Prompt library
Copy-paste prompts that tell Claude or ChatGPT what ads to make this week, and why. Each one is built on a real experiment Meta advertisers run every week.
11 prompts · Free to copy
Spot a customer group you aren't targeting and build a full funnel for them.
Look at the last 30 days of customer reviews and post-purchase surveys. Is anyone describing themselves as part of a group we're not already targeting or testing? Pick the most promising one and make 10 TOF, 10 MOF and 10 BOF ads for them, keeping the product the same.
See who competitors' long-running ads call out, and test that persona yourself.
Look at competitor ads in the Meta Ad Library that have been running for 30+ days. Who are they calling out that we aren't? Pick the most promising persona and make 10 TOF, 10 MOF and 10 BOF ads for them, keeping the product the same.
A prompt is only as good as the data behind it. These prompts ask for ad performance, reviews, comments and your catalog, so the setup that matters is getting that data in front of the model.
Create a project (both Claude and ChatGPT have them) and add your brand basics: what you sell, who buys it, your voice, your best ads. Every prompt you run in that project starts from that context instead of from zero.
Both tools can connect to outside apps. Connect the ones your prompt needs: Shopify for your catalog, Meta Ads for performance, and your reviews or helpdesk app for reviews and tickets. If a connector isn't available, export the data as a CSV and attach it.
Ask it to show the ads, reviews or products it picked before it writes anything. A wrong pick early on means a whole batch of wrong ads.
You'll get angles, headlines and creative direction. Build them in your design tool, or have an image model render them in 1:1, 4:5 or 9:16 for Meta placements.
Most AI ads look random because the prompt asks for "some ads". A good prompt names what changes, what stays the same, and what data it should decide from.
Product, persona, angle and format. Each prompt changes one and keeps the rest the same, so when an ad wins you know why.
Exploit wins more of the market you're in. Explore finds new markets, and a new market is really just a new persona. Seasonal captures moment-in-time spikes.
Performance data tells you what's working, and it feeds exploit. Customer data (reviews, surveys, comments) tells you what to try next, and it feeds explore.
You probably launch ads in weekly batches anyway. Each prompt looks at a different slice of data on batch day, which is why they all run on Mondays by default.
A set of instructions you paste into an AI tool like Claude or ChatGPT so it decides what ads to make and writes them. A good one names the data to look at, the one thing to change, and what to keep the same.
Yes, with the right data. Both can pick angles, write headlines and draft creative direction from your reviews, comments and performance. Connect your store and ad account, or attach exports, so they work from your numbers instead of guesses.
Most brands launch in weekly batches. Refresh an ad when its frequency rises and CTR drops, which the fatigue refresh prompt checks for every week.
Exploit ads win more of the market you already sell to: new versions, new angles and new formats for your current personas. Explore ads test a new persona, which is how you find a new market.
They work best with them. Without connectors the model will ask you to paste the data, so have your ad performance, reviews and catalog exports ready.
Yes. All 11 prompts are free to copy, change and share.
Yes. In StaticAds.ai any prompt here can become an Automation that runs on your schedule, Mondays at 5:00 AM by default, and builds the ads for you.
StaticAds.ai runs any of these prompts on a schedule with your catalog, swipe file and brand, and sends the new ads to Slack when they're ready.