Reach swarm · Multi-location-rsa-drafting agent · Build pillar · Published July 31, 2026
How to build multi-location compliant RSA drafting for franchise and multi-store SEM
A multi-location retail or franchise operator running 50-500 locations across Google Ads + Microsoft Advertising + Apple Search Ads generates per-location Responsive Search Ad copy at scale. Per-platform RSA hard format constraints (15 headlines + 4 descriptions + character limits + asset pinning + Ad Strength scoring) interact with FTC substantiation, FTC Franchise Rule Item 19 FPR, Lanham Act, naked-licensing doctrine, ADA Title III + WCAG 2.2 AA, and EU AI Act Article 50 in ways single-LLM drafting cannot survive. This guide walks the 4-skill bundle (Draft + Constrain + Gate + Audit) end-to-end.
The 4-skill bundle on the multi-location-rsa-drafting agent
Draft
Generate per-asset RSA copy via multi-LLM ensemble (OpenAI + Anthropic + Google + Mistral + Cohere) with operator -counsel-approved system prompt injection covering brand voice spec, claims allowlist (sibling #496), forbidden phrase library (sibling #507), required disclosures per vertical, per-location context (location name + service area + hours + offer + landing page), per-jurisdiction context (per-state legal disclaimers + per-state cannabis + per-state alcohol restrictions + Prop 65 where applicable), and per-platform editorial constraint summary. RAG retrieval from Pinecone + Weaviate + Qdrant + Chroma + Milvus + pgvector indexes covering per-location substantiation, FDD Item 19 FPR substrate, FINRA Rule 2210 substrate when financial scope, and FTC Endorsement Guides AI-content disclosure substrate. Per-vendor LLM zero-retention posture verified per call.
Constrain
Enforce per-platform RSA format hard constraints at decode time via grammar-constrained decoding (Outlines + Guidance + LMQL + SGLang + Pydantic schema + JSON schema + regex + Context-Free-Grammar). Google Ads RSA: 15 headlines at 30 characters + 4 descriptions at 90 characters + 2 path slots at 15 characters + asset pinning rules + display URL format. Microsoft Advertising RSA equivalent. Apple Search Ads variant. A draft that violates a character or slot constraint is structurally impossible rather than caught downstream by Google Ads API validate-mutate. Cross-asset uniqueness via Levenshtein + Damerau-Levenshtein + Jaro-Winkler + embedding cosine + stylometric fingerprint catches near-duplicate headlines that erode Ad Strength even when each individual slot is policy-clean. Per-asset pinning strategy documented per-campaign per-ad-group with rationale.
Gate
Run per-platform editorial policy check (Google Ads Policies + Microsoft Advertising Policies + Apple Search Ads Policies + Google Healthcare and Medicines Policy when applicable). Run FTC compliance overlay (handoff to #516 marketing-compliance-overlay-for-regulated-industries for compound regulator scope when vertical reaches FDA + FINRA + CFPB scope). Run combinatorial-claim-closure substantiation check: every claim in any slot must be substantiable on its own AND in every plausible served combination (headline 7 with description 3 must not manufacture a deceptive net impression). Run FTC Franchise Rule Item 12 territorial cross-check (ad copy cannot imply exclusive geography contrary to FDD Item 12). Run franchisor agency-theory check (ad copy on franchisor-coordinated platform inherits franchisor exposure). Run ADA Title III + WCAG 2.2 AA landing page check. Run EU AI Act Article 50 AI-generated content disclosure check where served to EU. Produce allow/batch-review/escalate/reject with explainability trace and rule citation.
Audit
Per-asset canonical record (asset ID + campaign + ad group + per-location pointer + system-prompt snapshot + RAG retrieval snapshot + LLM ensemble snapshot + per -platform format constraint snapshot + per-slot character count + cross-asset uniqueness signals + asset-pinning strategy + per-platform editorial policy check snapshot + combinatorial-claim-closure substantiation snapshot + FTC Endorsement Guides AI-disclosure snapshot + FTC Franchise Rule Item 12 + Item 19 cross-check snapshot + Lanham Act + naked-licensing snapshot + ADA Title III + WCAG 2.2 AA landing page snapshot + EU AI Act Article 50 snapshot + Gate decision + canary rollout stage + Ad Strength prediction + Ad Strength realized). WORM storage. Per-asset record retains for FTC substantiation defense + FDD-amendment review + Google Ads editorial appeal + state-AG enforcement defense + EU AI Act supervisory authority review.
