Decision brief · SEM
Treat every text-guideline edit as a campaign change
The beta adds useful message controls, but their scope and blast radius are narrower and sharper than a normal brand guide.
- The filter covers Google-generated copyIt governs text assets made through text customization, not every headline, landing page or targeting choice in the campaign.
- The two rule types behave differentlyTerm exclusions block exact words or phrases by language; messaging restrictions govern concepts, associations and tone across languages.
- A new rule can remove live assetsExisting Google-optimized text that fails a guideline can stop serving and be marked as removed.
- Copied rules are snapshots, not a shared policyEach target campaign receives an independent copy, so governance requires versioning and a post-change asset review.
Google Ads text guidelines look like a place to paste a brand voice document. They are more operational than that. The beta is a campaign-level filter for text created by Google when text customization is enabled in Performance Max or AI Max for Search.
The useful question is therefore not whether the feature understands a brand. It is which generated assets a rule can suppress, what the rule cannot control, and how to change it without quietly shrinking the campaign’s usable copy pool.
The campaign contains two different copy systems
Text customization combines advertiser-provided responsive search ad assets with additional Google-generated headlines and descriptions. Google says those generated assets draw on the domain, landing pages, existing ads and ad-group keywords. The guidelines apply to that generated layer.
Inside the filter
Google-optimized headlines and descriptions
Term exclusions and messaging restrictions decide whether generated text remains eligible. The asset report identifies this layer with “Google AI” in the Added by column.
Outside the filter
Targeting, source pages and written assets
The feature does not rewrite landing-page claims, repair targeting or replace the advertiser’s own review of responsive search ad copy.
This boundary matters as Google moves legacy automatically created assets into AI Max. PolymathDesk’s earlier guide to the Dynamic Search Ads and AI Max transition explains the migration layer; text guidelines solve the narrower problem of generated message control inside it.
Exact terms and message rules are different instruments
Term exclusions are literal and case-insensitive. Google documents a maximum of 25 exclusions per campaign, with each API value limited to 30 characters. They are language-specific, so excluding an English word does not automatically block its Japanese equivalent.
Messaging restrictions are broader instructions about concepts, associations or tone. Google allows up to 40 restrictions, and the API permits 300 characters per instruction. Unlike term exclusions, an English messaging restriction can also apply to generated assets in German or Dutch.
Literal gate
Term exclusions
Use them for exact forbidden words, phrases and competitor references. A symbol such as “$” will not reliably prohibit price claims.
25Maximum exclusions per campaign
Semantic gate
Messaging restrictions
Use direct hard requirements such as avoiding prices, city names or discount implications. The feature does not support audience-specific instructions or different capitalization rules by asset type.
40Maximum restrictions per campaign
The blast radius begins with assets already serving
Adding a guideline is not merely a prompt for future generation. Google states that the new rule also applies to existing Google-optimized assets. Noncompliant assets can stop serving and be marked as removed. A broad restriction can therefore reduce available combinations before a marketer sees any improvement in tone.
That risk is different from the budget tolerance discussed in the recent Smart Bidding Exploration experiment guide. Here, the changed variable is the eligible copy pool. The correct baseline is an asset inventory, not only campaign CPA or ROAS.
Signature module · The 48-Hour Copy Watch
Release the rule only when its copy pool can be audited
Google says customized assets are reviewed and refreshed as needed at least every 48 hours. Use that cycle as a change window, not as a promise that every asset will change on schedule.
- Capture the current poolExport campaign-level assets marked Google AI and note the combinations receiving impressions.
- Write one observable rule changeSeparate literal exclusions from concept restrictions so a removal has one plausible cause.
- Watch removals and regenerationReview the asset report after release and again across the next refresh window.
- Read combinations before performanceConfirm that compliant copy still produces useful ad combinations before judging spend or conversions.
Exit rule If the copy pool collapses or approved concepts disappear, revise the narrowest rule rather than adding another compensating instruction.
Copied guidelines create policy drift by design
Google lets advertisers copy guidelines from one Performance Max or AI Max Search campaign to another. The target can add to or replace its existing rules. After that operation, the copies are independent. Editing the source campaign does not update the target.
Source campaign
One approved rule set
Record a version, owner, business reason and evidence date before copying. This is the policy artifact that can be reviewed later.
Target campaign
An independent snapshot
Future edits require their own release and Copy Watch. Similar text does not mean the campaigns remain synchronized.
This is the same governance problem that appears when generated media needs an AI asset ledger. The interface can automate production and disclosure, but it cannot supply the organization’s version history or approval record.
Reporting closes the control loop
At ad and campaign level, the Added by column separates Google AI assets from advertiser-provided ones. Impressions, spend and conversions are available for customized text assets, while combination reporting shows which groups of assets most often served together.
Campaign-level reporting is Google’s recommended scale view. Combination reporting answers a different question: whether a compliant generated line is appearing with the advertiser’s intended claims. Neither report proves that a restriction caused a performance movement, so change one rule family at a time and preserve the pre-change inventory.
The operating rule is simple. Treat text guidelines as a versioned eligibility layer for Google-generated copy. Keep the landing page accurate, keep advertiser-written assets under human review, and give every guideline edit its own asset-level change window.
Questions about Google Ads text guidelines
Do text guidelines change advertiser-written headlines?
No. Google describes the feature as controlling assets made exclusively through text customization. Advertiser-provided responsive search ad assets remain a separate input.
Can a term exclusion block all price claims?
Not reliably. Google says excluding “$” will not prevent a generated line such as “Shoes for less than $50.” Use a direct messaging restriction that prohibits prices.
Do English rules affect ads in other languages?
It depends on the rule. Term exclusions are language-specific. Messaging restrictions can also apply to assets written in languages other than the restriction.
Will copied guidelines stay synchronized?
No. Google states that copied guidelines are independent. Editing the source campaign later does not update a campaign that received a copy.
