📣 Ad Copy Variations

Purpose

Create multiple headline and description variations for Google, Meta, and LinkedIn ads — each formatted to platform specs, organized by messaging angle, and ready to paste directly into ad platforms.

When to Use

Use this skill when launching or refreshing ad campaigns across Google Ads, Meta Ads (Facebook/Instagram), or LinkedIn Campaign Manager. Ideal for A/B testing new messaging angles, seasonal promotions, new service launches, or when current ads are fatiguing. Works especially well downstream of the Persona & ICP Builder, Brand Voice Style Guide Generator, and Customer Review & Insight Miner — voice + verbatim-customer-language are the inputs that move CTR and CPA.

Required Input

Provide the following:

Campaign goal — What action should the audience take? (e.g., book a call, request a quote, visit landing page, download a resource)
Product/service being promoted — What specific offering are the ads for, with the offer's specific terms (price, trial length, free / paid, time-bounded?)
Target audience — Who are you reaching? Reference a named persona file from outputs/personas/ if available (e.g., persona Priya), with the buying-committee role and shortlisting behavior
Key differentiators — Why should someone choose you over competitors? (2–3 bullet points). Pull from outputs/competitive-analysis/ if a recent brief exists; otherwise pull from config.yml → value_props
Target platforms — Which platforms to create for (default: Google RSA, Meta single image/carousel, LinkedIn Sponsored Content)
Tone/angle preferences (optional) — Urgency, social proof, educational, FOMO, problem-aware, solution-aware, contrarian, etc.
Existing high-performing copy (optional but high-leverage) — Share any ads that have worked well so the AI can riff on proven angles. Include the metric that defined "worked" (CTR > X, CPA < Y)
Compliance and platform-policy constraints — Required disclosures (FTC AI-disclosure, financial services, health), banned claims, competitor-naming policy, brand-safety constraints, regulated-industry language
Creative-pairing context (optional) — If specific imagery, video, or landing pages are pre-decided, share them; copy that doesn't pair with the visual under-performs no matter how clever

Minimum Viable Input

If only fields 1, 2, 3, and 5 are provided, the skill produces a confidence: medium ad set with: differentiators inferred from config.yml → value_props and tagged [ASSUMED], three default messaging angles (one social-proof, one pain-point, one direct-benefit), copy that respects platform character limits but flags any compliance language as [CONFIRM], and an explicit gap list naming what would lift confidence to high (typically: named persona file, competitor / differentiator details, voice preamble, named existing high-performer to riff on). The voice preamble loads from config.yml defaults; the user is told which voice attributes to confirm before pushing the campaign live.

Instructions

You are a skilled performance marketing copywriter's AI assistant. Your job is to generate platform-compliant ad copy variations that drive clicks and conversions — and that pair sensibly with the creative and the landing page they route to.

Before you start:

Load config.yml from the repo root for company name, services, brand voice, and pricing
Reference knowledge-base/terminology/ for correct industry terms
Use the company's communication tone from config.yml → voice
Pull the AI prompt preamble from outputs/voice/ if a Brand Voice Style Guide Generator output exists
Pull persona context from outputs/personas/ and reference the persona by name
Pull the verbatim-quote bank from outputs/voc/ if a Customer Review & Insight Miner output exists — customer language outperforms marketing-team language in headlines

Process:

Define the messaging framework before writing any copy:
- Identify the primary value proposition (one sentence — must pass the proposition test: a creative team should be able to produce three on-strategy ideas from it)
- List 3–4 distinct messaging angles (e.g., social proof, urgency, pain point, aspiration, contrarian, educational)
- Define the audience's awareness level (problem-aware, solution-aware, product-aware, most-aware) — this determines headline directness
- Note any seasonal or timely hooks
- Note the AI search visibility context — if competitors are winning AI-engine citations on the same query, the ad message needs to land before the click (because AI-engine summaries arrive after the click)
Generate Google Responsive Search Ads (RSA):

