Microsoft Advertising has published dedicated guidance on AI-generated, AI-manipulated and other synthetic content in ads. The message: if AI helped create or materially alter an ad, consumers may need to know. Advertisers are responsible for complying with laws wherever campaigns run, including disclosure requirements, permissions and rules on using a person’s likeness or voice. AI-generated ad disclosure is now a baseline check for ad operations.
Advertisers should keep watermarks, metadata and other provenance information that identify how AI content was created. Interfering with machine-readable provenance can get ads rejected.
When disclosure is required, advertisers must make it clear and place it near the relevant content. Microsoft recommends embedding AI disclosures directly into image and video assets; its existing ad disclaimer feature can also be used with supported formats.
Slapping “AI-generated” on deceptive creative won’t make it acceptable. Ads can be rejected, restricted or removed if they contain prohibited deepfakes, impersonate people or organizations, use someone’s likeness or voice without authorization, omit required disclosures or interfere with provenance data.
Images, audio and video made with Microsoft AI tools may carry machine-readable provenance data, metadata and imperceptible watermarks. Because consumers can’t see those signals, advertisers may still need to add visible or audible disclosures.
Microsoft isn’t banning AI advertising; it’s drawing a clearer line between using AI to make an ad and using AI to mislead the viewer. The industry is moving the same way: the IAB’s revised AI disclosure rules focus labels on cases that matter to consumers, and YouTube reworked its AI likeness claims process. If you run campaigns across platforms, keep one document tracking each platform’s rules.
When AI created or materially altered an ad and local law requires it, advertisers must disclose clearly and near the content, ideally embedded in image and video assets.
Prohibited deepfakes, impersonation, unauthorized use of likeness or voice, missing required disclosures, or interfering with machine-readable provenance data.
No. Those signals aren’t visible to consumers, so advertisers may need to add visible or audible disclosures.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
Reddit Ads Manager is adding a campaign planner that forecasts reach and cost, a 0–100% optimization score and automatic budget shifting. Advertisers scoring 80%+ saw 19.6% lower blended CPA.
Netflix lowered its 2026 ad revenue forecast from $3B to about $2.7B even as ads grow 80%+. Here’s why it missed, plus what the new Netflix Pixel and The Trade Desk audiences mean for advertisers.
Meta’s new Terms of Service, effective Oct. 30, appeared to let it crawl businesses’ domains. After backlash, Meta clarified the scope. Here’s what changed and what advertisers should check.