Growth
Paid Ads Copy Variant Loop
Ads loop: generate copy variants, spend enough for signal, kill losers, scale winners.
Use when
Paid search or social ads are running and performance is flat, and you need a disciplined cut-losers / double-winners loop without letting the agent freely blow budget.
Intermediatedifficulty
Growthcategory
Greg Isenberg + Ellie interview: Making $$$ with Loop Engineering — ads agency-style experiment loop; AI stronger on copy variants than video creatives; humans as optional creative APIsource
Cadence
Every 2–3 days while spend is live; weekly when spend is low
Verification
Each active variant has enough spend or impressions for a decision rule, losers are paused, winners keep or increase budget share, and every copy/creative change is logged with the metric that justified it.
Structured loop spec
| Field | Value |
|---|---|
| Name | Paid Ads Copy Variant Loop |
| Category | Growth |
| Trigger | Every 2–3 days while spend is live; weekly when spend is low |
| Objective | Ads loop: generate copy variants, spend enough for signal, kill losers, scale winners. |
| Allowed inputs | Relevant files, source notes, logs, tests, screenshots, metrics, or task state for this loop |
| Allowed actions | Define the KPI (CPA, ROAS, CTR, or qualified leads) and a minimum spend/impression threshold before a variant can be judged.; Inventory live ads: creative type, copy, audience, budget, and last 7–14 day metrics. Mark approval boundaries (max daily spend, no new ad accounts, no billing changes).; Propose 2–5 copy-first variants (or drop human-made assets into a folder). Prefer text/headline tests over unvetted AI video unless a human supplies the asset.; Launch or schedule only within the spend cap. One change family per round (copy angle, not copy + audience + bid all at once).; On the next tick, apply the decision rule: pause below-threshold losers, keep/scale winners, log the result. Do not invent metrics.; Human-gate any budget increase, new campaign creation, pixel/account changes, or public brand claims. Stop when KPI target is met, spend cap hits, or two full cycles produce no winner. |
| Verification | Each active variant has enough spend or impressions for a decision rule, losers are paused, winners keep or increase budget share, and every copy/creative change is logged with the metric that justified it. |
| Stop condition | Stop when the verifier passes, the budget is exhausted, no progress is made, a blocker appears, or approval is required. |
| Budget | Set a time, turn, token, retry, file, or dollar cap before running the loop. |
| Approval boundary | Human approval required before publishing, sending, deleting, spending, changing accounts, touching production, or making reputational/legal/financial commitments. |
| Safe output | Draft, report, checklist, table, or approval-gated recommendation |
| Works with | Claude, ChatGPT, Gemini, any tool-using AI assistant |
Steps
- Define the KPI (CPA, ROAS, CTR, or qualified leads) and a minimum spend/impression threshold before a variant can be judged.
- Inventory live ads: creative type, copy, audience, budget, and last 7–14 day metrics. Mark approval boundaries (max daily spend, no new ad accounts, no billing changes).
- Propose 2–5 copy-first variants (or drop human-made assets into a folder). Prefer text/headline tests over unvetted AI video unless a human supplies the asset.
- Launch or schedule only within the spend cap. One change family per round (copy angle, not copy + audience + bid all at once).
- On the next tick, apply the decision rule: pause below-threshold losers, keep/scale winners, log the result. Do not invent metrics.
- Human-gate any budget increase, new campaign creation, pixel/account changes, or public brand claims. Stop when KPI target is met, spend cap hits, or two full cycles produce no winner.
Prompt
Run the Paid Ads Copy Variant Loop. KPI and spend caps must be explicit before any change. Read current ad set performance. Generate or select a small batch of copy-first variants (humans may supply images/video). Launch only inside the approved budget and one change family per round. After each review window, pause losers that cleared the minimum-spend threshold and failed the KPI, keep or scale winners, and write the decision into an experiment log. Never change billing, ad accounts, pixels, or daily spend ceilings without human approval. Stop on KPI hit, budget exhaustion, no signal after two cycles, or any red-zone account action.Run in Claude Code
Paste this into Claude Code (or any tool-using agent) to run the loop bounded: one change per round, the same verification every round, durable state files, and explicit stop conditions.
Run the "Paid Ads Copy Variant Loop" loop from AI Loop Library (https://ailooplibrary.com/loops/paid-ads-copy-variant-loop/) as a bounded loop.
Goal: Ads loop: generate copy variants, spend enough for signal, kill losers, scale winners.
Rules: one change per round; run the same verification every round (Each active variant has enough spend or impressions for a decision rule, losers are paused, winners keep or increase budget share, and every copy/creative change is logged with the metric that justified it.); append each round to docs/loops/paid-ads-copy-variant-loop/progress.md and update docs/loops/paid-ads-copy-variant-loop/state.json; stop on verifier pass, 8 rounds, 3 consecutive failed verifications, no progress, a blocker, or anything needing human approval (money, production, outbound, deletion). Finish with a proof report: rounds used, changes made, verification output, remaining risk, and the next human decision.