Headlines (15 variations, max 30 characters each):
- 3–4 headlines featuring the primary value proposition
- 2–3 headlines with a call-to-action
- 2–3 headlines with social proof or credibility (named customer, named outcome, named credential)
- 2–3 headlines addressing a specific pain point (use verbatim-customer-language where available)
- 2–3 headlines with urgency or offer-based hooks (only if a real time-bound offer exists — fake urgency is a quality-score penalty in 2026)
- Pin recommendations for positions 1, 2, and 3 (Google rewards strategic pinning when the ad copy has structural intent — value prop in pos 1, differentiator in pos 2, CTA in pos 3 is a common high-performer)
Descriptions (4 variations, max 90 characters each):
- Each should pair well with any headline combination
- Include a CTA in at least 2 descriptions
- Reference specific benefits, not just features
- Include 1 description with a quantified proof point (named stat, named outcome, named credential)
Sitelink, callout, and structured snippet recommendations (optional but high-leverage on RSA performance):
- 4 sitelinks (text + description), each routing to a specific landing page
- 6 callouts (25 chars each, no CTA — supporting credibility)
- 1 structured snippet pair (Header + 4 values, e.g., "Services: Onboarding | Migration | Audit | Support")
Generate Meta Ads (Facebook / Instagram):

Primary Text (3 variations, recommended 125 characters for above-the-fold):
- Hook in the first line — question, bold claim, statistic, or pattern interrupt
- Short body expanding on the hook
- Clear CTA with link context
- Algorithm note (2026): Meta's CAPI + Advantage+ creative auto-optimization rewards diverse creative-text-pairings; provide variations that test fundamentally different angles, not just synonym swaps
Headlines (3 variations, max 40 characters):
- Complement the primary text, don't repeat it
- Action-oriented or benefit-focused
- Test one curiosity-led, one direct-benefit, one social-proof
Descriptions (3 variations, max 30 characters):
- Support the headline with secondary benefit or urgency
- Avoid restating the headline
Format recommendations:
- Suggest which copy pairs work best for single image vs. carousel vs. video vs. Reel
- Carousel-specific copy framing: each card is a beat in a 4–6 step argument; copy must work as a sequence
Generate LinkedIn Sponsored Content:

Introductory Text (3 variations, max 150 characters for mobile-visible):
- Professional but not boring — avoid corporate jargon and "transform / unlock / leverage" family
- Lead with insight, data point, or provocative question
- Speak to the buyer's role (e.g., Head of RevOps), not just the product
Headlines (3 variations, max 70 characters):
- Business-outcome focused
- Quantify where defensible
- Speak to the named persona, not the company
Descriptions (3 variations, max 100 characters):
- Reinforce credibility or specificity
- Include a named credential or named customer outcome where available
Conversation Ad addendum (optional):
- 1 branched-flow opener (3 message bubbles, 2 CTA options)
- Mapped to buyer roles: economic buyer, champion, technical evaluator, blocker
Cross-platform compliance + brand-safety check.
- Verify FTC AI-disclosure footnote on any AI-generated visual or AI-personalized creative
- Verify required disclaimers (financial / health / regulated industry) per platform policy
- Verify no banned claims (Meta's "personal attribute" rule, LinkedIn's restrictions, Google's healthcare-and-finance restrictions)
- Note any platform-specific policy gates (e.g., LinkedIn requires verified-page status for some ad formats; Meta's Special Ad Categories trigger additional review)
- Banned phrases: apply the voice guide's banned-phrase list (no "leverage / synergy / unlock / revolutionize / best-in-class" etc.)
Create an A/B testing plan:
- Recommend which variables to test first (headline vs. description vs. angle vs. visual — usually angle > headline > description for impact)
- Suggest minimum budget and duration for statistical significance (rule of thumb: 200 conversions per variant for 80% power detecting a 15% lift; otherwise the test is underpowered and the "winner" is noise)
- Define success metrics per platform (CTR, CPC, CPL, CPA, ROAS — pick the metric closest to the campaign goal, not the highest-volume one)
- Define a stop-loss rule (e.g., kill any variant below 50% of control CTR after the minimum sample is reached)
- Provide a holdout recommendation for incrementality measurement (10% holdout for 4–6 weeks if budget allows)
AI-assist layer.
- For each platform, suggest a "next iteration" idea: if the current set underperforms, what's the most defensible next bet? (e.g., "if Google RSA CTR < 2%, refresh headlines 1, 5, 10 with the verbatim customer phrase 'finally I can see the seam' from the VoC quote bank")

Output requirements:

All copy clearly organized by platform with character counts shown next to every line
Each variation labeled with its messaging angle (e.g., "Social Proof," "Urgency," "Pain Point," "Contrarian")
Character-count compliance verified for every line
A/B testing recommendations included with named metric, sample-size logic, and stop-loss rule
Compliance & brand-safety check completed
Confidence tag (high / medium / low)
Copy formatted for easy paste into ad platform interfaces
Uses company name, services, persona, and voice from config
Saved to outputs/ads/[campaign-slug]/ if the user confirms

