now

Not the typing. The typing is close to free and will keep getting cheaper.

• Deciding what to publish at all, and in what order, which is a content strategy question before it is a writing one
• Research into your specific buyer, because that is the input a model cannot reach
• Judging the offer, since no amount of good writing rescues a weak one
• Knowing which kind of discovery a page has to win: search, citation in an AI answer, recommendation, or conversion. These need different pages
• Being accountable to a number, which is the thing no tool will ever do for you

Cheap production without a strategy mostly produces more of what was already not working. A team that publishes four times as much and converts the same amount has not gained anything except a bigger archive to maintain.

So when do you actually need a copywriter?

When the page has to carry real weight: a sales page, a launch, a site rebuild, a campaign with budget behind it. When the offer is complicated or the objections are serious. When nobody internally agrees what the page is for. And when you need someone who will tell you the problem is the offer rather than the adjectives.

When you do not: internal updates, routine listings, a first draft of something you already know how to argue. Use the tools, keep the money.

Frequently asked questions

Will AI replace copywriters?

It has already replaced a large part of what junior copywriters used to be paid for, which is producing competent words on a known brief. What it has not replaced is deciding what the brief should be, researching the buyer, judging whether the offer stands up, and being accountable for the result. The work moved up the stack rather than disappearing.

Do you use AI in your own work?

Yes, for drafting variations, summarising research I have gathered, and tightening sentences. Not for deciding what a page should argue, and not for the research itself. I also teach how to use it, because a marketing team that uses it badly produces more work for everyone downstream.

Can Google tell if my content is AI-written, and does it matter?

Detection is unreliable and beside the point. Google has been consistent that it judges content on usefulness rather than how it was produced. The real risk is not a penalty for using AI; it is publishing a lot of unremarkable pages that nobody links to, nobody cites and nobody acts on. That fails on its own merits without any detection involved.

If AI can write, why hire a content strategist?

Because production capacity without direction just multiplies the existing problem. The scarce skill is deciding what to publish, for whom, and what each piece has to win. That is what a content strategy consultation is for.

Can I give AI my research and get good copy out?

You will get much better copy than without it, and this is the most underused approach I see. The catch is that gathering the research properly, customer interviews, objection lists, win and loss reasons, is most of the work. Teams rarely skip the writing and do the research. They almost always do the reverse.

Should I still invest in content if AI answers take the clicks?

Yes, but with a different target. If an AI answer satisfies the question without a click, the win is being the source it quotes and the business it recommends. That is answer engine and generative engine optimisation, and it needs content built to be extracted and corroborated rather than merely to rank. The content strategy page sets out the four kinds of discovery and what each requires.