AI can help you. It can't be you.

AI can help you. It can't be you.

MoonMagicx isn't anti-AI. I'm anti-generic. I'm anti the feed filling up with the same soft sentences until none of you sound like anyone in particular.

This whole training is about connection. People need to connect with you. Not an AI version of you. Not a polished stand-in that never snaps, never forgets the tea, never has lasagne on the floor. If we're teaching you to build social media through actual connection, the people on the other side have to meet the actual person. Otherwise we've spent five weeks practising a relationship with a ghost.

Used well, AI is a helpful assistant who doesn't get tired. It can throw you twenty angles on a car-park morning. It can turn a rambling voice note into a list. It can help you find a hook, tidy your spelling, check your grammar, group twelve messy thoughts, show you a different way into the same story. Used badly, it becomes the person posting. Your people can tell it isn't you. They might not know why they're bored. They're bored because the real you isn't in the caption.

What AI is good for

You can use it. I want that said plainly, because some of you will hear this lesson as "throw the tool in the bin". Don't throw the tool in the bin. Use it for the jobs it is actually good at.

• Brainstorming when you have a moment and no shapes yet. "Give me ten ways to talk about forgotten tea that are not gratitude lists."
• Ideas when you are stuck. "I have a lost shoe, cold tea, and a journal in the door pocket. What could this become besides a rant?"
• Organising thoughts. You dumped twelve things. Ask it to group them. You still choose.
• Hooks. "Here is the story. Give me five opening lines. Do not add the word journey."
• Spelling and grammar. That's a kindness to the reader, not a personality transplant.
• Seeing different angles. "What would a sceptical friend say I am missing?"
• Unsticking. "I sound like a brochure. Ask me five questions about my actual morning."
• Accessibility. Alt text, a cleaner structure, a shorter version of a long ramble you already wrote.

Notice the pattern: you still bring the life. The machine cannot remember the raisins or the cupboard or the receipt unless you put them in. AI can help you find the words. It cannot be the person saying them.

What you have to put back in

This is the bit most people skip, which is how you end up with a grid that sounds like everyone else's.

After the tool has helped, you put yourself back in:

• Your actual experiences. The Tuesday. The snap. The tea. The shift. The lasagne. If it did not happen, it does not go in.
• Your opinions. Not a quote the model likes. A stance you would defend in a kitchen.
• Your humour, if you have it. If you don't, don't fake Bee. Dry is a humour. Quiet is a humour. "I do not do jokes" is also a voice.
• Your language. The words you actually say. Not "holding space" unless you have ever said "holding space" to a human being you live with.
• Your stories. The ones that happened to you, with your objects in them.
• Your personality. Priya dry. Jonah short. Bee sweary. Elise careful. You, whatever you are.
• Your actual life. The bits a stranger could not invent because they were not there.

If a stranger could swap their name into your caption, it isn't finished. Put yourself back in, and then it is.

Don't ask a tool to be you.

What it steals if you let it

It steals your nouns. Specific objects get replaced with "journey", "energy", "practice", "alignment". It steals your humour, because humour is risk and models prefer safe. It steals your opinion, because opinions create heat. It steals your rhythm — the way Bee swears, the way Jonah goes short, the way Priya goes dry. It steals your contradictions. You are allowed to be spiritual and fed up. The machine will try to iron that into a lesson.

That is why people scroll past. Not because the caption is "bad". Because nobody they know is in it.

A terrible AI caption, fully rewritten

Here is the kind of thing I get sent when someone says "I asked ChatGPT to write my post about the journal."

Terrible original:

"Are you ready to take your spiritual practice to the next level? This luxury moon journal is designed to help goddesses like you manifest your dream life, release what no longer serves, and step into your power one page at a time. Packed with high-vibe prompts, it is the ultimate self-care tool. Comment JOURNAL and I will send you all the details. Remember, what you seek is seeking you."

