AI is a power tool.
Here's the owner's manual.

Some owners will tell you AI writes everything for them now — estimates, emails, the job ad — and swear it's magic. Others won't let it near the business, because they heard it makes things up. Both camps are half right, and the half each camp has wrong is the half that costs money.

The first camp is right that it's fast. They're wrong that it's magic, and the day they let it invent a price or a credential, they'll find out how wrong. The second camp is right to be careful — these tools really will state things that aren't true, flat-out, with total confidence. Where they're wrong is thinking that makes the whole thing junk. A nail gun will put a nail through your hand. You didn't throw it out. You learned what it's for.

That's the frame this whole manual hangs on. AI is a power tool for words. Hand it your real facts — prices, service area, hours, what happened on the job — and it turns them into a clean draft, rewords them for a different reader, and tidies the spelling, faster than anyone in your office. Ask it for a fact you didn't give it, and all it has is a guess dressed up as an answer. It isn't lying. It just doesn't know anything about your shop, and it won't tell you that. You are always the source, and you are always the inspector. Every method in this manual is built on those two jobs staying yours.

Trying it costs nothing

This stopped being a fad while nobody was looking

Before the chapters, here's why it's worth your time. Real customers now show up at real businesses from these tools. As of July 2026, Google Analytics — Google's visitor-counting tool — includes a built-in "AI Assistants" traffic channel for exactly this, counting visitors arriving "from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok". Google doesn't build a counter for traffic that isn't happening. The details — what that channel catches, what it misses, and why visits from Google's own AI answers land in your regular search numbers instead of this one, as of that same July 2026 check — are a numbers question, already covered elsewhere. The plain-English story is on what SEO actually is. Reading your own counts is on reading your numbers. This page won't re-teach it.

The six chapters, and what each one does for you

• How AI actually works Rough-in What the machine is really doing when it answers, and why it makes things up, in the builders' own published words. Read this one first; everything else leans on it.
• The cutoff board Rough-in The date each big tool's built-in knowledge stops, read live off the vendors' own pages — including the two big vendors that won't publish one.
• The prompt method Rough-in The fill-in-the-blanks way to hand the tool your facts and get a usable draft back, with the verify pass that keeps you the inspector.
• AI copy jobs Rough-in The writing jobs around a shop it genuinely does well, plus the hard rules, including the federal one about fake reviews.
• The slop tell Rough-in How to spot canned AI text, in your own drafts before they embarrass you and in a vendor's "deliverables" before you pay for them.
• AI in the office Rough-in What these tools do with what you type into them, and the settings that stop your chats from being used as training material.

The standing rule

Everything above folds down to one table. This is the rule the rest of the manual just applies, one job at a time.

Left column: hand it the facts and let it work. Right column: you supply it, or it doesn't go in. That's the whole rule, and it doesn't bend for a better model or a bigger promise.