The tracker columns
Download the CSV and open it in Sheets or Numbers. One row per ad, one sheet per month.
| Column | What to write |
|---|---|
| date_observed, source | When you looked and which library (so you can re-find it). |
| brand_or_app, app_category, platform | Advertiser, category, the placement you saw (Reels, TikTok, YouTube, Search). |
| library_url, creative_url | Links back to the ad in the library; never a downloaded copy. |
| format | screen_demo, ugc_talking_head, hands_and_phone, before_after, brand_film, static. |
| hook_type, hook_text | One of the five types below, and the literal words or visual of the first two seconds. |
| offer_or_claim, cta_text | Free trial, discount, "no ads", a result claim; the CTA wording. |
| length_seconds | Measured, not guessed. |
| first_seen, last_seen, variants_seen | The dates you observed it in the library and how many near-identical versions exist. These three columns are the performance proxy. |
| notes | What you would change if it were yours. |
Five hook types
- Problem statement: "I kept forgetting to log meals." Works for utility and health apps.
- Question: "Is this the easiest way to edit photos?" Works for editing and consumer apps (see the Picsart example).
- Objection first: "No other technical identifiers." Works where trust blocks installs (see Flo).
- Outcome on screen: the finished screen in frame one, then how (see Opal, MyFitnessPal).
- Number or countdown: "30 seconds to detonation", "1 lb per week". Works everywhere; numbers are read before words.
Signals to study, without performance data
None of these prove an ad performs; libraries do not publish spend, CPA or ROAS. They tell you which ads are worth studying first.
- Longevity. An ad active for more than 30 days is probably not losing badly; after 90 days it is likely a core creative. Study its structure first, but a long run can also come from a big budget or a campaign nobody switched off.
- Variant batches. Five ads with the same start date and one changed element (hook text, first frame, CTA) are a test. Revisit in two weeks; the one still running hints at which element they kept.
- Cross-platform reuse. The same cut on Meta, TikTok and YouTube suggests they trust it across placements; one-platform ads may be placement-specific experiments.
- Reappearance. An ad that stops and restarts a month later with a new offer is their seasonal lever.
From findings to tests
Each week, pick the single structure with the strongest longevity signal that you have not tried, and build two variants of it on your own footage: same structure, different hook text. Run both for a week with equal budget. Log the result next to the competitor row so the sheet becomes your own evidence, not just theirs.
Mistakes we made
- Reading one striking ad as a strategy. It was a one-week test that died.
- Logging "first_seen" as the library's start date instead of the date we first observed it; it made old ads look fresh.
- Tracking twelve competitors and acting on none. Five, weekly, with one test, beat it.
- Placing copied captions at the bottom of the frame, where TikTok covered them. Hence the safe-zone checker.
Questions
How many competitors should I track?
Five is enough to see patterns and few enough to keep weekly. Two larger than you, two peers, one from an adjacent category whose users overlap with yours.
How often?
Weekly, same day, 20 minutes. Monthly reviews miss the test batches that reveal what a competitor is learning.
Do I need a paid spy tool?
Not to start. The official libraries plus this tracker cover Meta, TikTok and Google. Paid tools add history, estimated spend and networks without libraries; buy one when the weekly routine proves you act on the data.
Is copying a competitor's ad structure allowed?
Structure (hook type, beat order, caption placement) is not protected; footage, music, logos and distinctive copy are. Rebuild the structure on your own recording.
Sources
- Competitor ad tracker (CSV, free) — the columns described below with one sample row
- Meta Ad Library
- TikTok Creative Center: Top Ads
- Google Ads Transparency Center
Checked 2026-10-28. Platform rules change; the link is the authority, this page is the summary.