Microsoft Advertising Introduces New Rules for AI-Generated Ads

Microsoft Advertising Introduces Rules for AI-Generated Ads

On September 14, 2026, Microsoft Advertising published dedicated guidance for AI-generated and synthetic content in advertising. The policy addresses images, videos, audio, and other materials produced or materially altered using artificial intelligence.

Microsoft is not banning AI-created advertising. Its guidance focuses on transparency about content origins, preventing misleading impressions, and protecting the rights of people whose likenesses or voices appear in ads.

Why Microsoft Introduced the Guidelines

Generative tools have become part of everyday creative production. Microsoft itself offers AI-assisted creative capabilities through products such as Copilot and Ad Studio, while AI-powered campaign features can adapt ad content.

Advertisers also use third-party image generators, synthetic voices, digital characters, and AI video tools. Previously, they largely relied on general policies covering truthful advertising, intellectual property, and deceptive content.

Microsoft has now spelled out additional responsibilities for AI-assisted creative, including preserving information about how material was generated.

Key Requirements for AI-Generated Advertising

Preserve metadata and digital watermarks. Advertisers should retain machine-readable provenance information, embedded metadata, and watermarks that identify AI generation or substantial modification. Removing required provenance signals can result in ad rejection.

Disclose AI use when required. Where applicable laws require disclosure of synthetic content, advertisers must provide a clear notice. Microsoft recommends placing the disclosure directly within the relevant image or video; supported ad formats may also allow the Disclaimer in Ads feature.

Disclosure is not universally required for every advertisement that involved AI. The obligation depends on applicable laws, the content, and the market in which an ad is shown.

Avoid deceptive or unauthorized synthetic content. AI cannot be used to impersonate people or organizations, fabricate deceptive endorsements, or create prohibited deepfakes. Using a person's likeness or voice may require authorization.

An AI-generated label does not make an otherwise deceptive ad acceptable.

How Microsoft Labels Content Created With Its AI Tools

Images, videos, and audio created with Microsoft generative tools may include machine-readable provenance information, metadata, or imperceptible digital watermarks.

These technical signals can indicate that content was generated or changed using AI, but they are not necessarily visible or audible to consumers.

If a market requires consumer-facing disclosure, advertisers remain responsible for adding it. A hidden watermark alone may not satisfy that requirement.

What Changes for Advertisers and Agencies

The biggest impact is operational: creative production workflows need an additional quality assurance step before ads are submitted.

Teams should check where an asset came from, whether permissions for people's images and voices have been secured, whether a claim or testimonial is accurate, and whether export processes preserve embedded provenance information.

AI-generated UGC, synthetic customer testimonials, AI avatars, and digitally altered videos deserve particular scrutiny. An AI avatar falsely presented as a real customer describing a fabricated product experience may violate advertising rules even if the asset carries an AI-generated disclosure.

Microsoft can reject, restrict, or remove ads that fail to meet its standards. Generative tools remain useful for scaling production, but responsibility for truthful representation, necessary permissions, and appropriate disclosures rests with the advertiser.

A Practical AI Creative Compliance Checklist

Review the creative. Confirm the people, products, locations, events, and claims depicted are accurate and not misleading.

Verify permissions. Document authorization to use a person's likeness, voice, or other protected material when required.

Retain provenance. Avoid export or editing workflows that strip embedded metadata or watermarks from AI-generated assets.

Check local disclosure rules. Determine whether visible or audible disclosure is required in each advertising market.

Review ad format support. Use an appropriate built-in disclaimer where available, without assuming it replaces legal requirements.

Document approval. Maintain an internal record of asset source, permissions, disclosures, and final creative sign-off.

What This Means for Performance Marketing

The update does not change the basic mechanics of Microsoft Advertising bidding or targeting. It does, however, add compliance considerations to creative testing and AI-powered production.

Agencies producing dozens of image or video variants should ensure that provenance signals survive editing, resizing, and delivery. Brands running AI UGC need to verify that testimonials and endorsements are truthful and properly authorized.

The broader direction is clear: determining the provenance of synthetic media is becoming part of standard creative quality assurance, alongside format requirements, brand safety, and advertising policy review.