Microsoft Advertising draws the line on AI ads: creating is fine, misleading is not
Microsoft Advertising has added a new help center policy page on AI-generated and synthetic content in ads. Reported by Search Engine Land on September 14, 2026, the page asks advertisers to follow the law in every market where an ad runs, add required disclosures, secure rights and consents, and preserve watermarks and metadata. Deepfakes stay banned even when disclosed, and ads that strip provenance data can be rejected.
A new policy page in the help center, built around transparency
Microsoft Advertising Help has a new policy page titled "AI-generated and synthetic content in advertising" in its Generative AI section, alongside its Copilot ad pages; the subtitle promises to build trust through transparency. The page carries no date. Search Engine Land reported it on September 14, 2026 under the byline of Paid Media Editor Anu Adegbola, and PPC News Feed noted the same day that Hana Kobzová had spotted it on LinkedIn. The scope covers AI-generated, AI-altered or otherwise synthetic ad content, and the principle is consumer-facing: when an ad carries AI-generated or materially altered content, the person seeing it should be able to tell. Search Engine Land's reading is direct; Microsoft is not banning AI in ads, it is drawing a clearer line between making an ad with AI and using AI to mislead the viewer. PPC News Feed frames the guidance as a continuation of Microsoft's 2024 policies on synthetic content in political ads, synthetic content disclosure and deepfake sexual content in ads.
Four advertiser duties and Microsoft's invisible watermark
The page gives advertisers four responsibilities: comply with the laws and regulations of every market where the ad is shown; obtain the necessary rights, permissions and consents; provide disclosure where required; and preserve the watermarks, metadata and other provenance information. Deepfakes stay prohibited even with a disclosure, and every Microsoft Advertising policy applies, including the Unacceptable content policy. When Microsoft's own AI tools generate an image, audio or video, it applies machine-readable provenance information, metadata and an imperceptible watermark. That may not appear as a visible label, so adding the visible or audible disclosure the law requires remains the advertiser's job. Disclosures should be clear, easy to understand and close to the relevant content; Microsoft recommends embedding them directly in image and video assets, and the Disclaimer in ads feature is available in supported formats. A disclosure does not make misleading content acceptable: an ad can be rejected, restricted or removed for banned deepfake content, misleading consumers with AI, impersonating a person or organization, omitting a required disclosure, using someone's likeness or voice without authorization, or removing or tampering with the machine-readable information.
Not a ban but a line: technical layer on the platform, legal burden on the advertiser
What matters here is not a new prohibition but where responsibility lands. Microsoft applies provenance data automatically in its own tools yet leaves visible disclosure and local law compliance to the advertiser. A pre-submission review is expected as well; advertisers should check AI-assisted creative and confirm that the people, places, products, claims and events shown are represented accurately. The closing checklist comes down to three items: secure the required rights and consents, add the required disclosure, and keep existing watermarks, metadata and provenance information intact. Search Engine Land adds two practical notes: check disclosure requirements in every market where a campaign will run, and do not strip metadata or watermarks from AI assets. The sources also leave gaps: no separate effective date (the page is already live and reads as guidance), and no country list, penalty figures, named Microsoft spokesperson or name for the provenance standard. Microsoft says only machine-readable provenance information, metadata and an imperceptible watermark, and beyond that we do not know.
For businesses in Türkiye: not named in the sources, tied to local law
First the boundary: none of the three sources mentions Türkiye or any specific country. Microsoft's language is global, and by requiring compliance with the law wherever an ad is shown, the rule hands the burden to local legislation. Which disclosure obligation applies in Türkiye is not this story's subject; advertisers should confirm it with legal counsel, as Microsoft itself advises. The practical side is clear. For businesses and agencies in Türkiye running AI-generated images, video or audio through Microsoft Advertising, meaning the Bing search network, Microsoft Audience Ads or ads inside Copilot, the order is this. First, check that metadata and watermarks survive the production and export chain. Second, embed the required disclosure inside the image or video, not in the description text. Third, if a real person's likeness or voice is used, obtain written consent and keep it on file. Fourth, review the creative before submission against the real product, claim and event. These four steps reduce the risk of rejection whichever disclosure a country requires.
The UNALSOFT view
Our reading: an AI-generated ad is now judged not only by the quality of the creative but by the information it carries. Microsoft adds the invisible provenance data itself, asks the advertiser for the visible disclosure, and can reject the ad when either is missing. That is why, in our AI ad film production, we check three things in every delivered asset: the tools' metadata and watermark survive in the final file, the disclosure label sits inside the video and stays readable, and consent is documented for any real person or voice used. Microsoft itself leaves open how the requirements will vary by market and content type; treating its list as a delivery criterion today costs little.
Sources
Microsoft Advertising Help, official policy page · Search Engine Land, news report · PPC News Feed, news brief
Are your AI-generated ads ready for platform rules?
Let's review the disclosure, consent and provenance status of your current creatives together. A short conversation is enough to start.