AI prompts
for your business.

Ask an AI tool to 'write a Facebook ad for my plumbing business' and you'll get something that could be for any plumber in any country. It has nothing else to go on. The fix isn't a magic phrase. It's context: what you do, where, who for, and what's off limits.

Every prompt here does three things most templates skip. It names your service and area, so the answer is about your market. It tells the tool not to make up facts. And it carries the rule for your industry, so the first draft starts on the right side of it.

The square brackets matter more than they look. AI systems are known to produce inaccurate or false results, and in marketing that usually means a plausible detail nobody supplied: a star rating, a number of years in business, a guarantee you never offered. Asking for a gap in brackets turns an invented claim into a to-do.

Checking what AI says about you. The prompts ask the tool to name its sources and admit what it couldn't find, which shows where its picture of you comes from. If it can't find you at all, that's a finding. The AI visibility check turns the same idea into a tally.

Writing an ad. Google's responsive search ads allow headlines of up to 30 characters and descriptions of up to 90, so the prompt sets those limits. The rule-check prompt asks for a list of possible problems rather than a verdict, because the tool can't make that call for you.

Planning content. Google says its AI features run a set of related searches behind each question. The content prompts list what customers ask at each stage, then the follow-ups behind one question, then the gaps in your page. That's Fan-out Mapping in miniature.

Finding your customers' words. Reviews are the closest thing most businesses have to a transcript of what buyers care about. The prompts ask for exact quotes, grouped and counted, so your ads use their words instead of yours.

The OAIC's guidance on commercially available AI products is direct. As best practice, it recommends organisations do not enter personal information, and especially sensitive information, into publicly available generative AI tools. So no customer names, phone numbers, emails, addresses or account details in a prompt.

In some industries the line is closer than it looks. A patient's treatment, a client's debts, a family law matter or an NDIS participant's plan is sensitive. Under privacy law, health information is sensitive information, so pasting a real review from a dental patient may mean pasting health information. Strip it out, or write a made-up example instead.

Your own business details are fine. Your name, suburb, services and prices are public anyway. The test is simple: if you wouldn't print it on a flyer, don't put it in a prompt.

The same guidance says not to record or treat AI outputs as facts. For marketing, the risk isn't embarrassment. It's publishing a claim you can't back up. Under the Australian Consumer Law the business is responsible for false or misleading claims in its ads, whoever drafted them.

Some regulators say it outright. The Tax Practitioners Board's guidance on AI says practitioners stay responsible for the work and should verify AI-generated content. If you're in health, legal or finance, read every AI draft against your regulator's advertising rules before it goes anywhere. The ad rules checker is a fast first pass.

A routine that works: check every number, name and date against a source. Delete anything you can't prove. Read it aloud to hear whether it sounds like you. Then run ads through your industry's rules before they go live.

Give it an example. Paste one ad or page you're happy with and ask for more in the same style. One good example beats a paragraph of instructions.

Ask for options, then choose. Eight headlines and three descriptions give you something to pick from, and the best line is often a mix of two.

Tell it what to leave out: no exclamation marks, no 'solutions', no 'quality', Australian spelling. That saves a round of edits every time.

Run the visibility prompts in a private or logged-out window where you can. Your own chat history can shape the answer, so a clean window is closer to what a stranger sees.