AI SMM team
More and more brand posts include a generated image, an AI voice-over or a text a model helped to write. At the same time, platforms and regulators are asking for clearer disclosure of AI-made material, and audiences are getting better at noticing it. If you are not sure when to label, how, and where to look for the rules that apply to you, this guide gives a practical framework. It is general guidance, not legal advice: the rules differ by country and platform and change often, so always verify them on each platform’s official page and, if the stakes are high, with a qualified lawyer.
- Rules about labeling AI content differ by country and by platform, and they change; check each platform’s official help pages before relying on any summary, including this one.
- The safest practice is to label content that could be mistaken for a real photo, video, voice or event, and anything involving real people.
- Using AI as a writing or editing helper is treated differently from presenting a generated “photo” as a real one; platforms focus mostly on realistic media.
- Keep a simple internal policy: what you generate, who approves it, how it is labeled and where the original files are stored.
1. Why labeling matters for a brand
There are three practical reasons to think about labeling before the first generated image goes live:
- Trust. People forgive the use of AI far more easily than a feeling of being misled. A “photo” of a dish you do not actually serve, or a “customer review” video made by a model, can cost more trust than any saved hour is worth.
- Platform rules. Several major platforms ask creators to disclose realistic AI-generated or altered media, may add labels automatically and may reduce the reach or remove content that is misleading.
- Legal exposure. Depending on the country, advertising, consumer protection and personal data rules may apply to AI-made content, in particular when it depicts real people, makes claims about a product or imitates a testimonial.
2. What counts as “AI content”
It helps to split AI use into levels, because the expectations differ:
| Level | Examples | Typical expectation |
|---|---|---|
| Assistance | Spelling fixes, headline options, shortening a text, translation draft, brainstorming | Usually no label; you remain the author and check the result |
| Editing | Background removal, retouching, resizing, subtitles, noise reduction | Usually no label, unless the edit changes the meaning of what is shown |
| Generated illustration | Obviously stylised pictures, abstract backgrounds, icons, cartoon characters | Label is optional or recommended; the viewer is not misled |
| Realistic generated or altered media | A photorealistic image or video of a place, event or product; synthetic voice; a lookalike of a real person | Disclosure is expected or required on many platforms; label clearly |
| Synthetic people and testimonials | A generated “customer” praising the product, a fake expert, an invented before/after | Avoid; if used at all, label prominently and check advertising rules |
The pattern across platforms is similar: the closer a piece of content is to a believable record of reality, the stronger the case for labeling it.
3. How platforms approach it
Platforms describe their requirements in help centres and community guidelines, and update them regularly. In general terms you can expect:
- Self-disclosure tools. Some networks provide a toggle or a field when you upload to mark content as AI-generated or altered. Use it when your material is realistic.
- Automatic labels. Some platforms detect signals in files (metadata or watermarks of generation tools) and add a label themselves. Do not try to strip metadata to avoid it; that tends to look worse if discovered.
- Special rules for sensitive topics. Elections, health, finance and news-like content often have stricter requirements.
- Advertising has its own policy. Paid promotion is usually reviewed under separate rules that may require disclosure for realistic synthetic media.
Check the current wording on the official pages of the networks you use, for example the help or policy sections of Instagram, Facebook, TikTok and YouTube. Telegram channels are less formalised, but the laws of your country and basic honesty still apply to what you publish there. Do not rely on blog posts, forums or screenshots for the exact rule: platforms change them.
4. Country differences
Some countries and regions have introduced or are preparing requirements for labeling synthetic media, deepfakes or AI-generated advertising; others have none specific to this. Rules may apply based on where the publisher is, where the audience is, or both. For a small business this means:
- find out whether your country has specific rules for AI-generated content in advertising and media;
- if you sell or advertise in several countries, check the strictest of them;
- keep records of what was generated and when, in case you need to show it;
- when in doubt about a campaign with real people, medical, financial or political claims, ask a lawyer who knows local rules.
5. When to label: a simple decision list
Before you publish, ask yourself:
- Could a normal viewer take this for a real photo, video, recording or event? If yes, label it.
- Does it show a real person, or look like one, saying or doing something they did not? Do not publish without the person’s consent; label it in any case.
- Is it a product shot or a result of your service? Do not show an AI image as the actual product or work result. It can be a mood image, not evidence.
- Is it a review, testimonial or case study? It must come from a real client; never generate one.
- Is it obviously stylised art or a decorative background? A label is usually not needed, but a short note does no harm.
- Is it a sensitive topic (health, money, politics, children)? Be stricter: label and check the platform’s rules and the law.
6. How to label so that it actually helps
- Use the platform’s own toggle where it exists. It is the form the network understands.
- Add a short, plain note in the caption: “Illustration created with AI”, “Image generated with AI, not a photo of the product”. Put it near the start or the end, not buried in a pile of hashtags.
- For video, mention it in the caption and, if the video is realistic, consider an on-screen note at the start.
- For voice, say that the voice is synthetic if it could be mistaken for a real person’s.
- Be consistent. Pick one wording and use it everywhere so that regular readers learn to recognise it.
- Do not overdo it. You do not need a disclaimer on every spell-checked caption.
7. A short internal AI policy for your team
Even a one-person business benefits from a half-page policy:
- Allowed: brainstorming, drafts, headlines, translation drafts, background and layout images, subtitles, retouching.
- Allowed with a label: realistic generated images or video, synthetic voice.
- Not allowed: fake reviews, fake experts, AI images passed off as real product photos or real results, imitation of real people without consent.
- Review: a human reads every text and looks at every image before publishing, checking facts, prices, claims and details such as hands, text on images and logos.
- Records: keep the prompt or source and the final file for important posts.
For how to keep a consistent voice when AI helps with drafts, see our guide to training AI on your brand voice.
8. Common mistakes
- Presenting a generated image as a real photo of your dish, salon or project.
- Generated “before and after” images for a service where results vary.
- Using a real person’s face or voice without permission.
- Removing metadata or watermarks to avoid platform labels.
- Copying a rule from a blog post or another country and assuming it applies to you.
- Publishing without reading: a model can produce a wrong fact, a wrong price or a distorted logo.
9. How AI SMM fits in
The AI content studio helps you write drafts in your brand’s tone and create images in your brand’s style, which are then edited and approved by a person before they go anywhere. Our own advice here is simple: the studio produces drafts, and what you publish and how you label it remains your decision. The content then waits in the planner where you can read it again, add a note about generation to the caption, and schedule it for each network. When several people work on content, an approval step before publishing helps; see the social media approval workflow. For images, see AI images for social media posts. We do not give legal advice and do not guarantee that content meets every platform’s or country’s rules; verify them on the platforms’ own pages.
FAQ
Do I have to label text written with the help of AI?
In most cases platforms focus on realistic media, not on text. Some countries or industries have their own disclosure rules, so check yours. As a matter of honesty, you stay the author: read, fact-check and edit.
Will labeling reduce my reach?
Platforms say labels are meant to inform viewers, not to punish creators. Misleading content, on the other hand, can be demoted or removed. Reach depends mostly on how useful the post is.
Where do I find the official rules?
In the help centre or policy pages of each network, usually under sections about AI, synthetic or altered content, advertising and community guidelines. Search the platform’s help for “AI-generated content” and read the current text.
Is a label needed for a stylised illustration?
Often not, because nobody takes it for reality. A short caption note is still a good habit for transparency.
Write your one-page AI policy this week and check the help pages of the networks you use. When you are ready to produce drafts and visuals faster, try the AI content studio and keep the final say yourself.
Posts, chats, video and sales in one workspace. 7 days free.
Start free
