The Meta Ad Library shows you every ad a page is currently running, when each one started, whether it is still active, which platforms it runs on and how many versions it has. It does not show spend, impressions, click-through rate or targeting for ordinary commercial ads: those numbers exist only for political and social-issue advertising, and every tool that quotes a competitor's spend on a shampoo ad is modelling it, not reading it. Which leaves you with one real performance proxy, longevity, and it is more useful than it sounds.
What the library publishes, and what it does not
| Signal | Available for commercial ads | The catch |
|---|---|---|
| The creative itself, video and copy | Yes | Only what is live now, unless it is a political ad kept in the archive |
| Start date, the "started running on" stamp | Yes | It is the date the ad went live, not the date the concept was first tested |
| Active or inactive status | Yes | An ad can be paused and relaunched, which restarts the clock |
| Number of versions inside one ad | Yes | Tells you they are running variations, not which one gets the budget |
| Platforms and placements | Yes | Facebook, Instagram, Messenger, Audience Network flags only |
| Advertiser page details | Yes | Page transparency shows name changes and page age |
| Spend and impression ranges | Political and social-issue only | Not available for retail, DTC, SaaS or any normal advertiser |
| Targeting and reach breakdowns | Political ads, plus limited EU transparency data | Not the audience settings, and not comparable across regions |
| CTR, CVR, ROAS, hook rate | No | No platform publishes competitor performance, and no tool can see it |
| Budget split across their variations | No | Frequently implied by third-party tools, never published |
Two habits follow from that table. Filter properly before you conclude anything: country, platform, active status and date range change the result set enough that two people can look at the same brand and disagree about what it is running. And record the start date at the moment you look, because it is the only number in the library that carries information and it changes while you watch.
Why longevity is the only performance proxy you get
An ad that has been live for eight weeks is the strongest public signal you will find that something is working, and the reason is arithmetic rather than faith. Our 2026 fatigue benchmark, a synthesis of published platform and agency figures, puts CTR decline at 15 to 20 percent in a creative's first two weeks, with week three landing 45 to 70 percent below the launch baseline and most creative effectively dead inside three weeks. A competent team kills a losing ad well before that. So an ad still running past six weeks has either survived the decay curve or is being kept alive by someone not watching, and the first explanation is more common among advertisers who spend seriously.
How to use that without over-reading it:
- Sort by start date, not by what catches your eye. The visually impressive ads in any library are usually the newest, because a fresh brand film is what a quarterly process produces.
- Compare inside a vertical, since decay speed varies enormously. Food and beverage reaches a 40 percent CTR decline in about nine days and B2B SaaS in about 28, so six weeks means something very different in each.
- Treat repeats as a stronger signal than duration. The same hook rebuilt three times over four months is a team that tested it and kept winning.
- Discount anything running in one country only, which is often a test rather than a proven winner.
The honest limit: longevity tells you an ad is working for that advertiser's offer, funnel and price point. It does not transfer. Plenty of teams have rebuilt a competitor's six-month winner and watched it die in a week because the offer behind it was different.
Spend numbers, and who is allowed to publish them
Meta publishes spend ranges and impression ranges only for ads about elections, politics and social issues, where a legal transparency requirement applies. For a commercial advertiser there is no published spend figure, no impression count and no per-ad budget.
Third-party trackers that show you a competitor's monthly spend are estimating from whatever they can observe, usually ad counts, run duration and category assumptions. Those estimates can be directionally useful for judging whether a competitor is scaling or retreating. They are not measurement, and we would not put one in a board deck without saying so. If a tool cannot tell you how a number was derived, treat it as a guess with a chart around it.
The five-step loop from library to brief
Most competitor research fails at the last step. People collect 40 screenshots, feel informed, and produce nothing. This is the loop we use, and the output is a brief someone can shoot or build against.
One, write the question first. Not "what is Competitor X doing" but something answerable: "which hook types are surviving past four weeks for direct-to-consumer supplement brands in Germany". A question with a filter in it is a question you can close.
Two, sample deliberately. Six to ten advertisers, the same country and platform filters for all of them, everything currently active plus start dates. Include one adjacent category, because the hook that is about to arrive in your vertical usually shows up next door first.
Three, code what you see. One row per ad in a sheet: advertiser, start date, days live, hook type, opening shot, offer type, whether there are captions, whether the offer lands before the halfway point, format. Coding is the step everyone skips and it is where the value is, because it turns impressions into counts you can sort.
Four, sort by days live and read the pattern. You are looking for hook types that appear repeatedly among the long-running ads and are absent from the short-lived ones. Three or four patterns usually stand out, and one of them is normally something your team has been avoiding.
Five, write the brief. One page: the hypothesis in a sentence, the hook line, the proof the footage must show, the shot list, the offer position in seconds, and the placements. Then ship five to eight variations of it rather than one, because a single execution of a good hypothesis fails often enough to be indistinguishable from a bad hypothesis.
Our notes on running this at scale live on the ad library research page, and the Facebook Ad Library tool page covers the search and filtering side.
What to do with the brief
Disclosure: Genyad is our product, and step five is where it fits. It works from a library of footage you already own, transcribes and tags every clip, then builds each variation as its own script, shot selection, voiceover, caption set and export rather than a re-cut of one timeline. So a brief that says "open on the problem, offer by second five, three hooks" comes back as several genuinely different builds from your existing footage, exported in 9:16, 4:5, 1:1 and 16:9. The Meta ad creative generator page shows the flow, and the free plan is five variations without a card.
Being clear about the boundaries: it does not import a competitor's ad and rebuild it, and it cannot use footage you do not have rights to. There are no AI avatars, no static banner formats, no product-URL import, no predicted performance scores and no publishing to Meta. Copying a competitor's creative frame for frame is also a poor strategy, since you would be launching a worn-out concept into an audience that has already seen it several times, and Meta's internal research puts the CTR drop at 45 percent after a fourth exposure to the same creative.
Frequently asked questions
Can you see how much a competitor spends in the Meta Ad Library?
Only for political and social-issue ads, where Meta publishes spend and impression ranges under transparency rules. Commercial advertisers have no published spend, impressions or budget split. Any tool showing you a retail brand's monthly Facebook spend has modelled that figure from ad counts and run duration.
How do you tell which competitor ads are performing well?
Use how long they have been running. Most creative is effectively dead within three weeks according to our benchmark report, so an ad still live at six or eight weeks is the strongest public signal available that it is working. Compare within a vertical, because decay speed runs from about nine days in food and beverage to about 28 in B2B SaaS.
What does the Meta Ad Library actually show for a normal brand?
The live creative, the start date, active status, the platforms it runs on, how many versions the ad has, and page transparency details for the advertiser. That is genuinely useful for creative research and useless for performance analysis. Set your country, platform and date filters before drawing conclusions, since the unfiltered view mixes markets.
How do you turn ad library research into something usable?
Code the ads into a spreadsheet rather than collecting screenshots: advertiser, start date, days live, hook type, offer position, format. Sort by days live and look for hook types that recur among survivors and are missing from short-lived ads. Then write a one-page brief with the hypothesis, hook, required proof and shot list, and produce five to eight variations of it.