How to find what's actually working in your competitors' ads

The Meta Ad Library, at facebook.com/ads/library. Every ad currently running on Facebook and Instagram is there, free, with no account required. Meta built it for political transparency and then extended it to all advertising, which means your competitors publish their entire active creative strategy whether they think about it or not.

Search by advertiser name rather than by keyword. Keyword search queries ad text and misses anything whose message lives in the image, which for good static ads is most of them. Pick the eight or ten brands you actually compete with and work through them one at a time.

One practical trap. If you search with the country filter set to the United States, creatives in sensitive categories such as health, supplements, and finance often come back login-gated and you see a grey box instead of an ad. The same ads are usually visible from a global vantage, because the gating is applied per country view rather than per ad. If a brand looks like it is hiding everything, widen the country filter before concluding anything.

Performance. For US commercial advertising the library shows no spend, no impressions, no click-through rate, and no conversions. You can see exactly what a brand is running and have no idea whether any of it works.

The exceptions are narrow and worth knowing. Political and issue ads carry spend ranges and impression bands. Ads delivered in the EU carry a total reach figure under the Digital Services Act. If your competitors advertise in Europe, that reach number is the only genuine performance metric available publicly, and it is worth checking for that reason alone.

This matters because a whole category of tools sells “winning ads” dashboards built on the same public data. They are not measuring performance either. They are inferring it, usually from the two signals below, and the good ones say so.

Longevity. How long has this ad been running? The library shows a start date. Advertisers are not sentimental: an ad that loses money gets switched off within a week or two. An ad still running after 60 or 90 days is being kept alive by somebody watching a dashboard, and that is the closest thing to a public performance signal you will get.

Variant count. How many versions of the same concept is the brand running? Meta groups near-identical ads, and you can see when a brand has produced eight crops of one idea versus one of another. Teams scale what works. A concept with many variants has usually earned the extra production effort.

Weight them together rather than separately. A three-month-old ad with a single variant might just be a forgotten evergreen. Twelve variants launched last Tuesday is a brand testing hard, not a brand that has won yet. The interesting quadrant is old and heavily varied, and it is usually a very short list per category.

Not the ad. The decision behind it. Copying a competitor's creative produces a worse version of their ad; understanding why it works produces yours.

For each ad that passes the longevity and variant filter, write down four things. Where the product sits in the frame and how much of it you see. What the first three words do, and whether they name a problem, a result, or a price. Whether the background is studio, lifestyle, or abstract, and how much of the image it occupies. And what the ad asks you to do at the end.

Do that for fifteen ads in your category and patterns appear quickly. Most categories converge on two or three formats that everyone successful is using, plus one outlier doing something genuinely different. Both are useful: the convention tells you the baseline, the outlier tells you where the room is.

Doing it manually is genuinely worth it once. Fifteen ads in your category, read properly, will teach you more about your market than any tool output, and you only need to do it once per category rather than once per campaign.

After that the work is repetitive, which is where software helps. Dovey does this half automatically: it pulls the active ads in your category, ranks them by exactly the two signals above, and reads each one for format, hook, and product placement. The difference from a spy tool is what happens next. Instead of handing you a dashboard to interpret, it uses that analysis as the brief for your own creative, so what it learned about your category shows up in what it makes for your brand. It is $1 for three days, and the ad research runs on your category from the first generation.