The problem this solves
Your competitors publish their entire creative strategy in public ad libraries, and almost nobody on your team has time to read it. Checking one competitor on one platform is a coffee-break task; checking five across Meta, LinkedIn, and TikTok, capturing what they run, and organizing it into something usable is a day of clicking that never survives contact with the sprint board. So creative planning proceeds on memory and screenshots someone took in March.
The waste shows up in testing. Teams burn budget rediscovering angles their competitors already validated at scale, and miss the offers the whole category is converging on. A maintained swipe file inverts that: competitor spend becomes your market research, long-running creatives reveal what sustains performance, and your tests start from the frontier instead of from scratch. The file just never gets built by hand.
How the mission runs
- Sweep the ad libraries. Ad Libraries pulls the current active ads for your five named competitors across Meta, LinkedIn, and TikTok: creative, copy, formats, and observable signals like how long each ad has been running. The raw feed of their strategy, collected in minutes.
- Group by angle and offer. The agent classifies the haul: which persuasion angles each competitor leans on, which offers they push, and which formats carry the spend. Patterns emerge that no single ad shows - who is testing broadly, who has converged, and where the category herd is heading.
- Flag the proven runners. Ads that have run for extended periods get flagged as likely workhorses, since sustained spend on a creative is the strongest public signal it performs. These are the structures worth studying closely rather than the experiments that will vanish next week.
- Build the Notion swipe file. Everything lands in Notion as a browsable database: each ad with its creative, copy, platform, angle, and offer tags, organized so your team can filter by competitor, angle, or format when planning the next test cycle.
- Prioritize what to test first. The file closes with recommendations: the angles validated by competitor persistence that you have never tested, the gaps nobody in the category is exploiting, and a suggested first test slate - each tied to the evidence in the file.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
A Notion swipe file cataloging your five competitors' active ads across Meta, LinkedIn, and TikTok - every ad tagged by angle, offer, format, and platform, with long-running workhorses flagged - closed by a what-to-test-first shortlist grounded in the patterns. Your next creative planning session starts from the category's live evidence instead of from a blank whiteboard.
Make it yours
- Track a different roster: swap in emerging challengers or an adjacent category whose playbook you suspect transfers to yours.
- Narrow to one platform: a Meta-only deep file with more competitors covered, when that is where your budget concentrates.
- Make it a monitor: rerun monthly and have each update highlight what changed - new angles entering, old workhorses retired.
Frequently asked questions
Is collecting competitor ads like this allowed?
Yes. Ad libraries are public transparency surfaces the platforms themselves operate, and reading them is exactly what they exist for. The mission gathers only what any person could browse manually - its value is doing so systematically across five competitors and three platforms at once.
How do we know which competitor ads actually work?
Public libraries show what runs rather than what converts, so the mission uses longevity as the honest proxy: creatives sustained for months are very likely earning their spend. The file labels this as inference, and the test recommendations treat competitor evidence as hypotheses for your own validation.
Should we copy the best-performing angles directly?
The file is built for adaptation. Copying a competitor's ad wholesale wastes the chance to differentiate; the recommendations identify the validated underlying angle, then suggest how to run it through your own positioning and offer so you inherit the insight without the sameness.