Where the ads come from
Public ad libraries. Platforms are required to publish what is running, so the raw material is free and complete in a way almost no other competitor data is: this is their own site's worth of honesty, not a sample of a platform we had to guess at.
Two things follow. First, the count should come from the library's own counter rather than from however many a tool managed to download. One brand we track runs 5,796 live ads while we read fifty of them in detail, and a tool that reports fifty as their ad count is reporting its own settings as though it were their spending. Second, a library shows what is live now, not what was killed last month, so every lifespan you read is a floor rather than a measurement.
Longevity is the only honest scoreboard
You cannot see their click-through, their cost per acquisition or their return. Nobody outside the account can. But nobody keeps paying to serve an ad that loses, so time live is the one outcome signal that leaks out, and it is the one most swipe files throw away when they save the image.
Read it comparatively rather than absolutely. Within one brand, in one week, on one budget, the formats that survive longest are the formats that are working, and the differences are usually large enough to act on. That is a conclusion they paid for and you did not.
Count ideas, not sentences
A hook that appears once is a test. A hook that appears dozens of times is a conclusion. But counting identical sentences overstates how few ideas there are, because near-identical phrasings get counted separately and a long tail of one-offs looks like variety.
Group by overlap instead, so that two rewordings of the same idea count once, and then read the concentration: what share of the creative sits in the largest few families. High concentration is a company that has found its message and is scaling it. A flat spread is a company still looking, which is a different competitor to face and a different one to copy.
The half nobody checks
Creative gets the attention and the destination decides the outcome. Where the ads land, split by type, is often where the copyable idea actually is: a dedicated landing page per campaign, or every ad dumped onto the homepage, or a quiz, or a collection.
Offer share belongs next to it. A brand whose creative rarely carries a discount usually has somewhere cheaper to send a hesitant buyer, and the shape of their price ladder tells you whether that is true. Rising offer share is a company buying volume with margin, and that is a position you can outlast rather than match.
What this does not do
No spend, no impressions, no results. Platforms do not report spend for commercial ads. A range can be modelled from live creative and a cost band, and anchored in real reach where EU transparency data exists, but it stays labelled an estimate. A tool showing a precise spend figure has invented it.
No archive of what was killed. Short-lived creative is missing from the record unless somebody was watching while it ran, which is the argument for a weekly reading rather than a one-off look.
How to do it here
Ads are pulled weekly, tagged by format, offer, hook family, destination and days running, and the pattern lands on Details with the week over week movement on Changes. Every figure opens into the creative it was counted from, because a claim about somebody's advertising that you cannot check is just an opinion with a percentage sign.
The full description of the source and its limits is on the ads page.
Pick one competitor. See what we find.
Your first benchmark is free. One domain, measured against you, no card and no call.