AIVX Labs
AI news for agencies · Ads in the AI era

Writing brand rules for AI-written ads: Microsoft's 40 messaging guidelines

By Alan Magliocca · Published November 2026

Quick answer: On October 28, 2026, Microsoft Advertising opened a pilot that lets advertisers set up to 40 messaging guidelines for AI-generated ad copy in Performance Max campaigns and Search campaigns with AI Max. Write each client a short, tested rule set covering claims, prices, tone and banned words, and keep reviewing the copy Microsoft generates, because guidelines steer the AI but don't replace your review.

Key takeaways

  • Microsoft's messaging guidelines are in an open pilot announced on October 28, 2026.
  • They apply to Performance Max and to Search campaigns with AI Max switched on.
  • You can add up to 40 guidelines, set in the campaign creation flow.
  • Microsoft suggests using them mainly for restrictions on claims, themes, tone and language, plus some positive guidance.
  • Term exclusions, added with AI Max in August, block specific words; guidelines handle broader rules.

What happened

On October 28, 2026, Microsoft Advertising announced in its October product update an open pilot for messaging guidelines on AI-generated ad copy.

What Microsoft says about them:

  • Where they apply: Performance Max campaigns and Search campaigns with AI Max enabled.
  • How many: up to 40 guidelines.
  • What they're for: mainly restrictions, meaning broad rules about claims, themes, concepts, tone or language to avoid. You can also add positive guidance on preferred wording, themes or tone of voice.
  • Example: a high-end retailer could add "Never imply that products are low cost."
  • Where to set them: in the campaign creation flow.

The post doesn't say how to join the pilot. Ask your Microsoft account team or Microsoft Advertising support.

Guidelines sit on top of the controls Microsoft added when it made AI Max available to everyone in August. Those include brand inclusions and exclusions, and term exclusions that block specific words or phrases, such as "discount," "buy now" or "cure," from AI-written copy. At the time, Microsoft listed prompt-based messaging guidelines as coming soon. This pilot is that feature.

What it means for agency owners

When an ad platform writes copy for a client, the risk is a claim the client can't stand behind: a price they don't offer, a promise they can't keep, a word their regulator dislikes. Until now, your options in Microsoft were to block single words or to switch the AI off.

Guidelines let you write rules in plain sentences instead. "Don't promise same-day service" covers dozens of phrasings that a word list would miss.

This is also a new piece of work you can sell. A written brand rule set for AI-generated ads is a short, concrete deliverable, and the same rules carry over to Google's AI Max text guidelines and to any AI tool your team uses. It also protects you. Overstating what a service can do is an enforcement risk, as the FTC's case covered in how not to oversell an AI service shows.

What to do this week

  1. List the clients running Microsoft PMax or AI Max and ask your account team about joining the pilot.
  2. Collect each client's no-go list: claims they can't make, offers they don't run, competitor names, words their industry rules restrict.
  3. Write the rules using the template below. Aim for 10 to 20 clear rules before you go anywhere near 40.
  4. Add term exclusions for single words that must never appear, and keep guidelines for broader rules.
  5. Review the generated copy in the campaign after it starts serving. Microsoft's help center has a page on reviewing generated copy, and you should read the first batch line by line.
  6. Get the client to sign off on the rule set and keep it on file.

How AIVX Labs helps

AIVX Labs doesn't connect to Microsoft Advertising, so guidelines are entered in each account. It can help with the copy those rules govern. The AI Ad Creator writes 3 ad versions for Google, Facebook, Instagram or LinkedIn, the Ad Hook & Angle Generator gives you 20+ angles, and the Ad Variant / Split-Test Generator builds the variations to compare. Approved, human-checked assets give the platform's AI better material to start from. The Google ad set for a roofer demo shows RICO producing a full set from one request, with the credit cost shown before you click Go.

What goes in a client's rule set?

Group the rules so they're easy to check. A workable template:

  • Claims: what the client can and can't promise. "Don't claim results within a specific time." "Don't say 'guaranteed' unless the offer includes a written guarantee."
  • Prices and offers: "Don't mention prices." Or, "Only mention the free estimate; no other discounts."
  • Comparisons: "Don't name competitors." "Don't call the business the best or number one."
  • Regulated language: rules from the client's industry. A health client may need "Don't describe treatments as cures," and a finance client "Don't promise approval."
  • Tone: "Plain and friendly. No slang. No exclamation marks."
  • Positive guidance: the phrases the client wants, such as "Mention the 24-hour emergency line" or "Refer to the team as licensed technicians."

Write each rule as one sentence with one idea. Long rules with several conditions are harder to check against the ads that come out.

A worked example: a dental practice

Here is a 12-rule set for a family dental practice running Microsoft PMax:

  1. Don't promise pain-free treatment.
  2. Don't mention prices or payment plans.
  3. Don't describe any treatment as permanent.
  4. Don't name or compare other practices.
  5. Don't call the practice the best, top or number one.
  6. Don't imply emergency appointments are always available the same day.
  7. Don't use before-and-after language.
  8. Don't mention insurance brands.
  9. Keep the tone calm and reassuring.
  10. Mention that new patients are welcome.
  11. Refer to the location as the town name, not the region.
  12. Use "appointment," not "consultation."

Pair them with term exclusions for single words the practice never wants, such as "cheap" or "cure." Then read the first round of generated ads. If one breaks a rule, rewrite the rule more plainly and note the change.

If the same client uses Google Ads, set the matching rules in Google's AI Max text guidelines; see Microsoft AI Max is on by default for how settings move between the two. Meta has had its own debate over who answers for AI-altered ads, covered in why an Advantage+ audit still matters.

The same discipline helps wherever AI writes sales copy, including promotions for live events. BrightStage's piece on time-limited offers without fake urgency applies the idea to webinar offers.

Sources

Written by Alan Magliocca, founder of AIVX Labs and BrightStage AI (Podavinci LLC). Written with AI assistance and checked against the linked sources.

Read our featured article on LinkedIn

Get AIVX Labs verified for $1: RICO, your AI operator, with 7 days of full access.

Get AIVX Labs Verified for $1

Follow along:LinkedInYouTubeTikTokXFacebookInstagramSkool community

Frequently asked questions

Do messaging guidelines guarantee the AI never breaks a rule?

Microsoft's post doesn't promise that, and it points advertisers to a help article on reviewing generated copy. Keep reading the generated ads, especially in the first weeks.

What's the difference between term exclusions and messaging guidelines?

Term exclusions block specific words or phrases from AI-written copy. Messaging guidelines are plain-sentence rules about claims, themes, tone or language, so they can cover ideas a word list would miss.

Which campaigns can use the guidelines?

Microsoft says they apply to Performance Max campaigns and to Search campaigns with AI Max enabled, and you set them during campaign creation.