AI transparency
Last updated: October 9, 2026
What we use AI for
| What | For | Provider |
|---|---|---|
| Writing | Captions, hashtags, ideas, ad texts and video scripts. | OpenAI, Anthropic |
| Images | Ad backgrounds and scenes, and presenter portraits. Your text, logo and colours are placed on top without AI. | OpenAI |
| Video | Talking presenters, lip sync and moving product shots. | fal.ai (with the models shown before you start) |
| Voice | Voice-overs and, if you choose, a clone of your own voice. | OpenAI, ElevenLabs |
| Analysis | Reading a website into a brand and product card, summarising competitors' public ads, and insights about your posts. | OpenAI, Anthropic |
AI only runs when you ask for it; in the Ad Studio you see the cost in credits before it starts. SocialLoopy does not train AI models on your content. We send your input to the providers above through their business APIs, under terms that limit how they may use it.
How AI-made content is labelled
Under article 50 of the EU AI Act, people must be able to recognise realistic AI-made images, video and audio. That is why SocialLoopy:
- writes a machine-readable mark into every AI-made ad image (IPTC "digital source type" in the XMP metadata) and ad video from the Ad Studio. Meta, TikTok and other platforms can read these marks;
- puts a visible "AI-gegenereerd" label on AI-made ads by default. You can only switch it off after confirming that your use still meets the disclosure rules;
- switches on TikTok's own "AI-generated" label when you post an AI video made here to TikTok, and lets you switch it on for other AI videos.
Platforms may add their own labels, such as Meta's "AI info".
AI presenters are not real customers
The presenters in AI videos are made with AI or come from a photo you uploaded. They present your product; they never pretend to be a customer. Scripts are written that way, and a check warns you when a script says "I've been using this for weeks" or something similar.
You can only use a photo of a real person after confirming that this person agreed to it. That consent is stored with the photo, and deleting the avatar deletes the photo too.
What AI does not do here
- No invented proof. Reviews, ratings and figures are only used when they appear on your own website and you have confirmed them.
- No AI decisions about your budget. Ad advice comes from fixed rules about your results (for example: this ad costs twice as much per result as the others). The explanation you read is written out from those numbers. Nothing is changed without your confirmation, unless you switch on autopilot, which is limited to small budget steps within your cap and pausing clearly weak ads.
- No copying of competitors. When you write ads inspired by a competitor, only the idea goes to the AI, never their text, and results that are too similar to their ads are dropped.
- No decisions about people. SocialLoopy does not use AI to make decisions about individuals.
Your part
AI output can be wrong. Read texts, check images and watch videos before you publish them, especially product claims, prices and anything about health or money. You stay responsible for what you publish; see our Terms of Service. How we handle the data you give the AI is described in our Privacy Policy. Questions or a concern about AI content? Use our contact form.