The real ecosystem this sits above
Ad platforms + SEM management
Google Ads, Microsoft Advertising, Apple Search Ads ad platforms. SA360, Optmyzr, Adalysis, Skai, Marin, Kenshoo, WordStream, Acquisio, Adzooma, Channable, Smartly.io SEM management. Google Ads API validate-mutate + Microsoft Advertising Bulk API + Apple Search Ads API for pre -publish editorial check.
LLM + grammar-constrained decoding + RAG
OpenAI, Anthropic, Google, Mistral, Cohere, Meta, AWS Bedrock, Azure OpenAI, Vertex AI LLM under per-vendor zero-retention. Outlines, Guidance, LMQL, SGLang, Pydantic, Zod, JSON schema, regex, CFG grammar-constrained decoding. Pinecone, Weaviate, Qdrant, Chroma, Milvus, pgvector RAG for per-location + FDD Item 19 + FINRA 2210 + FTC Endorsement Guides substrate.
Copy assistants + policy-as-code + WORM
Jasper, Copy.ai, Anyword, Persado, Phrasee, Ad-Lib copy assistants — useful as one input, not as gating authority. OPA Rego, AWS Cedar, Casbin, Cerbos, Oso, Styra DAS, Permit.io policy-as-code at Gate. AWS S3 Object Lock, Azure Blob immutable, Google Cloud Storage Bucket Lock, Wasabi compliance WORM for Audit.
The 5-anchor compliance overlay
Anchor 1 — Per-platform RSA format hard constraints + combinatorial-claim-closure substantiation (operationally distinctive)
Google Ads RSA: 15 headline slots at 30 characters + 4 description slots at 90 characters + 2 path slots at 15 characters + asset pinning rules + Ad Strength scoring. Microsoft Advertising RSA equivalent. Apple Search Ads variant. Google composes ads at serve time by picking from the slots based on context and Ad Strength signal; the operator can only constrain the slot library. Combinatorial-claim-closure substantiation: every claim in any slot must be substantiable on its own AND in every plausible served combination. Headline 7 paired with description 3 must not manufacture a deceptive net impression that no individual slot creates. Google Ads Policies + Microsoft Advertising Policies + Apple Search Ads Policies (healthcare + restricted + prohibited + personalized) + Google Healthcare and Medicines Policy + FTC Section 5 + substantiation doctrine (Pfizer 1972) + FTC Endorsement Guides 16 CFR Part 255 (2023 update on AI-generated content + influencer disclosure) + FTC Fake Review Rule 16 CFR Part 465 (Oct 2024) + FTC Health Products Compliance Guide + FTC Made-in-USA 16 CFR Part 323 + FTC Green Guides 16 CFR Part 260. Operationally distinctive — naive drafting treats each slot independently and produces 30-60 percent of combinations individually defensible but combinatorially problematic.
Anchor 2 — FTC Franchise Rule + per-state Franchise Investment Law + FDD Item 12 + Item 19 FPR
FTC Franchise Rule 16 CFR Part 436 + Item 12 territorial rights (ad copy cannot imply exclusive geography contrary to FDD Item 12) + Item 19 Financial Performance Representation when RSA copy implies outlet performance. Per-state Franchise Investment Law enforcement covers California Corporations Code 31000 et seq + Michigan MCL 445.1501 + Maryland Franchise Registration and Disclosure Law + Illinois Franchise Disclosure Act 815 ILCS 705 + Minnesota Minn Stat 80C + New York General Business Law Article 33 + Virginia Retail Franchising Act + Washington RCW 19.100 + 6 additional registration states.
