It is 9pm on a Sunday and the kitchen table is the office again. Laptop open. A mug of tea that went cold an hour ago, still there.
The doodle you did on Friday came up beautifully, and you have been looking at the caption box for eleven minutes without typing a word. Behind it sits the message to the client whose dog arrived matted again, started and deleted three times. And the price rise you have meant to announce since April.
None of it is difficult. All of it is why you are still up.
ChatGPT for dog groomers works when you give it four things: who it is, what it has, what to produce, and what to avoid. That is the honest version of what a chatbot does for a groomer. It will not run your salon. It is good at the first draft of something you already know how to say. All 20 prompts below do those four things, and every one is written to copy straight out.
How to use ChatGPT for dog groomers without getting mush back
A one-line instruction gets a one-line answer: the average of everything ever written about dogs on the internet. That is where "AI writing sounds like AI writing" comes from. It is the instruction, not the tool. (What a prompt actually is, if the words here are new.)
Every prompt below has the same four parts:
- Who it is. One sentence puts it in your world, not a generic marketing one.
- What it has. The slot, the price, the notice period. Leave a fact out and it fills the gap itself.
- What to produce. A format and a length. Vague in, vague out.
- What to avoid. The most-skipped part, and the one that changes the output most. No emoji. No apology. And, over and over, do not invent anything I have not told you.
Each one sits in a box you can copy in one tap. Paste it as it is the first time. Then change the figures and the notice periods to your own before you use it again, because the ones here are a worked example, not your salon.
Why it gets things wrong is a post of its own. The short version: a blank it cannot fill is a blank it will fill for you, confidently. Saying so in the prompt is cheaper than catching it later.
The examples use Steph, a two-groomer salon in Sydney's Inner West. She is a worked example we follow through this series, with numbers you would recognise rather than a real client's books. Her casual full groom is $95, a nail trim is $20, and a de-matting surcharge is $30.
The one rule that comes before all 20
Every prompt below is written so you never paste a client's details into a chat box. Where the job needs client context, the prompt says "the client" or asks for initials, and you add the real name afterwards in your own app.
The Office of the Australian Information Commissioner recommends that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools, because once it is in there it is very difficult to track and potentially impossible to remove. The part you need here:
Never paste into an AI chat:
The full list, and the law behind it, is the next post in this series.
You do not need a paid plan for any of this
Every prompt below is words in, words out: no uploads, no images, no subscription. The free tier runs all 20. OpenAI does cap how much a free account gets and publishes no number for it, so the app tells you when you have hit a limit and when it resets.
OpenAI's plans are Free, Go, Plus, Pro, Business, and Enterprise, and "Team" is the retired name for Business, so anything telling you to buy ChatGPT Team is out of date. Business starts at two users, which matters when you are the only one in the building. Every price OpenAI publishes is in US dollars rather than Australian, which matters more than it sounds: what these tools actually cost an Australian business is its own post.
The model picker matters less than the prompt. For which tool suits which job, start with the decision chart.
Five ChatGPT prompts for the messages you dread writing
These are the five that clear the 9pm backlog. Each one names the situation rather than the person, so nothing about a client leaves your own app.
1. Filling the slot that just opened. Good output reads like a text you would send.
2. The matted coat conversation. Moose, Steph's Groodle, arrives pelted every time. Good output never blames.
If it keeps happening, the real fix is a shorter maintenance cycle, not a better plea.
3. The late cancellation. Good output states the fee once, then moves on.
4. The unhappy client. Good output takes the feeling seriously and asks nothing of them.
5. The client who quietly drifted. Good output never mentions the gap.
Four ChatGPT prompts for social posts and captions
Captions are where a chatbot is at its most useful and its most generic. The fix in all four is the same: give it one real detail from the day, and tell it plainly what it may not make up.
6. A week of captions from one line of notes. Good output gives five real angles, not one reworded.
7. The seasonal post that is useful. Good output names one thing to do today.
8. Explaining your price without arguing about it. Good output walks the two hours in order, never sounding sorry.
9. Turning a good review into a post. Strip both names before you paste. Good output quotes one short line and adds nothing the reviewer did not say.
Four ChatGPT prompts for the admin pile
This group is not writing, it is sorting. Ask for a shape you can act on, a list or a set of three emails, and say what to do with the items it cannot work out on its own.
10. The brain dump, sorted. Good output leaves the vague items vague.
11. The job ad for a bather. Good output describes the actual day instead of selling it. If it invents a wage, run it again.
12. Chasing the invoice you have been avoiding. Good output hardens across the three emails without turning nasty.
13. The closure notice, three places at once. Good output repeats the same two dates without drifting.
Four ChatGPT prompts for pricing and policy
Money is where invention costs you most, so each of these four either hands it your real figure or forbids it from producing one. Check the arithmetic yourself before any of it reaches a client.
14. The price rise notice. Good output carries no apology, and predicts the objections you dreaded.
15. The questions you should ask before quoting. It refuses to give you a price on purpose.
16. The mid-groom surcharge call. Good output sounds like speech, not a letter.
17. Checking the maths on a package. Good output shows every step, and stops if a number is missing.
You are retyping the same three numbers into every prompt here. There is a way to hand over your prices, your policies, and the way you write once: the next rung up from here.
Three ChatGPT prompts for when you are stuck
These three are the opposite of the rest. You are not asking for a draft, you are asking it to interview you, argue with you, or find the hole in something you have already written.
18. Make it interview you. The best prompt for a blank page. Good output asks one question and waits.
19. Make it argue against you. Good output is uncomfortable to read.
20. A second pair of eyes on your own draft. Good output hands you no rewrite.
Two settings every dog groomer using ChatGPT should find first
Data Controls. On a personal account your conversations are used to improve OpenAI's models unless you turn that off. The setting is called "Improve the model for everyone", under Settings, then Data Controls. Switching it off leaves your chat history where it is and only stops future conversations being used for training. It does not reach back over what you typed already.
Temporary Chats and memory. Memory sits under Settings, then "Personalization", which is how OpenAI's menu spells it, and lets it remember details about you between conversations. Useful, and also somewhere details end up. For anything you would rather was not kept, use a Temporary Chat: those draw on no existing memories and create none.
Projects, on every plan including free, hold their own instructions. One called "Salon messages", carrying your tone and notice period, stops you retyping that context every time.
What this actually reclaims for you
Say it takes you 20 minutes to write a message you have been avoiding, and five minutes to edit one already sitting on the screen. That is 15 minutes back. Do that four times in a Sunday session and you finish an hour earlier.
Steph's admin evening is two hours, twice a week. An hour off each of them is two hours a week and close to 100 hours a year, and not one of those hours came from working faster on the bench.
That is the whole promise of this level. Not a cleverer business. Just the Sunday back.
Before you paste anything else into a chat box, read what you must never put into one. Eight lines, two minutes, and the one part of this series that is not optional.