Updated September 10, 2026/Google AdsAd Copy

Google Ads Character Limits 2026: RSA Compliance Checklist

AI ad generators can violate Google's character limits when their output is not checked programmatically. Here's why token generation is not a reliable character counter and what a compliant workflow should do.

DG

Digraph Team

Ad Generation Research

📏

Character Compliance

Google Ads has a hard constraint that most AI tools pretend doesn't exist: character limits. Headlines max out at 30 characters. Descriptions cap at 90. These aren't suggestions—they're binary.

Why LLMs Can't Count

Large language models generate token sequences rather than executing a guaranteed character-counting routine. A prompt can improve first-pass compliance, but the result still needs a deterministic check that uses the relevant language and counting rules before it is sent to an ad platform.

The Compliance Stack That Actually Works

Reliable ad generation requires a multi-layer validation system that treats the LLM's output as a first draft, not a finished product:

  • Layer 1: Prompt optimization - Maximizes first-pass compliance
  • Layer 2: Post-generation validation - Programmatic character checks
  • Layer 3: Policy filtering - Content policy compliance checks
  • Layer 4: Auto-retry - Targeted regeneration of failed outputs

Beyond Character Counts

Character limits get the most attention, but Google Ads compliance is broader. Punctuation restrictions, trademark policies, editorial standards, and claims substantiation all matter.

A practical pre-publish checklist

  1. Count every asset deterministically: validate each headline, description, and path field with the account’s language and format rules.
  2. Check the served combination: responsive search ads can assemble assets in different combinations, so review pinning, repetition, truncation, and the actual preview.
  3. Verify the claim: compare prices, guarantees, comparisons, and superlatives with an approved source and the landing page.
  4. Review policy context: flag trademarks, restricted categories, misleading urgency, and offers whose conditions are not visible on the destination.
  5. Keep an audit trail: store the generated draft, validation result, reviewer decision, final asset, and the specification version used.

This is a compliance workflow, not a guarantee that Google will approve every ad. Platform specifications can change, and final review belongs with the account owner. For the broader relationship between AI answers and paid performance, see how to measure AI search visibility and paid performance. For message testing, read scarcity versus social proof in paid search.

Validate with a programmatic checker

Treat generated copy as a draft. Count each headline, description, and path with the same rules used by the destination platform, flag double-width characters where relevant, check required assets, and reject or regenerate failures. Store the original output and the final approved variant so the team can audit what changed. A validator should fail closed on uncertainty instead of silently truncating the text.

Compliance is more than length

Check the landing page, trademark use, substantiation for superlatives, disclosure of offers, and the match between the ad and the destination. Keep a human review for claims that affect safety, price, eligibility, or regulated categories. The best generator saves review time by producing structured, testable drafts—not by pretending that a model can guarantee policy approval.

Current Google Ads references

A repeatable generation pipeline

A production workflow can be: retrieve approved product facts, generate multiple candidate assets, count and validate them, check claims and policy constraints, compare the landing page, route exceptions to a reviewer, and log the final variant. Keep the model away from private customer data unless the workflow is explicitly approved. The goal is controlled variation around verified facts, not maximum copy volume.

Make the checker language-aware

Character handling can differ for double-width scripts and other localized assets. Test the same strings with the account’s language and format rules, preserve Unicode correctly, and validate the final rendered asset rather than only a prompt response. Keep the checker versioned because platform specifications and ad formats can change.

Google’s current responsive-search guidance lists up to 30 characters for headlines, up to 90 for descriptions, and up to 15 for path fields, with additional language-specific counting considerations. Specs can change, so validate against the live documentation and the account format before shipping.