Anchor 3 — Lanham Act + naked-licensing doctrine + franchisor agency-theory
Lanham Act 15 USC 1125(a) trademark misrepresentation when competitor marks appear in headlines or descriptions or when own marks are misused. Naked-licensing doctrine (Dawn Donut v Hart Food Stores 1959 + Stanfield v Osborne Industries 1995 + Doeblers Pennsylvania Hybrids 2010) when franchisees produce ad copy without franchisor quality control; franchisor that does not gate franchisee copy risks losing mark protection. Franchisor agency-theory exposure (Restatement Third of Agency Sec 7.07) for ad copy on franchisor-coordinated platforms with franchisor-approved templates.
Anchor 4 — ADA Title III + WCAG 2.2 AA when RSA points at non-compliant landing
ADA Title III places of public accommodation + WCAG 2.2 AA accessibility apply to landing page targets. Robles v Dominos Pizza 9th Cir 2019 confirmed that web accessibility is enforceable under Title III. Gil v Winn-Dixie 11th Cir 2021 added circuit-split clarity. Serial-plaintiff exposure is real. Gate runs landing page WCAG check before approving any RSA pointing at a landing page; landing page accessibility failure routes to engineering with the ad held in batch review until landing remediation lands.
Anchor 5 — EU AI Act Article 50 + Article 13 + 14 + 15 + NIST AI RMF + ISO 42001 + per-vendor LLM zero-retention
EU AI Act Article 50 transparency for AI-generated content (where served to EU + August 2026 phased applicability) + Article 13 transparency + Article 14 human oversight + Article 15 accuracy and robustness + Article 26 deployer obligations. NIST AI RMF Govern + Map + Measure + Manage + ISO 42001 AI Management System. Per-vendor LLM zero-retention posture verified before any operator content or per-location context is sent to LLM endpoint at Draft. Verification record retained per LLM call.
The 6-workstream pre-engagement-baseline reporting cycle
Completions does not commit to numeric CTR or quality-score targets before engagement scope is documented. The Q6 pre -engagement-baseline reporting cycle covers the six workstreams that ship in every engagement.
- Draft coverage. Per-location context completeness + per-vertical context completeness + per -jurisdiction context completeness + brand-voice spec freshness + claims-allowlist freshness + forbidden-phrase library freshness + RAG retrieval coverage + per-vendor LLM zero-retention verification freshness.
- Constrain quality. Per-platform format constraint coverage + grammar-constrained decoding correctness + per-slot character compliance + cross-asset uniqueness signal freshness + asset-pinning strategy rationale freshness.
- Gate quality. Per-platform editorial policy check freshness + FTC compliance overlay handoff freshness + combinatorial-claim-closure substantiation completeness + FTC Franchise Rule Item 12 + Item 19 cross -check freshness + Lanham + naked-licensing check freshness + ADA + WCAG landing page check freshness + EU AI Act Article 50 disclosure freshness.
- Audit quality. Per-asset canonical record completeness + WORM storage posture + per-asset rule citation freshness + per-asset chain-of-thought trace freshness.
- Compliance posture. Per-platform editorial policy operator-counsel signoff + FTC Section 5 + substantiation + Endorsement Guides 2023 + Fake Review Rule + Health Products Compliance + Made-in-USA + Green Guides + FTC Franchise Rule + per-state Franchise Investment Law + Lanham Act + naked-licensing doctrine + franchisor agency-theory + ADA Title III + WCAG 2.2 AA + EU AI Act Article 50 + 13 + 14 + 15 + 26 + NIST AI RMF + ISO 42001 + per-vendor LLM zero-retention freshness.
- Audit-trail completeness. Per-Draft + per -Constrain + per-Gate + per-Audit canonical record retention in versioned-history substrate readable by FTC substantiation defense + FDD-amendment review + Google Ads editorial appeal + state-AG enforcement + EU AI Act supervisory authority review.
Frequently asked questions
What problem does multi-location compliant RSA drafting solve for a franchise or multi-store SEM operator?
