The worst way to use AI: letting it publish posts nobody checks.

Here's a pitch you've probably seen. A tool for forty-nine bucks a month writes four blog posts, drops in the keywords, and publishes them on a schedule. No writing. No checking. Nobody reads a post before it goes live with your name on it.

Here's what I think of that. Of every place you could point AI at your business, letting it publish generic posts nobody checks is close to the worst one. Not because the writing comes out bad. It usually comes out fine. That's the problem.

To be clear about where I stand: this blog is written with AI assistance, and it says so under every byline. So this isn't a post about AI being off-limits for writing. It's about the difference between using AI to draft and letting it publish.

"Fine" is the failure state

A blog post exists to do one thing: make a stranger believe you know what you're talking about. It's not there to hit a word count. It's there so somebody who's never met you reads four paragraphs and thinks, okay, this person does this for a living.

Competent, inoffensive, average writing does not do that. It does the opposite. It signals that nobody was home.

And average is exactly what you get when nobody steers, because that's what the tool is built to produce. It has never been to a job site. It doesn't know that you stopped offering the cheap package because it attracted customers who called you nine times before signing. It doesn't know that in Northeast Florida the humidity destroys a certain kind of installation in four years. It doesn't know any of the things that make you worth hiring instead of the guy two exits down.

It knows the average of everything ever written about your industry. Your competitor bought the same tool. He's publishing the average too.

Unchecked is worse than generic

Generic posts waste your time. Unchecked ones can cost you a customer.

AI writes with the same confidence whether or not it's right. Left alone, it will happily give your business a price you don't charge, a service area you don't cover, a turnaround you've never promised, or a "happy client" who doesn't exist. On an auto-publish schedule, nobody catches it. It goes live with your name on it, and the first person to notice is the customer who quotes it back to you.

That's the line for me. Every claim a post makes about your business has to match what you sell, at the price you charge. If nobody is checking, the tool shouldn't be publishing.

You spent your AI budget in the room with no money in it

Set aside quality entirely. Say the posts are good. You've still aimed your one new capability at the thing on your list with the least leverage.

Because for a lot of small businesses, the blog is not what's broken.

What's broken is that a request for a quote came in on Thursday and went out the following Wednesday, and by then they'd hired somebody else. What's broken is that you take good notes on a site visit and then they sit in a truck console for a week. What's broken is that four people asked the same question this month and you typed the answer from scratch four times.

Nobody sells you a $49/month tool for that. It doesn't demo well. But that's where the hours are, and that's where the lost jobs are.

Where I point it

A workflow audit isn't about content. It's about the ugly middle of the business: the stuff that's boring enough that nobody has ever thought to fix it.

• Turning a conversation into a document. Record the walkthrough on your phone, run it through something like Otter, and you have a transcript that becomes the scope, the estimate, and the follow-up email, the same afternoon instead of next week.
• Answering the same question the same way every time. A custom GPT loaded with your real pricing, your real scope, and your real policies gives a consistent answer at 9pm when you're at your kid's game. Not a made-up one. Yours.
• Draining the inbox. Sorting what came in overnight into "needs you," "needs a reply I can draft," and "ignore." Ten minutes instead of an hour of dread.
• The follow-up you never send. Nine drafts, one per open estimate, each referencing what you discussed. You read them, fix them, hit send. Fifteen minutes for something you've been meaning to do since March.

None of that shows up on your website. Nobody will compliment you on it. It just quietly gives you back hours and stops jobs from leaking out the side of the funnel while you're on a ladder.

If you want to find those leaks in your own week, the free AI Time-Leak Checklist is a good place to start.

How I use AI for writing

I use AI constantly when I write, including here. But I use it in a specific way, and the distinction is the whole point.

I use it to draft and research, and as an argument, not an author. I'll tell it what I think and ask it where I'm wrong. I'll have it cut a paragraph I'm too attached to. I'll ask it to find the four sentences doing no work. I'll dump a rambling voice memo into it and get back an outline of what I was trying to say.

Then the post gets checked. Every price, timeline, and claim about my business has to match my published services and prices, or it comes out. Nothing goes up because a tool said so.

What I never do is ask it what I think. That part isn't automatable, and it's the only part with any value. If the post doesn't start from something I believe and could defend to a client on the phone, no tool is going to rescue it. It'll just make the emptiness grammatical.

It's the same way I handle blog writing for clients: I use AI to draft and research, then edit every post for your voice. You approve each post before it goes live.

Google's own guidance on AI-generated content says roughly the same thing, for what it's worth: it isn't out to punish you for using AI, it's out to reward content that's helpful. Which means auto-generated posts don't lose because a robot wrote them. They lose because they're not useful to anybody, and that was true long before AI existed.

The test

Before you publish anything, read it and ask two questions. First: is every fact in here true about my business? Then: would I say this out loud to a customer sitting across from me?

Not "is this on topic." Would you say it (in those words, with that little bit of opinion in it) to a person who's deciding whether to hand you money?

If the answer to both is yes, publish it. If you'd be embarrassed to say it out loud because it sounds like a brochure, nobody wants to read it either.

Four checked, honest posts a year that sound like the person who owns the business will do more for you than forty-eight that nobody read before they went live. And the time you save is time you can spend on the quote sitting in your drafts folder right now.

---

Not sure where AI would earn its keep in your business? Here's how my AI integration work runs, or book a free call and tell me what eats your week. I'll tell you straight whether it's worth automating, or whether you just need to hire somebody.