Signs AI Wrote Your Copy

There’s a number you’ve probably seen everywhere lately – in marketing emails, in Instagram captions, in reels from coaches and consultants who swear they “barely use AI.”

The number is 47.

“You’ve got 47 browser tabs open.” “You’re trying to do 47 things at once.” “47 new AI tools dropped this week.” Sound familiar? Because once you notice it, you cannot unsee it.

Here’s what’s happening: AI writing tools – ChatGPT, Claude, Manus, all of them – have certain defaults baked in. Patterns they reach for when they’re filling in gaps. 47 is one of them. It’s specific enough to feel real, but round enough that no one questions it. And it’s everywhere. Reddit threads full of marketers going, “…wait, is 47 following me?”

Spoiler: it’s not. It’s following your AI tool.

Why does this matter for your business?

Because if you’re using AI to write your content – and at this point, most of us are – the output is only as good as the prompting, editing, and brand-training behind it. If you’re letting AI write and posting without reading it back, your audience will start to notice patterns you don’t. And 47 is just the most obvious one.

There’s also the rocket ship emoji. Long associated with “hype content” (you know the type – “ This is going to change everything!!”), it’s now so overused by AI tools that it’s become a shorthand for “I did not write this.” Same goes for the fire emoji used unironically, and anything described as a “game-changer.”

These aren’t just aesthetic problems. They’re trust problems. Your audience is smart, and they know when content feels like it came from a template – even if they can’t articulate exactly why.

So what do you actually do about it?

Train your AI on your actual voice. The single biggest fix is not better prompting – it’s better brand training. If you give AI your tone of voice document, real samples of your writing, your banned words list, and your typical sentence structure, the output stops sounding generic. It starts sounding like you. Not perfectly, but close enough to edit rather than rewrite.

Read it out loud before you post it. This sounds obvious, but most people skip it. If you wouldn’t say “47 tabs open” in conversation – or you’d say it differently – that’s your gut telling you to rewrite the sentence. Trust it.

Do a quick “AI tell” scan before publishing. Numbers that feel oddly specific (but aren’t actually your numbers), rocket ships, fire emojis, phrases like “fast-paced world” or “game-changer” or anything described as “transformative” – these are all signals to slow down and edit.

Ask AI to check its own work. Genuinely – paste your draft back in and ask: “Does this contain any overused AI patterns or generic phrasing? Flag them.” It’s surprisingly self-aware about this when you ask directly.

The bigger picture

The irony of the 47 situation is that AI tools are genuinely brilliant at a lot of things. Research, structure, first drafts, repurposing, brainstorming – I use AI for all of it, every day. But “produce content I can post without reading” isn’t something any of them can do reliably yet. Not because they’re bad, but because they’re trained on a massive pool of content that includes a lot of… other AI content. The number 47 is literally the result of AI learning from AI.

Your job isn’t to avoid AI. Your job is to edit better than your competitors do.

Because here’s the honest truth: the gap between AI-generated content that builds trust and AI-generated content that erodes it is mostly editing. And the editing is still a human skill.