AI agent for copywriters
Headline Test Iteration Agent
Each headline test reaches a clear result and feeds a better next round
What it does
Copywriters launch headline and subject line tests, then rarely have time to read results properly or write the next round. This agent reads test results from the ad or email platform, waits until each variant has enough impressions, and checks whether the difference is likely real rather than noise. If a winner is clear, it names the angle that won, such as a number, a question or a benefit, and drafts three new variants that push that angle further. If no winner is clear, it either extends the test or proposes a bolder contrast. It checks the new drafts against the brand word list and character limits before you see them. You choose which variants go live. Edge case: a variant wins only because it ran on a holiday weekend, so the agent compares dates before calling it.
How it works
Follow the arrows from top to bottom. The orange dashed arrow is the loop: when a check fails, the agent goes back and tries again.
Read the steps as a list
- Daily check on running tests
- Pull impressions, clicks and conversions per variant
- Does each variant have the minimum sample?If not: leave the test running and check again tomorrow. Back to step 2.
- Test whether the gap between variants is significant
- Compare run dates and audiences to rule out timing effects
- Name the winning angle and draft three new variants
- Do new drafts meet character limits and the brand word list?If not: rewrite the failing drafts and recheck. Back to step 6.
- Copywriter picks variants to launchThe agent waits here for your OK.
- Log result and angle in the test library
- Test closed and next round ready
How it decides
A result is called when each variant has the minimum sample and the gap passes a significance check. The winning angle is the feature the winner has that the losers lack.
- Require at least 1,000 impressions or 2,000 sends per variant
- Call a winner at 95% confidence
- Extend a test once by 7 days if no winner; after that, propose a bolder contrast
- Ignore results where variants ran on different date ranges
Make it yours
Every agent is a starting point. You choose these settings for your own situation.
- Minimum sample per variant
- Confidence level (default 95%)
- Character limits per channel
- Brand word list
- Number of new variants per round (default 3)
What keeps you in control
It always asks you first
- Launching any new variant
Hard limits
- Never launch or pause campaigns
- Never call a result below the minimum sample
It stops when
- Done: winner called, next round approved
- Stop: two extensions with no winner, test marked inconclusive
Set it up
We guide you through the set-up, step by step
Members get the full set-up guide for this agent. No technical skills needed: you copy, paste and upload.
- One set of instructions to paste into your AI, with the clicks for ChatGPT, Claude, Microsoft 365 Copilot, Gemini and Grok
- The agent then walks you through connecting your own data, one source at a time
- A downloadable copy with the flow chart, the rules and the full guide