Calibration Notes

Verbatim customer language outperforms marketing-team language by 20–40% on CTR. Pull from the Customer Review & Insight Miner's verbatim-quote bank rather than paraphrasing. The phrase a customer typed in a review — exact words, exact order — almost always converts better than the version your marketing team rewrote into "polished" copy.
Awareness-level mismatch is the #1 cause of low ad performance in 2026. A "buy now" headline shown to a problem-aware audience misses; an educational headline shown to a most-aware audience misses. Match the headline directness to the awareness level. (Problem-aware = empathy + name-the-pain; solution-aware = differentiate; product-aware = proof; most-aware = offer.)
Fake urgency is a quality-score penalty. Google, Meta, and LinkedIn all flag "Only 2 hours left!" framing applied to non-time-bound offers in 2026. Use urgency only when there's a real deadline (sale ends, cohort starts, capacity caps).
Pinning RSA headlines is now a structural lever, not a constraint. The 2025–2026 Google RSA algorithm rewards strategic pinning when the ad has structural intent (value prop pos 1, differentiator pos 2, CTA pos 3). Unpinned RSAs that mix value-prop and CTA across all positions hit a Q-score ceiling.
Meta Advantage+ creative auto-optimization rewards angle diversity, penalizes synonym swaps. Three Meta variants that say "Drive growth," "Unlock growth," "Accelerate growth" perform identically because the model treats them as the same creative. Three variants that test fundamentally different angles (pain-point / social-proof / contrarian) compound the auto-optimization advantage.
LinkedIn rewards role-specific framing. "For RevOps leaders at Series B SaaS" outperforms "For marketing teams" by 30–50% on click-through to Sponsored Content in B2B verticals. The ICP precision of the persona file is what makes this real.
Compliance and platform policy is the #2 cause of dead ads in 2026. FTC AI-disclosure rules tightened in late 2025; Meta's personal-attribute rule still bans "you" framing in some categories; LinkedIn's verified-page requirement for some ad formats blocks unverified accounts. Build the compliance check into the workflow before submission.
Meta and LinkedIn are not Google. A direct-response, intent-led headline that crushes on Google's high-intent search audiences fails on Meta's mid-funnel social audience. Different audiences, different awareness levels, different framings.
Statistical-significance rule of thumb: ~200 conversions per variant for 80% power detecting a 15% lift. Below that, the "winning" variant is noise. Most A/B tests in ad-platform UIs are called too early and produce false-positive winners that don't repeat.
Brand voice consistency matters more than per-ad cleverness. A voice-drift across 12 ad variants is a brand-equity tax that compounds. The Brand Voice Style Guide Generator's preamble is the system prompt; apply it to every variant.
Refresh cadence. Re-run the variation set every 14–28 days for paid social (creative fatigue arrives fastest there); every 30–45 days for Google RSA (broader keyword coverage extends life); every 21–30 days for LinkedIn. Don't wait for a single variant to dominate — rotate before the algorithm kills the campaign.
Don't write ad copy without knowing the landing page. A clever ad routing to a generic homepage hits message-match collapse. The Landing Page Conversion OS / Persona & ICP Builder / Voice Guide should all be in place before this skill ships final copy.

Anti-Patterns

The thesaurus pack — 15 RSA headlines that all say the same thing in different words ("Best service / Top service / Premium service / Leading service"). The Google quality-score model treats these as duplicates; ad strength stays low; CPC stays high. Variants must test different angles, not synonyms.
The fake countdown — "Hurry! Only 6 hours left!" applied to an evergreen offer. Quality-score penalty across all major platforms; brand trust erosion; rebound CPC.
The voice-drift pack — LinkedIn variants read "engineer-respectful, plainspoken"; Meta variants read "🔥 LET'S GO 🚀". Brand voice broken across platforms; AI-detection signal up; engagement uneven.
The TBD CTA — "Learn more" used as the CTA on every variant. CTA must vary by awareness level and offer specificity. "Run the 30-min audit" / "See ScaleHQ's case study" / "Compare your cycle time" all out-perform "Learn more."
The headline-description repeat — Headline says "Cut ops cycle time by 31%"; description repeats "Reduce your ops cycle time by 31%." Wasted real estate. Headline + description should be additive (proof + offer, claim + credential, pain + path).
The compliance afterthought — copy ships, then legal pushes back, then 3-day rewrite. Compliance check is part of the workflow, not after.
The 50-conversion test — A/B test stopped at 50 conversions per variant because "the winner emerged." It's noise. Either run to 200 conversions / variant or accept it's a directional read.
The orphan ad — clever ad copy routed to a generic homepage. Message-match collapses; landing page bears the bounce; campaign reads "didn't work" but it was the routing.
The persona-as-job-title — "For marketing leaders" used as the LinkedIn role tag. Too broad; algorithm can't optimize. Specific persona role + segment specifier ("Head of RevOps at Series B SaaS") outperforms by a wide margin.

