AI content policy

What the AI does

Tunsource uses AI models from Google (Gemini, through Google Cloud Vertex AI) to write ad copy, headlines, short video scripts and SEO content from the link, photo and text you give it, to make and edit images in the studio (product photos, posters, logos and restyles), and to check studio requests against the rules below. On paid plans we may also use a model from Anthropic (Claude) to write campaign copy. The ad layouts are drawn from your own photo and the copy. AI output is a starting point, not professional advice.

Check everything before you publish

AI can get facts wrong, invent features or prices, or make claims your product can't back up. Read every ad, fix anything that isn't true, and make sure it follows the rules of the platform where you post it, including any rule to label AI-generated content. You are responsible for the ads you publish.

What you can't use Tunsource for

• Illegal products or services, or anything the law where you advertise forbids.
• False or misleading claims, fake reviews, fake scarcity, or ads that pretend to be news.
• Medical, financial or legal claims you can't prove, or promises of guaranteed results.
• Pretending to be another person, brand or organisation, or using their trademarks, photos or content without permission.
• Hate, harassment, violence, sexual content, or anything that sexualises or targets children.
• Political or election ads, or content meant to mislead people about public issues.
• Scams, phishing, malware, or collecting people's data without consent.

Requests that clearly break these rules are refused, either by an automatic check before generation or by the AI itself. Your first few refused requests each day don't use a campaign or image; after that a refused request uses one, and after many refusals in a day we pause your requests until the next day. We may also remove content and close accounts that break these rules.

AI images and logos

AI images can change small details of your product, misspell words, or look like existing work by chance. Check every image before you use it, and check a logo for trademark conflicts before you register or rely on it. Images made with Google's image model carry an invisible SynthID watermark that marks them as AI-made. AI-made people are not meant to look like any real person; don't use the studio to imitate anyone.

Report AI output

If an ad, image, script or SEO text is harmful, offensive, untrue or infringes someone's rights, use Report on the campaign page or studio result (you need to be signed in), or email support@tunsource.com. We read reports and act on them, which can include removing content, blocking requests or changing how we generate ads.

Your data and AI training

We send your inputs to Google (and, on paid plans where it is used, to Anthropic) only to generate and check your content. We don't use your content to train AI models. Under their commercial terms, Google does not train its models on data sent through Vertex AI, and Anthropic does not train its models on content sent through its API.