In short
Ad Radar is a research tool for Meta ads. It shows only ads that are already winning, with $50k or more in estimated spend, and tags each one on ten dimensions, from hook to landing page. The blog is written by the Ad Radar Research Team from the same board. Questions and corrections go to hello@ad-radar.dev.
What Ad Radar is
Ad Radar is a competitor research tool for Meta ads, which means ads on Facebook and Instagram. It follows the ads of DTC and e-commerce brands in the public Meta Ad Library, records how each ad performs over time and keeps only the ones that are already winning: ads with $50k or more in estimated spend.
Each ad on the board comes with its estimated spend, how long it has been live, how many variants the brand launched, the full copy, the video transcript and the landing page it sends to. Members search and filter the board in the app, in the Chrome extension, or inside Claude through Ad Radar's MCP server.
It does less than a general ad library, on purpose. It doesn't try to show every ad. It shows the ones a brand kept paying for.
What the board covers
The research on the blog cites a fixed snapshot of the board, so the numbers in different articles add up. The current snapshot is from October 9, 2026:
2,443
live Meta ads, each with $50K+ in estimated spend
749
brands behind them
$223K
median estimated spend
10
dimensions every ad is tagged on
The live board in the app keeps moving every day, as new ads cross the line and old ones stop.
How every ad is tagged
Every ad is tagged the same way, so ads can be counted and compared instead of scrolled. The dimensions the research uses most:
| Dimension | What it captures |
|---|---|
| Hook | The opening line or first seconds: its form (question, number, story, warning) and why it grabs attention |
| Angle | The line of argument the ad builds: root cause, systematic breakdown, story, expert-based and four more |
| Format | How the ad was made: UGC selfie, founder or expert, footage with voiceover, product demo |
| Awareness | What the opening assumes the viewer knows, on Eugene Schwartz's five levels |
| Landing page | Where the click goes: product page, dedicated landing page, listicle, quiz, advertorial |
The other five are hook mechanism, proof level, video type, niche and CTA type. Definitions and examples for all ten are in the methodology.
Who writes the research
The blog is written by the Ad Radar Research Team: the people who build the board, write the tag definitions and check the data. Articles are signed by the team rather than by one person, because each one draws on the same board and the same rules.
How we work:
- Every figure comes from the board. Each article names the snapshot it uses, and counts, shares and medians are computed from it, not quoted from elsewhere.
- Spend is always "estimated". Meta doesn't publish spend for commercial ads, so every spend figure is a model, and we say so every time. How it is built is in the methodology.
- Examples are real ads. Ads quoted in articles ran on Meta and were on the board on the snapshot date, with their estimated spend and days live.
- No paid placements. No brand or tool pays to appear in an article. Comparison pages list competitor prices with the date they were checked.
- Limits are stated. Small samples are flagged as small, and patterns are described as patterns, not causes.
Where to start
New to the research? Start with Facebook ad examples, Facebook ad hooks and marketing angles, or browse the blog by topic.
Contact
Questions, corrections and press requests: hello@ad-radar.dev. If a number in an article looks wrong, tell us which one, and we will check it against the snapshot and fix the article.