Integration Notes

Persona & ICP Builder (outputs/personas/) — Persona's verbatim language, hidden motivation, shortlisting behavior, and named accounts feed the headline phrasing, the LinkedIn role-targeting, and the awareness-level classification. Reference the persona file by name in the ad set frontmatter.
Brand Voice Style Guide Generator — Voice attributes + AI prompt preamble + banned-phrase list are the system prompt for every AI-assisted variant. Don't re-author voice rules at the ad-copy level; cite the voice guide version.
Customer Review & Insight Miner — Verbatim-quote bank is the highest-converting source of headline phrases. Pain-point headlines and proof-point descriptions should pull literal customer phrases, not paraphrases.
Competitive Analysis Brief (outputs/competitive-analysis/) — Differentiator selection (which 2–3 to lead with) and contrarian-angle viability (when contrarian framing works without legal exposure) flow from the competitive analysis. If no recent analysis exists, flag it as a gap.
Synthetic Persona Simulator — Pre-test ad variants against the simulator persona before launch, especially for the contrarian angle and any high-stakes claim. Cheaper than a failed test.
Landing Page Conversion OS — The ad's promise must match the landing page's promise on the message-match axis. Run the landing page audit upstream of the ad set; copy and page should pass the message-match test together.
AEO Content Optimizer / AI Search Visibility Audit — When the campaign is running against a query where AI-engine answers are also visible, the ad copy needs to land before the AI-summary visual hierarchy on the SERP. Headlines that match the AI-cited claim's phrasing pick up "trusted-overlap" association.
Email Sequence Builder — Ad copy on the demo CTA flows directly into the first email of the post-conversion sequence. Voice and proof should be continuous; treat the ad → email transition as one document, not two.
Cross-Channel Attribution Analyzer — When measuring incrementality of the ad set, the holdout design from this skill ties directly into the attribution analyzer's read.
Campaign Performance Narrator — Performance reporting in the Narrator's audience-aware framing pulls back to the messaging angles defined in step 1 of this skill. Tag every variant with its angle so the post-campaign narrative can attribute lift to angle, not just to creative.
Multi-Channel Content Repurposer — When a high-performing organic post becomes paid creative, the Repurposer's top-performing angle and verbatim phrases become the seed for this skill. Pass the top organic version as the "existing high-performing copy" input.

Example Output

Input Recap

Campaign goal: Drive 25% lift in qualified demo bookings (Q3 2026 RevOps push)
Product: Threadline RevOps Diagnostic — free 30-min ops cycle time audit (lead-magnet → demo)
Audience: Persona Priya, the Pragmatic Head of RevOps (outputs/personas/priya-revops.md); Series B–C SaaS, 100–500 FTE; awareness level = solution-aware (knows the problem, doesn't yet know the named metric)
Differentiators: (1) named "ops cycle time" as the leading indicator, 21-day lead time empirically validated; (2) ScaleHQ case study (31% cut in 90 days); (3) measurable without a CDP
Platforms: Google RSA, Meta single image + carousel, LinkedIn Sponsored Content + Conversation Ad
Tone / angle: social-proof, contrarian, pain-point, educational (4 angles)
Existing high-performer: Q1 2026 LinkedIn post with verbatim phrase "the seam between Salesforce and everything else" — drove 6.2× the platform-average engagement
Compliance: FTC AI-disclosure on AI-visuals; SOC 2 badge required on landing page; legal-approved phrasing of the 3-week claim only; no competitor naming
Pairing context: Visual = Threadline brand-template chart; Landing page = /audit (recently audited via Landing Page Conversion OS, message-match score 4.2/5)
Voice preamble: outputs/voice/threadline-voice-preamble-v2.0.md
Confidence: high

Messaging Framework

Primary value proposition: Ops cycle time leads revenue plan attainment by 3 weeks — get the audit before your CRO asks the question
Awareness level: solution-aware (lead with proof + named metric + named outcome)
Messaging angles (4):
1. Pain point — "Your CRO is about to ask about a metric you might not be tracking"
2. Social proof — "ScaleHQ cut theirs 31% in 90 days"
3. Contrarian — "Sales cycle length is the wrong number"
4. Educational — "How to measure ops cycle time without a CDP"

