Microsoft tests a way to tell its AI what your ads should never say

Microsoft is testing a way to tell its AI which claims and phrases to avoid when writing your ads. The messaging guidelines feature is in open pilot for Performance Max campaigns and Search campaigns with AI Max enabled.

In its October 28 update, Microsoft says advertisers can supply up to 40 guidelines covering tone, themes, concepts and language.

Its example for a premium retailer is simple: “Never imply that products are low cost.” That gives the system a broader instruction than banning one word, such as “cheap.” An ad can suggest a bargain without using that word.

You can also tell the generator what to include. Microsoft’s screenshot shows separate instructions for descriptive language, headline capitalization and how to write the brand name.

You still need to check what the AI writes

Microsoft describes the feature as a way to reduce the risk of unwanted wording. The announcement does not promise that every generated ad will obey every rule.

That’s consistent with Microsoft’s existing position that advertisers remain responsible for AI-generated ads, including content produced with its own tools.

For a campaign team, a useful starting point is the language you already reject during reviews: unsupported claims, promises your service cannot meet and descriptions that misrepresent the product. Turn those recurring corrections into guidelines, then check the resulting ads against them.

The pilot gives you somewhere to put those instructions before generation. It leaves the final judgment about whether the ad is accurate with you.

Source: Microsoft