A multi-location retail or franchise operator running 50-500 locations across Google Ads + Microsoft Advertising + Apple Search Ads generates per-location Responsive Search Ad (RSA) copy at scale. Per-platform RSA hard format constraints (15 headlines + 4 descriptions + 30 character per headline + 90 character per description + 2 path slots at 15 character + asset pinning rules + Ad Strength scoring) interact with FTC substantiation discipline, FTC Endorsement Guides (2023 AI-content disclosure), FTC Fake Review Rule (October 2024), FTC Franchise Rule Item 19 FPR when ad copy implies outlet performance, FDD Item 12 territorial rights when ad copy implies exclusive geography, Lanham Act when competitor marks appear, naked-licensing doctrine when franchisees produce ad copy without franchisor quality control, and ADA Title III + WCAG 2.2 AA when ad copy points at a non-compliant landing page. Single-LLM drafting that does not encode these constraints either produces copy that fails Google Ads editorial review or produces copy that passes editorial review but creates regulatory exposure. The skill ships the substrate that makes per-location RSA copy defensible at portfolio scale.
What is the 4-skill bundle and what does each skill do?
Draft generates per-asset RSA copy via multi-LLM ensemble (OpenAI + Anthropic + Google + Mistral + Cohere) with operator-counsel-approved system prompt injection covering brand voice, claims allowlist (sibling skill #496), forbidden phrase library (sibling #507), required disclosures per vertical, per-location context, and per-jurisdiction context. Per-vendor LLM zero-retention posture verified per call. Constrain enforces per-platform RSA format hard constraints (Google Ads 15 headlines + 4 descriptions + 30 char headline + 90 char description + 2 path 15 char + Microsoft Advertising same + Apple Search Ads variant) via grammar-constrained decoding (Outlines + Guidance + LMQL + SGLang + Pydantic schema + JSON schema + regex-constrained + Context-Free-Grammar) so a draft that violates a character or slot constraint is structurally impossible rather than caught downstream. Constrain also runs cross-asset uniqueness (Levenshtein + Damerau-Levenshtein + Jaro-Winkler + embedding cosine + stylometric fingerprint) to surface near-duplicate headlines that erode Ad Strength. Gate runs the per-platform editorial policy check + FTC compliance overlay (handoff to #516 marketing-compliance-overlay) + franchisor agency-theory check + asset pinning strategy + Ad Strength prediction + canary rollout staging. Audit ships per-asset canonical records to WORM storage for FTC substantiation defense + FDD-amendment review + Google Ads editorial-appeal preparation + state-AG enforcement defense.
Why is per-platform RSA format hard constraint combined with FTC substantiation the operationally distinctive anchor for this skill?
A Google Ads RSA has 15 headline slots (30 character each) and 4 description slots (90 character each). Google composes ads at serve time by picking from the slots based on context and Ad Strength signal. An operator cannot predict which headline-description combination will show; the operator can only constrain the slot library. This interacts with FTC substantiation: every claim in any slot must be substantiable on its own AND in every plausible combination because Google may serve headline 7 with description 3 to the same query that serves headline 12 with description 1. A claim of fastest service in headline 7 needs substantiation even when paired with description 3 mentioning low price; the combination cannot manufacture a deceptive net impression that no individual slot creates. Operationally distinctive frame: the RSA slot library is a combinatorial claim surface, not a sequential one, and FTC substantiation applies to the combinatorial closure. Naive drafting treats each slot independently and produces 30-50 percent of combinations that are individually defensible but combinatorially problematic. The skill encodes the combinatorial closure at Constrain + Gate so the slot library survives FTC review under any served combination.
What real regulatory and standards-body hooks does the compliance overlay anchor on?
