Ask which Facebook ads are performing best and you get two kinds of answer. Lists of famous campaigns, which tell you what a writer admired. Or dashboards showing a competitor's ROAS to two decimal places, which tell you that somebody is comfortable making things up.
The second is worth pausing on, because it is everywhere. Meta does not publish CTR, CPC, clicks, conversions, or ROAS for ads you do not own. That data never leaves the advertiser's account. So when a tool shows you a rival's 4.2x return, there was no input for that number. It was generated to fill a tile. We wrote about why we ban fabricated metrics because the practice has become normal enough to look credible.
Which leaves a real question. If the numbers are not public, how do you find the ads that are actually working?
What Meta does publish
Three things, and one of them matters.
The creative and copy. Every ad currently running is in the public archive, in full.
Spend and impressions, sometimes, as ranges. Only for political and issue ads in most regions, and the ranges are so wide they are near useless for commercial work.
Whether the ad is running, and since when. This is the one.
The start date is public and the ad's status is public. Check every day and you know exactly how long each ad has been continuously live. That is a hard number, not an estimate, and nobody has to invent it.
Why run time beats engagement
The usual substitute is engagement. Likes, comments, shares, so pick the ad with the most.
Engagement is a weak proxy for two reasons. It measures reaction, not purchase, and the most-commented ad in a category is frequently the most annoying one. It also accumulates with time and spend, so an old ad looks impressive next to a new one that is converting better.
Run time works differently, because it reflects a decision rather than a reaction.
Someone at that company is watching a dashboard you cannot see. They know the real CTR, the real cost per purchase, the real return. Every day the ad stays on is that person, with their own money at stake, choosing not to switch it off. After 150 days that is roughly 150 consecutive decisions to keep paying.
You are not measuring the ad. You are reading the advertiser's revealed preference, and they have information you never will. That is the argument in more depth in why days-running beats every other performance proxy.
Reading it properly
Run time is strong, not infallible. Some ways to use it well.
Look at the distribution, not the single winner. One ad running 200 days could be a forgotten evergreen nobody turned off. Eleven ads over 100 days is a strategy, and the pattern across them is what you want.
Watch what repeats. If every long-runner opens on a problem before showing the product, that ordering is doing the work. If they all put a face in the first second, that is the finding.
Separate old from newly launched. Long-runners tell you what already won. The ads from the last three weeks tell you what they are testing now. Both are useful and they answer different questions.
Compare within a size class. A giant's long-running ad may survive on brand recognition you do not have. A competitor your size is a fairer read.
Do not confuse volume with success. A brand shipping forty new creatives a month is testing hard, which tells you about their process, not their results.
There is a practical walkthrough in finding a brand's longest-running Meta ad, and a piece on what the surviving hooks have in common in long-running Meta ad hooks.
What to do once you have found one
Identifying the winner is only useful if it changes what you ship. A screenshot in a folder does not.
Clone Studio is built for that step. Their ad sits on the left of the screen, sorted to the top of their library by run time, and your version builds on the right.
For video, the AI watches their ad and writes your script scene by scene before rendering anything: the timing, the voiceover line, what is on screen, plus a note on why their ad works. Every line is editable and nothing is charged until you approve it. Even at zero credits the script is free, so the analysis of a 150 day winner costs nothing to read.
For images, the clone comes back in about a minute.
Rendering video is priced by length: 1 credit is about 2 seconds of finished ad, 1 credit is $1, minimum 5 credits, and credits do not expire. Free accounts get the libraries, the scripts, and one image clone without a card.
There is a longer piece on the decision itself in what to do with a competitor's winning ad.
FAQ
How do I see the best performing Facebook ads?
Not by performance data, because Meta does not publish it for other advertisers. You find them by run time: pull a brand's active ads and sort by how long each has been continuously live. The oldest survivors are the ones the advertiser keeps choosing to fund.
Does Facebook show competitor ad performance?
No. CTR, CPC, conversions, and ROAS stay inside the advertiser's own account. The public archive shows the creative, the advertiser, and the dates. Any tool displaying a competitor's exact ROAS is making it up.
Is engagement a good way to judge an ad?
It is weak. Engagement measures reaction rather than sales, favours ads that provoke arguments, and accumulates with age and spend. Run time reflects a spending decision made by someone who can see the real numbers.
How long does an ad need to run to count as proven?
There is no fixed threshold, and it varies by category and budget. As a rough guide, past 90 days you are looking at something deliberate rather than accidental, and past 150 days it is a considered keeper. Patterns across several long-runners beat any single number.
What should I do with the winner once I find it?
Rebuild its structure for your product: the hook shape, the pacing, the way the offer is framed. Not its footage or copy. That is what Clone Studio does, and the boundaries are in cloning a competitor's ad ethically.
The scoreboard you can actually see
You will never see your competitor's ROAS, and you do not need it. You can see which ads they refuse to turn off, which is the same information filtered through somebody who had all the data and made a decision with their own budget.
Open Clone Studio, add the competitor you most want to beat, and sort their library by longest running. The ad at the top is their answer. Reading the script we write for your version of it is free.
You cannot see their dashboard. You can see what their dashboard made them do.