Google Ads Text Guidelines: Put Guardrails Around AI-Generated Copy

Google Ads API v23.1 introduced campaign-level text guidelines for Performance Max and Search. The fields include term exclusions and messaging restrictions, giving teams a programmatic way to steer AI-generated text instead of relying only on a brief stored elsewhere.

What text guidelines should contain

A useful rule is specific enough to test. “Stay on brand” is not testable. “Do not use the words free, guaranteed, or instant” is. Separate hard prohibitions from preferred language so reviewers know which failures block launch.

Control	Use it for	Do not use it for
Term exclusions	Prohibited claims, competitor names, retired products, sensitive phrases	Long explanations of brand voice
Messaging restrictions	Required framing, audience limits, offer conditions	Replacing legal review
Human approval	Context, implied claims, tone, landing-page match	Checking only spelling

Build a three-layer guardrail system

1. Block risky language

Start with terms that create legal, policy, or trust risk. Include absolute claims, unsupported superlatives, medical or financial implications, and words that conflict with current offer terms. Keep the list short enough to maintain.

2. Define acceptable message boundaries

State what the campaign may promise, which audience the promise applies to, and what qualification must remain visible. If a discount is conditional, the restriction should preserve that condition. If the product is not suitable for every customer, do not let the generated copy imply otherwise.

3. Verify the served experience

Generated text can be individually compliant yet misleading beside a particular image, price, or landing page. Review combinations, not isolated assets. Compare the final promise with the destination page and the conversion event that bidding uses.

A practical QA workflow

1. Export approved and prohibited phrases from the current brand and legal guidance.
2. Convert only enforceable rules into term exclusions or restrictions.
3. Create a test campaign or controlled change window.
4. Review previews and served combinations across devices.
5. Log violations, false positives, and wording that weakens performance.
6. Update the rules through change control rather than ad hoc edits.

This complements MTC’s AI Brief guidance: the brief gives the system direction, while text guidelines define boundaries.

How to maintain text guidelines as the account changes

Treat the guideline set as a living control document, not a one-time setup. Assign one owner, record the reason for every restriction, and review the list whenever offers, legal language, product names, or landing-page promises change. A useful monthly check compares generated assets and search terms against the approved messaging matrix. If the same undesirable phrase appears more than once, decide whether to add a restriction, revise the source assets, or narrow the campaign’s URL scope.

Keep an exception path, too. A phrase that is unsafe for one market may be required in another. Document campaign-level exceptions and their expiration dates so temporary rules do not become permanent blind spots. The goal is controlled flexibility: enough room for Google’s systems to assemble relevant copy, with clear boundaries around claims, tone, and regulated language.

Frequently asked questions

Do text guidelines guarantee compliant ads?

No. They constrain generation but cannot interpret every claim, context, or jurisdiction. Human review remains necessary.

Should every disliked word be excluded?

No. Excessive exclusions can make copy generic or reduce useful variation. Reserve hard blocks for real risk.

Sources

• Google Ads API release notes (v23.1 text guidelines)
• Google Ads AI Max guidance

Free weekly newsletter

Get one useful marketing lesson each week.

Practical guidance for making better marketing decisions without the hype.