Playbook
Everyone knows how to find ads. The Meta Ad Library and a dozen spy tools will bury you in them. The hard part — the part that actually costs money when you get it wrong — is telling which of those ads is a real winner worth copying. Here's the old way, and the way that just became possible.
You open a spy tool, search your niche, and stare at a wall of ads. You sort by engagement. You look for the ones that have run longest, because "if it's still running, it must convert." You save a dozen to a swipe file, model two, spend a week and a few thousand dollars — and find out one of them was an autopilot dud the whole time. The database never lied to you. It just never had an opinion.
A long-running ad is the most over-trusted signal in the game. Ads coast on budgets nobody reviews. Brands run creative for awareness, not response. Real winners get rotated out early to beat fatigue. Run time is a hint — and you've been treating it like proof (the full unpacking: how to tell if a competitor ad is working).
Here's what changed. Connect the PreTestAds MCP to an AI agent and say "spy on this advertiser's Facebook ads." The agent opens the public Meta Ad Library, pulls every live ad, and scores each one for predicted attention — hook strength, the second it drops, whether it holds through the CTA. Then it crosses that score with how hard the advertiser is paying to run it, and hands you a one-word verdict on every ad:
Every competitor ad, sorted into four buckets
HIDDEN GEM
Great creative nobody's backing — steal it early
MODEL IT
Proven winner — copy the structure
SKIP
Nothing to learn or beat
OUT-HOOK IT
They're committed to weak creative — beat it
Creative attention score →
No scrolling. No guessing. No week of spend to learn what a score told you in minutes.
Finding the winner is half of it. Build your version, score it, and run it head-to-head against the competitor's — second by second — so you launch knowing you out-hook the ad you learned from. That's the whole loop, and you can watch it on 7 real competitor ads we scored (every one had a broken hook — the openings were the opportunity).
Score any competitor ad for attention in minutes. First analysis free.
Score an AdStart in the Meta Ad Library or a spy tool to see what competitors run, then judge the creative itself — because run time and engagement are weak proxies for whether an ad works. The reliable method is to score each ad for predicted attention: hook strength, where it drops, and how it holds through the CTA. The strong ones are your models; the weak long-runners are ads to out-hook.
You can't from run time alone — plenty of mediocre ads coast on autopilot budget for months. Cross the creative's attention score with how hard the advertiser is paying to run it: high score plus high conviction is a real winner to model; high conviction but low score means they're stuck with weak creative you can beat.
Yes. With the PreTestAds MCP and a browser-capable agent, you can say 'spy on this advertiser's Facebook ads' and the AI opens the Meta Ad Library, scores every live ad for attention, and returns a verdict on each — model, out-hook, hidden gem, or skip — so you skip the manual scrolling and guessing.