What is wrong, specifically, not vibes:

• It addresses "goddesses like you", which is second-hand costume.
• "Next level", "dream life", "what no longer serves", "step into your power", "high-vibe", "ultimate self-care", "what you seek is seeking you" — a graveyard of phrases with no owner.
• It claims luxury and prompts the journal may not even be about.
• The ask is a lead magnet, not a person offering a thing.
• There is no Tuesday in it. No body. No cost. No humour. No "this is not for you if".

Now we put a human back in. I will do this as Elise, because she is the one most tempted to let a machine sound confident for her.

Step 1. Facts from her life. She writes at night after the baby sleeps, sometimes. She is scared of morning-pages culture. She bought the Moon Journal because the blank posh notebook made her feel like a failure. She still skips days. She will not claim manifestation.

Step 2. Opinion. Prompts that boss her around would make her quit. She likes that this one does not shout.

Step 3. Language she actually uses. "I cannot do 5am." "If it has a streak tracker I will lie to it." "Please do not call me a goddess, I had Weetabix in my hair."

Step 4. Product as furniture. It lives on the sofa arm, not on an altar.

Step 5. Ask in a way they can say no. Don't shove.

Rewritten:

"I asked a machine to write this caption and it called you a goddess and promised you a dream life. I cannot in good conscience. I use the Moon Journal on the sofa after the baby is down, if I am honest enough to pick up a pen instead of the rest of the biscuits. It does not make me consistent. It does make me slightly less vicious in my own head. If you want a notebook that will not shout at you for skipping Tuesday, I will send this one. If you want a manifestation system, I am the wrong woman, and I am still glad you are here."

That is a rewrite. The machine could have helped her list sofa / baby / biscuits / skipped Tuesday if she had fed it those words. It could not have meant them.

Do a second pass if you need to, out loud. Elise's first human draft still had "holding space" in it because the phrase had got into her mouth from two years of scripts. She cut it. The sentence became "I sat with it" which is what she actually did. That is the work: not sounding clever. Sounding like the person who was there.

How to add yourself every time

A checklist, not a ritual:

1. 1Put in one object that exists. Pans. A receipt. Locker tin. Lasagne.
2. 2Put in one feeling you would usually tidy. Guilt. Fury. Horniness for sleep. Embarrassment.
3. 3Put in one opinion. Not a quote. A stance.
4. 4Put in one joke or one dry line if that's your mouth. If it isn't, don't fake Bee.
5. 5Put in a limitation. Who it isn't for. What it won't fix.
6. 6Read it out loud. If you wouldn't say it to a neighbour, cut it.
7. 7Search for words you never say. Delete them.

Tasha: she used AI for everything

Tasha is bright. She was drowning in a 9-to-5 and a toddler. AI felt like relief. Then her grid became a wellness brochure. People who knew her in real life said "this doesn't sound like you". Strangers didn't speak at all. She was posting daily and vanishing.

Recovery, in order, not all in one weekend:

• Stop generating full captions for two weeks. Allowed: spellcheck, structure, questions.
• Speak five stories as voice notes. Transcribe. That's your language.
• Steal from yourself: old texts to friends are closer to your voice than any model.
• Rewrite one old AI post in the workbook until someone who loves you would say "there you are".
• Tell the audience, if you want: "I've been letting a robot talk and it made me boring. I'm back. Here's the washing up." You don't have to confess. You do have to change.
• Measure: could I have written this on a bus with no signal? If no, it isn't ready.

She doesn't need to be anti-tool. She needs to be the one writing.

A week later Tasha posted a photo of toddler yoghurt on a work blouse and four sentences about using the journal in a car park because the house was too loud. Eight people she actually likes replied. That's not going viral. That's people recognising her, which is the only kind of engagement that turns into a conversation — and the only kind of conversation that can hold a product later.

Don't ask a tool to be you. Next: comparison, which will try to nick your voice in a different way.