Google AI Max Text Guidelines: 27% Conversion Lift Guide
Google AI Max text guidelines went global Feb 26, 2026. 25 term exclusions, 40 messaging restrictions, 27% lift. Brand voice-to-restriction framework.
Conversion Lift
Term Exclusions
Messaging Rules
BYD Lead Lift
Key Takeaways
On February 26, 2026 — one day after the Google Ads API v23.1 release exposed Campaign.text_guidelines to developers — Google completed the global rollout of AI Max text guidelines to all advertisers worldwide. The published benchmarks are real: an average 14% conversion lift across AI Max campaigns, up to 27% lift for campaigns that historically relied on exact and phrase match keywords, and a flagship BYD case study reporting 24% more leads at 26% lower CPL. Those numbers are the headline; the craft is in how you configure the 25 term exclusions and 40 messaging restrictions you get per campaign.
This guide is the implementation reference. We walk through what “27% lift” actually measures, the term-exclusion vs. messaging-restriction tradeoffs, a 25-slot prioritization framework, a brand-voice-to-restriction translation pattern, and the match- type strategy that maximizes incremental lift. For the programmatic control side — how to ship this configuration from an Ads API mutation — cross-reference the Google Ads API v23.1 playbook.
Two numbers, two audiences: use 14% for conservative internal forecasting. Use 27% when pitching campaigns that live mostly on exact + phrase match — those campaigns are where AI Max unlocks the most growth because they had the least flexibility before.
Interpreting the 27% Lift Benchmark
Google's published numbers are defensible but require the right baseline. Three tiers:
| Campaign type | Reported lift | Why |
|---|---|---|
| Exact + phrase match heavy | Up to +27% | Was artificially constrained; AI Max unlocks nearby intent. |
| Mixed match type | +14% avg | Already captured broad intent; AI Max adds creative matching. |
| Broad-match-only | Lower (not officially cited) | Already most flexible; incremental creative gain diminishes. |
What Actually Shipped February 26
- Global availability — all languages, all verticals, all advertisers.
- 25 term exclusions per campaign — exact strings, case-insensitive, language-specific, 30 char max each.
- 40 messaging restrictions per campaign — natural-language rules, semantic interpretation, cross-language, 300 char max each.
- UI + API parity — the same day v23.1 exposed the API surface, the UI rolled out to all accounts.
- AI Max for Search + Performance Max — both surfaces use the
Campaign.text_guidelinesfield.
Term Exclusions vs. Messaging Restrictions
- 25 per campaign, 30 char max each
- Case-insensitive
- Language-specific — no cross-language coverage
- Best for: competitor names, banned superlatives, specific compliance strings
- 40 per campaign, 300 char max each
- Semantic interpretation
- Cross-language — one rule applies everywhere
- Best for: tone, brand voice, topic exclusion, content categories
The 25-Slot Term Exclusion Framework
Treat the 25 slots as budget. Four tiers we prioritize by:
| Tier | Slots | Purpose | Examples |
|---|---|---|---|
| A — Compliance hard stops | 5-8 | Regulated claims you legally cannot make | “guaranteed cure,” “FDA approved,” “risk-free” |
| B — Brand rails | 5-8 | Voice guardrails — words your brand never uses | “cheap,” “best ever,” “amazing” |
| C — Competitor names | 5-8 | Named competitors you don't want to reference | Specific product + company names |
| D — Channel-specific strings | 3-7 | Strings that make sense in product copy but break ad copy | “click here,” “terms apply,” internal code names |
Don't burn slots on translations. Term exclusions are language-specific. If your compliance team needs the same ban across English, Spanish, and German, use a messaging restriction (cross-language) instead of three separate term exclusions.
Brand Voice → Messaging Restrictions
Brand guidelines are usually English prose written for humans. The translation to messaging restrictions is what actually controls AI output. Four-step pattern:
- Extract the negative statement.Brand books say “We are warm, not chilly.” The messaging restriction is “Don't use distant or clinical language.”
- Make it specific.“Avoid hype” is too vague. “Don't use superlatives (best, greatest, ultimate) unless substantiated by a cited study” is actionable.
- Map compliance to rules.“Never promise outcomes,” “Don't imply FDA endorsement,” “Avoid claims of instant results.”
