How to read a competitor's ad library in ten minutes

Ad libraries are public. Most people open one, scroll for a minute and leave. Here is what to look at instead.

Where ad libraries live

Meta, TikTok, Google and LinkedIn all publish public libraries of the ads they run. Most were built for transparency, not competitor research. They show you what ran and leave the reading to you.

• Meta Ad Library. Search the active ads any brand is running across Facebook, Instagram and Meta's other apps. Each ad shows when it started running, and near-identical versions are grouped together.
• TikTok. The Commercial Content Library covers ads shown in the EEA, Switzerland and the UK. Top Ads, inside TikTok's Creative Center, is a different thing: a curated set of high-performing ads that advertisers have agreed to feature, filtered by region and industry.
• Google Ads Transparency Center. The ads an advertiser has run across Search, YouTube and Display, the regions they were shown in, their format and the last date each one ran.
• LinkedIn Ad Library. Search by advertiser, keyword, country or date range. Ads stay listed for a year after their last impression.

None of them tell you what an ad earned, and for ordinary commercial ads you will not see clicks or sales either. You are reading the brand's decisions, not its results. The rest of this guide is about reading those decisions well.

One detail matters more than it looks. For most ads, Meta's library only lets you search what is active right now. Ads about social issues, elections or politics are kept for seven years, and ads delivered in the EU are archived for a year after their last impression. Outside those cases, once an ad stops, it drops out of the search.

So a library shows you today, not last quarter. If you want history, someone has to record it. That is part of what our Ad Library does: it keeps ads after they stop running, from the day we start tracking a brand.

The method below works in any library. The examples use Meta's, because it is the one you are most likely to open first.

Sort by days running

Brands stop paying for ads that lose money. That is the simplest idea in this guide and the most useful one.

Say a brand reviews its ad budgets once a week. An ad still live after 60 days has survived roughly eight of those reviews. Each time, someone looked at numbers you cannot see and decided to keep paying. One review could be luck or a busy week. Eight is a decision.

So before you look at anything else, work out how long each ad has been live: today's date minus its start date. Read the longest-running ads first, and ignore the rest until you have read those. It feels slow. It is the fastest way to reach the ads that matter.

A rough way to group what you find:

What longevity does not prove

Days running is a signal, not proof of profit. Keep these in mind:

• An ad can run for months on a tiny budget. Long life, little spend.
• Retargeting and brand ads are often left on and judged by different rules.
• Sometimes an ad is still running because nobody switched it off.
• Seasonal ads stop because the season ended, not because they failed.
• The start date is when delivery was set to begin. It does not prove the ad ran every day since.

That is why you never read one number alone. You read it next to variants and relaunches, which is what our Signal score does: a 0 to 100 estimate built from all three. It is labelled "Estimated, not spend", because that is what it is. For the longer argument, read why days running beats every other metric.

Count the variants

One ad with seven near-identical versions is a brand scaling a concept, not testing it. Testing looks different: several unrelated ideas, each with one or two versions, many of them gone within a few weeks.

When you find a cluster, line the versions up and look for what stays the same. The opening shot? The first line of copy? The person on screen? The offer? The thing that never changes is the part the brand believes in. The things that change are what it is still testing.

A hypothetical example. Say a skincare brand is running seven versions of one video. The first three seconds are identical in all seven: a close-up of serum dropping onto skin. The voiceover changes from version to version, and so does the offer in the last frame. What that tells you: the opening shot is settled. The words and the offer around it are still being tested.

Variants also show format decisions. If one concept appears in 9:16, 4:5 and 1:1, the brand has paid to resize it for every placement. Resizing takes work, so note which ideas a brand thinks deserve it.

Two patterns to recognise:

• Different ads, same idea. Versions are not always exact copies. A new hook on the same body, or the same script with a different presenter, is still one concept.
• Many ideas, one version each. That is a brand in testing mode. Interesting to watch, but nothing is proven yet. Come back in a month and see which ones survived.

In most markets, longevity is the best performance signal a public library gives you. Use it.

Watch for relaunches

An ad that disappears and comes back weeks later worked once, and someone remembered. Relaunches are some of the cheapest research you can do. The brand already paid for the test, looked at the results and chose to pay again.

The catch is that you can only spot a relaunch if you saw the first run. As above, most libraries show what is live today. If a brand brings an old creative back as a new ad, it arrives with a fresh start date and looks brand new. Your days-running count resets to zero, even though the idea has been proven before.

Three habits catch most of them:

1. Save the ads you find interesting, with the date you saw them.
2. Check the same brands again every week or two.
3. When a creative you have saved turns up again, note what changed: the hook, the offer, the format or nothing at all.

The ones that come back with a change are the most useful. They show you the part the brand kept, and the part it thought could do better. There is more on reading them in Relaunches: the cheapest research in paid social.

Write it down

Save the three longest-running ads, note the hook of each in one line, and you have a brief. That is the whole method. The rest of the ten minutes is about keeping yourself honest.

Here is how the time splits:

1. Minutes 0 to 2. Search the brand. Note how many ads are live and which formats they use.
2. Minutes 2 to 5. Work out days running and open the three oldest ads.
3. Minutes 5 to 7. Check the variants on those three. Write down what stays the same.
4. Minutes 7 to 8. Compare against what you saved last time. Has anything come back?
5. Minutes 8 to 10. Write one line per ad.

Each line should answer three questions: what happens in the first three seconds, why someone would buy, and what the deal is. For a hypothetical serum ad, that might read: "Hook: close-up of serum on skin, no voice. Angle: one step replaces three. Offer: 20% off the first order."

Three lines like that are enough to brief a designer, an editor or a studio. You know what is working in your category, which parts are proven, and what you would change.

Keep the notes next to the ads. A folder of screenshots and a separate document drift apart quickly. In Marketing Window you can save ads to your swipe file with a note on why, and group them into boards as you go. Every plan includes a swipe file and boards, and every account starts with a 7-day free trial.

One last rule. Take the structure, not the ad. Copying a competitor's ad frame by frame is lazy, and it can get you into trouble. What you want is the pattern underneath: the kind of hook, the angle, the format. Your version should use your product, your claims and your voice.

Once you have more than three ads saved, the next step is turning them into a proper brief. We walk through it in from six saved ads to one brief.