- AI writes from what it knows. Give it a brand brain: your offer, audience, tone, facts and words you never use.
- Separate facts that must be exact (prices, hours, services) from general advice the AI can write freely.
- Never let AI invent testimonials, results, statistics or customer stories. Fake reviews are illegal in the US under an FTC rule.
- Review every post before it goes live with a short checklist; approval is the most effective quality control.
- Feed your edits back into the brand brain so the same mistake does not repeat.
Why AI posts go generic or wrong
Language models are good at producing text that sounds right. Without specific information about your business, they fill gaps with the average of everything they have read. That leads to three typical problems:
- Generic voice. "Elevate your self-care journey" could be any salon in any city.
- Invented details. A price, opening hour, award or "since 2010" that nobody gave the AI.
- Repetition. The same hook, structure or topic every week, because the tool does not remember what it already wrote.
Each problem has a fix, and none of them requires you to write the posts yourself.
Build a brand brain
A brand brain is the information the AI should use every time. It can live in a document, in your tool's settings, or both. Include:
- What you sell. Services or products, with the names you use and what is included.
- Who it is for. Your typical customers and what they worry about.
- Where and when. Location, service area, opening hours, how to book.
- Proof you can use. Real reviews you have permission to share, real certifications, real numbers.
- Tone. Three adjectives ("warm, direct, a bit funny") and two example posts you love.
- Do not say. Words, claims and topics that never appear, such as "cheapest", "guaranteed results", competitor names.
- Visual style. Colors, fonts, logo, and the kind of photos you use.
Your website is usually the best starting source. Our guide on turning your website into social media posts shows which pages hold the most useful facts.
Set fact rules
Split content into two kinds and treat them differently:
| Type | Examples | Rule |
|---|---|---|
| Brand facts | Prices, hours, services, staff, offers, awards, years in business | Only from your brand brain. If it is not there, it does not go in the post. |
| General knowledge | Hair care tips, how a cleaning works, why stretching helps | The AI may write it, but you check it fits your practice. |
| Numbers and statistics | "80% of people..." | Only with a real source you can name. Otherwise, remove it. |
| Time-bound details | Dates, deadlines, seasonal offers | Check they are still true on the publishing date. |
Write these rules into your instructions for the AI, and use them as your review lens.
What AI must never invent
- Testimonials and reviews. In the US, the FTC's final rule on fake reviews and testimonials (announced August 14, 2024) bans businesses from creating or buying fake reviews, including AI-generated ones that misrepresent who wrote them. Other countries have similar consumer protection rules.
- Customer stories. "Sarah came to us with..." is only fine if Sarah exists and agreed.
- Results and guarantees. Especially for health, beauty, fitness, finance and real estate.
- Statistics. Made-up numbers damage trust when someone checks.
- Before and after images. Never generate or alter them to show results that did not happen.
- Partnerships and awards you do not have.
If a draft contains any of these and you cannot confirm it, delete the sentence. A slightly less exciting true post always beats an impressive false one.
A 60-second review checklist
- Facts: are every price, time, name and offer correct?
- Date: will it still be true on the day it is published?
- Voice: would you say this to a customer at your counter?
- Claims: any promise, number or result you cannot back up?
- Visual: is the image yours or clearly illustrative? Does text on the image match the caption?
- Call to action: one clear step that works (link, keyword, booking)?
- Hashtags: up to 5 relevant ones on Instagram.
If two or more answers are "no", skip the draft rather than fixing everything by hand. Approval workflows are covered in social media on autopilot.
Before and after: fixing a generic AI draft
A typical draft from an AI with no brand context, for a physiotherapy practice:
"Unlock your body's full potential! Our expert team has helped thousands of happy clients live pain-free. Book now and transform your life today!"
What is wrong: "thousands of happy clients" is an invented number, "pain-free" is a health promise, and the voice could belong to anyone. The same post written from a brand brain:
"Sitting all day and your lower back aches by 4 pm? Try this: stand up every 45 minutes and do 5 slow hip circles each way. If the pain lasts longer than a week, book an assessment with Anna or Mark at our Elm Street practice. Link in bio."
It is specific, useful, uses real names and a real address from the brand brain, and makes no promise the practice cannot keep. That is the standard to aim for.
Extra care in regulated industries
Health, dental, aesthetics, finance, legal and real estate businesses face stricter rules on advertising claims, before and after images and testimonials, which vary by country and profession. Ask your professional body or adviser what applies to you, write those rules into your brand brain as "never say" items, and make sure the person who approves posts knows them. Our playbooks for real estate agents and coaches include examples of careful wording.
AI images and video: extra care
- Do not let AI images pretend to be your work. Use them for ideas, backgrounds or illustrations. Use real photos for results, your space, your team and your products.
- Label where appropriate. Meta explains on its AI content labeling page that it adds an "AI info" label when it detects industry standard AI indicators or when people disclose that they are uploading AI-generated content. Disclose realistic AI media honestly.
- Check text in images. AI images can contain misspelled or invented words. Zoom in before approving.
- Voice and presenters. If an AI presenter speaks for you, the script still needs the same fact check as a caption.
Make it better every week
- Turn edits into rules. If you change "clients" to "guests" twice, add it to the brand brain.
- Track topics. Keep a list of what was posted so new drafts do not repeat the same angle.
- Answer open questions. When the AI lacks a fact, add it once instead of correcting each post.
- Save your best posts as examples. Good examples teach tone faster than adjectives.
How Moonya handles it
Moonya's Brand Brain reads several pages of your website and learns your offer, voice, audience and brand colors. You can review and edit what it learned, add documents, and answer the questions it asks when a fact is missing. Posts are written from that Brain in your language and designed in your brand colors and fonts.
Every draft then waits for your approval: approve, edit or skip. Nothing is published without you. Instagram captions use up to 5 hashtags, and each network gets a fitted caption. Plans start at $29/month after a 7-day trial for $1. See pricing, compare Moonya vs Ocoya, or read the playbooks for med spas and dental clinics, where accuracy matters most.
Keep reading
See the best AI social media tools for small businesses and social media manager vs AI tool.
FAQ
Is it okay to use AI to write social media posts for my business?
Yes. Many businesses do. Give the AI accurate information about your business, and have a person review every post before it goes live.
How do I make AI posts sound like my brand?
Give the AI your offer, audience, tone in a few words, example posts you like and a list of words you never use. Then turn your edits into rules.
Can AI write customer reviews or testimonials for my posts?
No. In the US, the FTC's rule on fake reviews bans creating fake reviews and testimonials, including AI-generated ones. Only share real reviews with permission.
Do I have to label AI-generated images on Instagram?
Meta adds an "AI info" label when it detects AI indicators or when you disclose AI content. Disclose realistic AI images and video honestly, and never use AI images to fake results.
How do I stop AI from inventing prices or opening hours?
Keep exact facts in one source, such as your brand brain or website, and tell the AI to use only those. Check every price and time during review.
How long does reviewing AI posts take?
With a short checklist, about a minute per post. Skip weak drafts instead of rewriting them.
Sources
- U.S. Federal Trade Commission: final rule banning fake reviews and testimonials, including AI-generated (August 14, 2024) (checked October 29, 2026)
- Meta Transparency Center: labeling AI content ("AI info" label, detection and disclosure) (checked October 29, 2026)