A/B Test Runner
Set up a clean A/B test (one variable, proper split, a decision rule and duration for significance), read it honestly, and call the winner only when the data supports it.
Works across your stack
Supported AI platforms
8 supported clients. Connector requirements are shown separately below.
Add the ad testing skill to your AI client
Set up the Adspirer "A/B Test Runner" skill for me. If you can run commands: npx adspirer-skills add ad-testing If not: fetch https://www.adspirer.com/skills/ad-testing.md and follow it. Then read the skill and ask me what you need to get started.
Paste it into Claude, ChatGPT, Cursor, Codex, Gemini, or any agent — it works out how to install itself.
Or install it yourself
npx adspirer-skills add ad-testing Writes SKILL.md to .claude/skills/ in the current project. The bundled references, scripts, and assets ship with the ZIP. Download ZIP →
Upload the ZIP in Claude → Settings → Skills. The bundle preserves every reference, script, and asset.
Install the Adspirer plugin first: Cowork tab → Customize (left sidebar) → Plugins → Browse Plugins → search "Adspirer" → Install. Requires Claude Max, Team, or Enterprise. Then upload this ZIP as a skill.
In ChatGPT, open Plugins → Skills → Create → Upload from your computer. ChatGPT scans the skill before making it available. Availability and upload permissions depend on your workspace plan and admin settings.
npx adspirer-skills add ad-testing --agent cursor Writes the skill to .cursor/rules/ in the current workspace, so Cursor loads it on every session.
npx adspirer-skills add ad-testing --agent codex Appends the skill to AGENTS.md, which Codex reads on every session — so it persists instead of living in one conversation. The append is idempotent; re-running it will not duplicate the block.
npx adspirer-skills add ad-testing --agent gemini Appends the skill to .gemini/GEMINI.md, which the Gemini CLI reads on every session. Connect Adspirer as an MCP server first so the live ad-account steps can run.
npx adspirer-skills add ad-testing --agent windsurf Writes the skill to .windsurf/rules/ in the current workspace, so Windsurf loads it on every session.
npx adspirer-skills add ad-testing --agent other Writes SKILL.md to adspirer-skills/ in the current folder — point any agent at that file. No shell? Fetch the raw file with curl -fsSL https://www.adspirer.com/skills/ad-testing.md, which prints the skill instead of installing it.
Starter prompt
Client setup guide →Use the A/B Test Runner skill. I’m working on [campaign or account] and want to understand [goal or problem] for [date range]. Check that Adspirer is connected, start by looking without changing anything, and explain the story in plain language. Show me the evidence and ask before making any live change.Where it helps
When to use the ad testing skill
A plain-English test readout, the evidence behind it, and a clear recommendation to scale, keep testing, or stop.
Use case 1
A creative decision is stuck
Two concepts are spending against the same audience and the team needs to know whether the apparent winner is real or simply had a good few days.
Use case 2
A landing-page test needs a fair read
Traffic was split between two pages, but budget delivery or conversion volume may have made the comparison uneven.
Use case 3
A winner is ready to scale
The test has reached its agreed window and you want an evidence-backed recommendation before pausing the loser or moving budget.
See it in a campaign
Ad testing examples
The numbers, account, and recommendation will be yours. These show the kind of story the skill helps you work through.
Meta creative test
A skincare brand runs a creator testimonial against a polished product demo at $150 per day for 14 days.
The skill checks delivery and conversion volume, explains whether the CPA gap is meaningful, and asks before reallocating budget.
Google message test
A B2B campaign compares “book a demo” copy with a savings-led message while everything else stays the same.
The skill spots any setup differences, reads the agreed metric, and calls the result a win, a loss, or honestly inconclusive.
Keep reading: Meta Ads setup docs ·Claude ad management ·PPC campaign automation guide
Keep building the workflow
Agent skills related to ad testing
The Account Takeover Audit: A PPC Audit for Any Account You Inherit
When taking over an existing account — a new client handoff, an internal transfer, or an account you have just acquired — produce a full evidence-backed audit: structure, tracking health, wasted spend, quick wins, and risk, across Google Ads, Meta, and TikTok, before you spend another dollar.
The Budget Reallocator
Find capped winners and weak spenders across a campaign set, and propose moves sized to avoid resetting learning.
The Daily Pacing Check
Every morning, compare spend vs plan across every connected account and raise honest, evidence-backed alerts only when something needs a human. Read-only; hands off to the reallocator for changes.
Before you install
Frequently asked questions
What is the ad testing skill?
It is an Agent Skill — a folder of instructions your AI assistant loads when you ask it to do this job. It follows the open Agent Skills specification, so the same package works across different assistants. Set up a clean A/B test (one variable, proper split, a decision rule and duration for significance), read it honestly, and call the winner only when the data supports it. Unlike a prompt you paste in, the skill carries its own procedure, checks, and stop conditions, so the assistant works the same way every time you run it.
Which AI assistants can run it?
Claude, Claude Code, Claude Cowork, ChatGPT, Codex, Cursor, Gemini and Windsurf. Download the ZIP and upload it in assistants that accept skill folders, or point the assistant at the raw SKILL.md file. The install card above has the exact step for each one.
What do I need to connect before I run it?
Adspirer is required, because the skill reads live account data rather than guessing. The skill confirms the connection before it touches a real account, and tells you plainly if something is missing.
Do I need an Adspirer account, and is there a free tier?
The skill itself is free to download and MIT-licensed. To let it read or change real ad accounts you need Adspirer connected: the free tier includes 15 tool calls a month, Plus is $49/mo for 150, Pro is $99/mo for 600, and Max is $199/mo for 3,000. You can install the skill and read what it does before signing up for anything.
Can it change my campaigns without asking?
No. Any new campaign Adspirer creates is created paused, enabling a paused or live campaign needs your confirmation, and destructive actions are gated behind an explicit confirmation flag. Spend, publishing, and live changes wait for your approval every time — installing a skill or a connector is not itself approval to act.
How do I get the best results from it?
Give it the specifics it asks for up front — the account or campaign, the date range, the goal, and what "good" looks like for this business. Let the first pass run read-only so you can check the evidence before anything changes. The starter prompt above is a working template; edit the bracketed parts and paste it in.
How is this different from just asking the assistant directly?
A plain prompt gets you a plausible answer from whatever the model already knows. This skill makes the assistant pull real numbers from your ad accounts, follow a fixed procedure, show its evidence, and stop at the points where a person should decide. That is the difference between a confident guess and a ad testing you can act on.
Ready to run it against real campaigns? Connect Google, Meta, Amazon, LinkedIn, TikTok, or ChatGPT Ads to your assistant on the free tier — nothing spends or publishes without your approval. Create a free Adspirer account →