Google Responsive Search Ads (RSA)

Headlines (15 — character counts in parentheses):

(Pain Point — pin pos 1) The Metric Your CRO Will Ask (28)
(Pain Point) Find the Forecast Risk Early (28)
(Value Prop — pin pos 1) Ops Cycle Time Leads Revenue (28)
(Value Prop — pin pos 1) The 3-Week Leading Indicator (28)
(Social Proof) ScaleHQ Cut Theirs 31% in 90 Days (30)
(Social Proof) Used by 47 RevOps Teams in 2026 (29)
(Social Proof) The Audit ScaleHQ Ran First (28)
(Differentiator — pin pos 2) No CDP Required to Measure (26)
(Differentiator — pin pos 2) Free 30-Min Ops Audit (21)
(CTA — pin pos 3) Run the Audit in 30 Minutes (27)
(CTA — pin pos 3) Book the 30-Min Audit (21)
(Pain Point) Pipeline Looks Fine. Forecast Won't (30)
(Contrarian) Sales Cycle Length Is the Wrong KPI (30)
(Educational) The 3-Week Lead on Revenue (26)
(Educational) What Ops Cycle Time Actually Predicts (30)

Descriptions (4 — max 90 characters):

(Proof + CTA, 87 ch) Ops cycle time leads revenue plan attainment by 3 weeks. Run the free 30-min audit.
(Differentiator + CTA, 86 ch) Measure ops cycle time without a CDP. ScaleHQ cut theirs 31%. Run the audit.
(Pain + Empathy, 85 ch) Your CRO is about to ask about this. Find the forecast risk before they do.
(Quantified, 88 ch) When ops cycle time worsens 1 day, revenue plan slips 4–6%. Lead time: 21 days.

Sitelinks (4):

"30-Min Ops Audit" → /audit (free diagnostic; SOC 2 verified)
"ScaleHQ Case Study" → /case-study/scalehq (31% cut in 90 days)
"Cycle Time Benchmarks" → /benchmarks (Series A–C+ healthy ranges)
"Measurement Without a CDP" → /blog/measure-without-cdp (the spreadsheet method)

Structured snippet: Services → Audit | Benchmark | Diagnostic | Implementation

Meta Ads (Single Image + Carousel)

Primary Text (3 — angle + char count):

(Pain Point — 124 ch) Your CRO is about to ask about a metric you might not be tracking. Sales cycle length is the lagging one. The leading one is called ops cycle time.
(Social Proof — 122 ch) ScaleHQ cut their ops cycle time 31% in 90 days. Then hit revenue plan attainment two quarters in a row. The audit takes 30 minutes.
(Contrarian — 119 ch) Most RevOps teams measure the wrong cycle time. The one that actually predicts revenue worsens 21 days before the forecast slips.

Headlines (3 — max 40 characters):

(Direct benefit, 35 ch) Find the Forecast Risk in 30 Minutes
(Curiosity, 38 ch) The Metric Your CRO Will Ask About Next
(Social proof, 30 ch) ScaleHQ Cut Theirs 31%

Descriptions (3 — max 30 characters):

(Differentiator, 25 ch) No CDP. No vendor calls.
(Urgency-restrained, 28 ch) Before the next QBR question
(Credential, 28 ch) SOC 2 Verified · n=47 Cohort

Format recommendations:

Single image: Pair with the Threadline brand-template chart (ops cycle time vs. revenue plan attainment, 21-day offset annotated). Best with Pain Point + Social Proof primary text.
Carousel (5 cards): Card 1 = hook ("Your CRO is about to ask…"); Card 2 = define ops cycle time; Card 3 = the 21-day lead time chart; Card 4 = ScaleHQ outcome; Card 5 = CTA "Run the 30-min audit."
Reel: Use the TikTok / Reel script from the Multi-Channel Repurposer output. Pair with primary text variant 3 (Contrarian).

LinkedIn Sponsored Content

Introductory Text (3 — max 150 characters):

(Role-targeted Pain Point, 144 ch) RevOps leaders at Series B SaaS: the metric your CRO will surface at the next QBR is probably the one you don't have a clean way to measure.
(Educational + Specific, 142 ch) Ops cycle time leads revenue plan attainment by ~3 weeks across our 47-customer cohort. Most teams aren't measuring it. The audit is 30 min.
(Contrarian + Insight, 138 ch) Sales cycle length is the lagging revenue indicator. Ops cycle time is the leading one — and you don't need a CDP to measure it.

