MetaMiner

Competitor ad analysis, when it happens at all, happens by hand. Someone scrolls the Ad Library on a Friday, eyeballs a few rivals, and jots down what looks interesting. By the next time anyone looks, the market has moved and the notes are stale.

The signal that matters most is invisible to the naked eye: ad spend. Anyone can launch a creative. The ones that stay live for weeks are the ones that convert, because nobody keeps paying to run an ad that does not. Longevity is the vote. But you cannot see longevity by glancing at a feed; you have to track every ad over time.

And the hook is everything. In direct response, the first line of the ad is engineered to stop the scroll. Miss which hooks your competitors are scaling and you are guessing at the single decision that drives the whole funnel.

MetaMiner runs a headless miner against Meta's Ad Library on a daily schedule. It pulls the top active ads per tracked brand, sorted by longest-running, and grabs the hook copy, creative variants, and started-on dates. Because the Ad Library is public, this needs no special access: it sees exactly what the regulator and the competition see.

Raw hooks pass through deterministic pattern detection: lexical template matching first, then clustering of whatever does not match a known template. No black-box embeddings, because the same input must produce the same signature on every run. A pattern is a reproducible thing, not a vibe.

Each pattern carries a lifecycle. New patterns are flagged the run they appear. Converging patterns gain momentum as more brands adopt them. A pattern needs two consecutive misses before it retires, so one noisy scrape cannot kill a real signal. A signature retired and revived again and again gets tagged burned.

The output is a daily field briefing: which patterns moved, which brands are at risk, which new angles are maturing toward a template, and the actual hook copy behind each. Read it in two minutes, act the same morning.