- Test at the boundary. After publishing the restrictions, manually generate a dozen assets. Iterate until the output consistently respects the rule.
Sample starter set (B2B SaaS)
// Messaging restrictions (up to 40 per campaign)
1. Don't generate assets with prices.
2. Don't use superlatives unless substantiated by a cited study.
3. Never imply specific outcome guarantees (e.g., "will increase sales by X%").
4. Avoid medical or health-adjacent claims.
5. Don't reference unannounced product features.
6. Avoid the words "revolutionary," "disruptive," or "game-changing."
7. Keep CTAs action-first: "See how," "Talk to sales," "Book a demo."
8. Don't use technical jargon more than one layer below the product name.
9. Avoid competitor names or references to competitors' pricing.
10. Use US English spelling conventions consistently.
Pairing With Keyword Match Strategy
Text guidelines don't eliminate match-type strategy — they amplify it. Four pairings we run in production:
| Campaign intent | Match type | Text guidelines focus |
|---|---|---|
| Brand defense | Exact | Tight — lock brand voice; 25 term exclusions for competitor names |
| Core product | Phrase + exact | Moderate — 10-15 messaging restrictions; medium term exclusion use |
| Category discovery | Broad + AI Max | Loose — 20-30 messaging restrictions shape voice; minimal term exclusions |
| High-regulation vertical | Exact (legal-reviewed) | Maximum restrictiveness — all 25 term slots + 40 messaging rules |
BYD Case Study Walkthrough
Google publicly featured BYD as an early partner. The reported outcome: 24% increase in leads at 26% lower cost per lead across global campaigns — via text guidelines used to maintain brand standards while scaling AI-generated creatives.
The lessons from what BYD did (extrapolated from the public case):
- Consolidated global brand voice into messaging restrictions — not per-market term exclusions.
- Used term exclusions only for competitor brand names and non-localizable strings.
- Let AI Max expand match outside exact + phrase keywords, where the 27%-lift tier lives.
- Reviewed generated assets weekly across top-5 markets, iterating restrictions instead of pausing campaigns.
The 5 Most Common Mistakes
- Burning term slots on translations. 25 is a cap — save slots for language-specific compliance strings; everything cross-language belongs in messaging restrictions.
- Vague restrictions.“Keep it professional” is uninterpretable. “Don't use slang, exclamation marks, or caps-lock emphasis” is.
- Blocking prices with “$”. Doesn't work. Use the messaging restriction “Don't generate assets with prices.”
- Copy-pasting brand-book prose.Brand books are written for humans. Messaging restrictions need imperative negatives (“Don't X”).
- Shipping once, never iterating. AI Max output changes as the underlying models change. Review asset output at least monthly; update restrictions when drift appears.
Preparing for the DSA → AI Max Cutover
Google has announced that Dynamic Search Ads are being upgraded to AI Max in September 2026. Existing DSA campaigns will transition automatically. Every DSA you run today needs a text-guidelines plan before the cutover because AI Max generates copy more aggressively than DSA, which means brand-risk surface scales with match flexibility.
Need a turn-key DSA → AI Max migration plan? Our PPC advertising team audits existing DSA campaigns, builds the text-guidelines policy, and operates the post-cutover review cadence.
Agency Monthly Text-Guideline Workflow
- Week 1 — Pull asset samples. Export the 50 highest-impression AI-generated headlines + descriptions per priority campaign.
- Week 2 — Review against the policy. Flag drift. Catalog the specific words or phrases that crossed a line.
- Week 3 — Update restrictions. Patch the policy, push via the Ads API so the change is version-controlled.
- Week 4 — Monitor conversion lift. Compare week-over-week conversions and CPA. Policy tightening can cost lift if over-applied.
Conclusion
AI Max text guidelines turn Performance Max and AI Max from “creative wildcards” into “governed channels.” The 27% lift is real for the right campaign mix, but the unlocked upside depends on how you use the 25 term-exclusion and 40 messaging-restriction slots. Tight, specific, cross-language rules win. Vague brand-book prose does not.
Ship AI Max Text Guidelines That Actually Lift Conversions
We audit your campaigns, translate your brand voice into programmatic restrictions, and operate the monthly review cadence that keeps AI Max on-brand without sacrificing reach.
Frequently Asked Questions
Related Guides
More on programmatic PPC, AI Max, and brand-safe ad generation.