It's late afternoon. The newsletter needs to go out tomorrow and you've got nothing. So you open ChatGPT, type something vague about your business, and five minutes later you're staring at a draft that is technically fine – structured, professional, complete – and completely, unmistakeably, not you.

That's not bad luck. It's just how standard AI tools work. They're trained on enormous amounts of professional writing, so they produce something that sounds like professional writing. Your personality, your turns of phrase, the slightly dry observation you'd have dropped into paragraph three – none of that makes it in. What you get instead is a newsletter that could have come from any business, in any industry, aimed at anyone.

Most people post it anyway and wonder why it doesn't land.

The fix is training, not prompting

The reason most people get generic AI output is that they're working with a generic AI. You can wrestle with prompts – add more context, beg it to sound more casual, tell it your audience is "small business owners who value authenticity" — and you'll get marginally better results. Or you can build a custom GPT that already knows all of that before you type a single word.

A custom GPT isn't a different product. It's ChatGPT with a detailed brief built in — your tone of voice, your audience, examples of your actual writing, specific instructions about the phrases and habits that aren't you. The difference in output is significant. Instead of spending twenty minutes editing the robot out of a draft, you're making light adjustments to something that already sounds like yourself.

What actually goes into one

The setup takes more thought than most people expect, which is probably why a lot of custom GPTs underperform. Feeding it one blog post and a rough description of your brand isn't enough. The things that make your voice recognisable – the rhythm, the vocabulary, the level of directness, the things you'd never say – take more examples and more specific instructions to land properly.

Done well, though, it changes how content creation feels. A week of social posts in the time it used to take to write one. Blog drafts that need polishing rather than rebuilding. Emails that don't make you wince when you reread them the next morning. The speed gain is real, but the more useful thing is consistency – your content sounds like you, whether you wrote it on a clear Tuesday morning or asked an AI to draft it at midnight.

What it doesn't do

It won't salvage messaging that was muddled to begin with. If you're not clear on your offer, your audience, or what makes you different, a custom GPT will just produce muddled content faster. The voice training works; the strategic thinking still has to come from you.

Worth saying too: AI writing has tells, and even well-trained custom GPTs produce them if you're not paying attention. The output still needs a human eye before it goes anywhere.

If you want one built properly

I help businesses set up custom GPTs that are actually trained on their voice – not just configured with a system prompt and a vague aspiration. If your current AI content sounds fine but feels wrong, get in touch, and we'll talk about what a proper setup looks like.

Let's talk about creating your custom GPT