Anchor 1 is per-platform RSA format hard constraints + Google Ads Policies + Microsoft Advertising Policies + Apple Search Ads Policies (healthcare + restricted + prohibited content + personalized advertising) + Google Ads Healthcare and Medicines Policy + FTC Section 5 + FTC substantiation doctrine (Pfizer 1972 reasonable-basis) + FTC Endorsement Guides 16 CFR Part 255 (2023 update on AI-generated content + influencer disclosure) + FTC Fake Review Rule 16 CFR Part 465 (Oct 2024) + FTC Health Products Compliance Guide + FTC Made-in-USA 16 CFR Part 323 + FTC Green Guides 16 CFR Part 260 + combinatorial-claim-closure discipline (slot library is combinatorial not sequential). Anchor 2 is FTC Franchise Rule 16 CFR Part 436 + Item 12 territorial rights + Item 19 Financial Performance Representation when RSA copy implies outlet performance + per-state Franchise Investment Law FPR enforcement (California Corporations Code 31000 et seq + Michigan MCL 445.1501 + Maryland Franchise Registration and Disclosure Law + Illinois Franchise Disclosure Act 815 ILCS 705 + Minnesota Minn Stat 80C + New York GBL Article 33 + Virginia Retail Franchising Act + Washington RCW 19.100). Anchor 3 is Lanham Act 15 USC 1125(a) trademark misrepresentation when competitor marks appear in headlines or descriptions + naked-licensing doctrine (Dawn Donut v Hart Food Stores 1959 + Stanfield v Osborne Industries 1995 + Doeblers Pennsylvania Hybrids 2010) when franchisees produce ad copy without franchisor quality control + franchisor agency-theory exposure (Restatement Third of Agency Sec 7.07). Anchor 4 is ADA Title III + WCAG 2.2 AA when RSA copy points at a non-compliant landing page (Robles v Dominos 9th Cir 2019 + Gil v Winn-Dixie 11th Cir 2021 + serial-plaintiff exposure). Anchor 5 is EU AI Act Article 50 transparency for AI-generated content + Article 13 + Article 14 human oversight + Article 15 accuracy + NIST AI RMF + ISO 42001 + per-vendor LLM zero-retention verified before any operator content or per-location context is sent to LLM endpoint.
How does Constrain prevent format failures at scale?
Naive LLM drafting produces copy that violates Google Ads format constraints frequently: headlines over 30 characters, descriptions over 90 characters, paths that include disallowed characters, asset pinning that violates Ad Strength advice, dynamic keyword insertion that overflows when the keyword is long. Caught downstream by Google Ads API validate-mutate, these failures generate retry loops that burn LLM cost and timeline. Constrain enforces the format at decode time via grammar-constrained decoding (Outlines + Guidance + LMQL + SGLang + Pydantic + JSON schema + regex + CFG) so a draft that violates a character or slot constraint is structurally impossible. Cross-asset uniqueness via Levenshtein + Damerau-Levenshtein + Jaro-Winkler + embedding cosine + stylometric fingerprint catches near-duplicate headlines that erode Ad Strength even when each individual slot is policy-clean. Per-asset pinning strategy is documented per-campaign per-ad-group with rationale so an editorial reviewer can answer the question why this asset is pinned to position 1.
What does Completions ship and how does an engagement start?
Completions ships the multi-location-rsa-drafting agent + 4-skill bundle (Draft + Constrain + Gate + Audit) + 5-anchor compliance overlay (per-platform RSA format + Google Ads + Microsoft Advertising + Apple Search Ads editorial policy + FTC Section 5 + substantiation + Endorsement Guides 2023 + Fake Review Rule + Health Products Compliance + Made-in-USA + Green Guides + FTC Franchise Rule Item 12 + Item 19 FPR + per-state Franchise Investment Law + Lanham Act + naked-licensing doctrine + franchisor agency-theory + ADA Title III + WCAG 2.2 AA + EU AI Act Article 50 + 13 + 14 + 15 + NIST AI RMF + ISO 42001 + per-vendor LLM zero-retention) + the Q6 6-workstream pre-engagement-baseline reporting cycle. Tier 1 AI Readiness Assessment ($10k, 2-3 weeks) audits the current per-location RSA drafting posture, the combinatorial-closure substantiation chain, and the per-asset audit trail. Tier 3 Fractional CMO with AI Swarm ($15-25k/month, 6-month minimum, 1-2 days/wk embedded) runs the multi-location-rsa-drafting agent on the operator SEM stack on an ongoing basis.
Engage Completions on the multi-location-rsa-drafting agent
Tier 1 AI Readiness Assessment ($10k, 2-3 weeks) audits the current per-location RSA drafting posture, the combinatorial -closure substantiation chain, and the per-asset audit trail. Tier 3 Fractional CMO with AI Swarm ($15-25k/month, 6-month minimum, 1-2 days/wk embedded) runs the multi-location-rsa -drafting agent on the operator SEM stack on an ongoing basis.