How to see what your competitors advertise on Facebook

You can't see what a competitor spends. Outside political and social-issue advertising and ads delivered into the EU or the UK, no ad library publishes spend, reach or targeting for an ordinary commercial ad. What you can see is how long each ad has been live, how many versions exist and where they run. Read those properly and they're enough.

Can you see how much a competitor spends on ads?

No, and almost every guide on this subject implies otherwise. Meta's Ad Library shows spend ranges and impression ranges for one category of ad only: political, electoral and social-issue advertising, where the disclosure exists for electoral integrity reasons rather than for competitive research. A skincare brand's prospecting ad carries none of it.

TikTok's Creative Center is the same story in a different shape. It'll show you top-performing ads by industry, region and objective, with engagement indices attached, and it withholds exact spend and the targeting behind them. That's a curated showcase, not a ledger.

So the first correction to make is to stop looking for the number. Third-party estimates exist, they're modelled from impression proxies, and nobody publishes the method well enough for you to check it. Building a competitive read on an unauditable estimate is worse than building one on the observable record, because it feels precise.

What is actually in the public record?

More than most people use, and less than most people assume. For a commercial ad, the Meta Ad Library gives you the creative itself, the page running it, the date it started running, whether it's currently active, the platforms it can appear on, the destination link and a library ID. Where an advertiser has several versions of one ad under a single entry, the library says so and lets you page through them.

That's the whole of it. There's no spend, no impressions, no audience, no frequency, no cost per result and no indication of budget. The library was built as a transparency tool for public accountability, and the competitive-research use is a side effect people discovered afterwards.

Why is how long an ad has run the best signal you have?

Because an ad costs money every day it delivers, and the decision to keep paying is renewed continuously. An ad that's been live for four months has survived four months of somebody looking at its cost per purchase and deciding not to turn it off. That's a revealed preference, and it's the only thing in the library that behaves like one.

Creative fatigue makes the signal stronger rather than weaker. Video creative decays as the audience that has already seen it grows, so an ad that keeps delivering is an ad that keeps beating the account's own alternatives in an internal auction that gets harder over time. Survival is expensive.

What can you learn from the EU version of the same library?

This is the part almost nobody uses. Article 39 of the EU's Digital Services Act requires very large online platforms to keep a public repository of the advertising they carry, holding each entry for a year after the ad was last displayed. The repository has to contain the ad itself, who it was presented on behalf of, who paid for it if that's someone different, the period it ran, whether it was aimed at particular groups and the main parameters used, and the total number of recipients reached.

Reach is the field that changes the read, and it is reach specifically. Spend and impression ranges stay restricted to political and social-issue ads no matter where an ad ran; what EU delivery adds is an estimated total reach figure. So a brand that sells into Europe hands you a rough size for the same creative you can only see the start date of in the US view. It isn't the same campaign and the budget split will be different, but it's the same brand making the same argument with a number attached to it.

The practical move is simple. Look your competitor up twice, once in the ordinary view and once filtered to an EU country, and see which creatives appear in both. Anything running in both places with a meaningful reach band in Europe is a creative the brand has committed to globally.

What else is observable, and how does each signal mislead?

Five things beyond longevity are visible from outside: how many versions of an ad are live, which placements it is eligible for, how fast new creative appears, where the ads point and what the ads are arguing. Each one carries real information and each one has a specific way of being wrong.

How do you read one competitor library end to end?

The read takes about twenty minutes per brand once you know what you're doing, and it produces a list of angles ranked by how confident you can be that they work.

1. 01

   Pull everything and sort by start date

   Not by relevance, not by the order the library returns. Start date is the axis the whole read is built on, so it has to be the sort.

2. 02

   Cut everything under three weeks old

   A new ad hasn't survived anything yet. Set those aside as a separate list called cadence, because their count per week is the useful thing about them, not their content.

3. 03

   Group the survivors by argument, not by format

   Two founder-to-camera videos making the same claim are one angle with two executions. Grouping by format hides the thing you came for.

4. 04

   Check the same brand in an EU country

   Any creative that appears in both views with a reach band attached moves to the top of the list. That's the closest thing to a confirmed winner you can see.

5. 05

   Open the destinations

   A brand running six angles into one generic collection page is testing creative. A brand running six angles into six different landers is testing the offer, and that's a much more expensive programme to copy.

6. 06

   Diff it in thirty days

   Which of the long-runners died, which of the new ads survived. One snapshot is a photograph. Two snapshots are a trend, and the trend is where the real information lives.

Doing this by hand means the Meta library in one tab, the TikTok Creative Center in another and a spreadsheet you maintain yourself. That works, and it's how most people do it. RIVYL pulls a rival's full ad library across Meta and TikTok into one view sorted by how long each ad has run, which is the same read with the sorting already done.

What does a worked read look like?

Here's a hypothetical supplement brand with 54 live ads. The numbers below are illustrative rather than measured, and the point is the shape of the conclusion rather than the digits.

Read it and the account resolves into two stories. Three of the five angles have something running past 90 days and the other two don't. The 22 discount ads look like the biggest thing in the library and are the least informative part of it, because their median age is nine days and a burst of promotional creative tells you what month it is rather than what works.

What you'd take away is that this brand converts on a founder story and on creator reviews, has never made before-and-after work despite trying fourteen times, and treats ingredient education as a lower-funnel job. That is a creative brief, and none of it required a spend figure.

Where does this reading go wrong?

In four places, and each one has produced a confidently wrong strategy at some point.

A long-running ad may have no performance bar at all. Larger brands run awareness campaigns judged on reach or brand lift, and those sit in the library for months next to performance creative that had to earn its slot. The tell is usually the destination: an ad pointing at a homepage with no offer isn't being judged the way an ad pointing at a product page with a discount code is.

Volume is not spend. That's the most common error. Dynamic and catalogue formats generate library entries mechanically, so a brand with hundreds of near-identical entries may be running one modest campaign. Count distinct arguments, not rows.

You're looking at survivors only. The library shows what ran. It never shows what was killed in week one or what somebody scripted and abandoned before it went live. So the angle mix you observe has already been filtered by a process you can't see, which makes it a good source of hypotheses and a bad source of certainty.

Their economics aren't yours. An ad that sustains a $70 cost per purchase for a brand with a subscription and a 65% margin is unfundable for a single-purchase product at 40%. Before copying an angle, check the arithmetic against your own numbers rather than theirs.

What do you do with what you find?

Turn survivors into arguments, then rebuild the arguments in your own voice. Copying an execution gives you a worse version of somebody else's ad, because you're missing the offer, the audience and the eighteen months of iteration that produced it. Copying the argument gives you a hypothesis you can test cheaply. The angle taxonomy is a way to make sure the ten you build from a competitor read are ten different arguments rather than ten different images.

Then plan the output rate before you plan the shoot. The arithmetic of how many creatives you need each week starts from a measured winner rate of about 5%, which sets the number of concepts a competitor read has to produce to be worth doing at all.

And when your versions go live, judge them on the right metrics. Hook rate and hold rate diagnose attention and not much else, which matters because the temptation after a competitor study is to grade your new creative on whether it looks like theirs. The launch arithmetic covers the budget floor a new set of creatives needs before any of its numbers mean anything.