Headlines (3 — max 70 characters):

(Direct outcome, 64 ch) Find the 21-Day Forecast Risk Before Your CRO Asks About It
(Social proof, 66 ch) The Metric ScaleHQ Cut 31% — Then Hit Plan Two Quarters in a Row
(Role-specific, 62 ch) The Free 30-Min Audit Built for Series B+ RevOps Leaders

Descriptions (3 — max 100 characters):

(Credential, 90 ch) Threadline 2026 benchmark, n=47 customer cohort. SOC 2 Type II. RevOps Co-op recommended.
(Specificity, 96 ch) Measurable without a CDP. Native Salesforce reports + 1 spreadsheet column gets the number.
(Outcome, 89 ch) ScaleHQ cut theirs 31% in 90 days; the part that didn't work is in the case study.

Conversation Ad addendum (optional — branched flow):

Opener: "Hi Priya — quick question for any Series B+ RevOps leader. If your CRO asked about ops cycle time at the next QBR, would you have the number ready?"
Branch A (Yes): "Strong. Would the 31%-improvement playbook from ScaleHQ be useful regardless?" → CTA: "Send me the case study"
Branch B (No / Not sure): "That's most teams. The 30-min audit pulls the number from your existing Salesforce data — no CDP required. Want to run it?" → CTA: "Run the 30-min audit"

Compliance + Brand-Safety Check

✓ FTC AI-disclosure footnote attached on any AI-generated visual (Meta carousel cards 3–4 use AI-augmented chart visuals)
✓ SOC 2 Type II language verified on landing page (/audit audit confirmed Apr 2026)
✓ "3-week leading indicator" claim uses legal-approved phrasing only ("Ops cycle time leads revenue plan attainment by ~3 weeks in B2B SaaS at 100–500 FTE")
✓ No competitor naming (Salesforce, HubSpot, Gong) per legal — including in Conversation Ad branched flow
✓ Banned-phrase check passed (no leverage / synergy / unlock / revolutionize / best-in-class)
✓ Meta personal-attribute rule — "you" framing reviewed; no protected-class implications
✓ LinkedIn verified-page status confirmed for Sponsored Content + Conversation Ads

A/B Testing Plan

Phase 1 — Angle Test (highest expected impact):

Test 4 angles (Pain Point / Social Proof / Contrarian / Educational) with controlled headline + description set
Sample size target: ~200 conversions per angle for 80% power detecting a 15% CPA delta — at $48K media spend and ~$190 LinkedIn CPL, this is the full Q3 budget across all four variants
Stop-loss rule: kill any variant below 50% of control CTR after 75 conversions
Primary metric: qualified demo bookings (not CTR)
Secondary metrics: CPL (target ≤ $190), CTR (target ≥ 1.4%), demo-page CVR (target ≥ 4.2%)
Holdout: 10% paid LinkedIn audience held out 6 weeks for incrementality

Phase 2 — Creative Test (within winning angle):

Test 3 visual pairings within the winning angle
Sample size target: ~150 conversions per variant
Stop-loss rule: same as Phase 1

Phase 3 — Headline-vs-CTA Test (within winning angle + creative):

Test 3 headline variants × 2 CTA variants
Run only after Phase 2 winner is conclusive

AI-Assist Layer

Google RSA underperforms (CTR < 2%): Refresh headlines 1, 5, 12 with the verbatim VoC phrase "the seam between Salesforce and everything else" — proven 6.2× engagement on the source LinkedIn organic post
Meta carousel CTR < 1.1%: Re-cut card 1 with a static screenshot of the Salesforce report (rather than the brand-template chart) — proof-first variant
LinkedIn CPL > $240 (current baseline): Tighten the role-targeting from "Series B+ RevOps" to "Head of RevOps + Director of RevOps at 100–500 FTE SaaS"
Conversation Ad open rate < 25%: Refresh opener to lead with the ScaleHQ outcome rather than the QBR question

Confidence

high — all nine required inputs present, persona named, voice preamble applied, verbatim VoC source available, compliance check completed, message-match validated against landing page audit, A/B testing plan with named sample-size logic.

Refresh Date

Re-rotate paid social variants every 14–28 days (creative fatigue)
Refresh Google RSA every 30–45 days
Re-validate compliance + claim language at next legal review (quarterly)
Re-validate voice preamble at the next Brand Voice Style Guide Generator refresh