How to find competitors' winning ads

Every ad-spy blog tells you the same thing: "if an ad has been running a long time, it's working". Nobody handing out that advice has ever measured it. We track 143,006 competitor ads with run-lengths recorded, so this guide replaces the folklore with numbers: how long ads actually run, where longevity is a sharp filter and where it's nearly useless, and the second signal most people miss entirely.

Why ad longevity is the best public proxy for "winning"

No platform publishes a competitor's conversion rate. What the public ad libraries do publish (free, for any brand) is when each ad started and, on some channels, when it was last shown. That matters because of one simple economic fact: advertisers kill ads that don't pay. Every extra week an ad survives is another week someone with full visibility of its performance chose to keep funding it.

Survival isn't proof of profit (we'll get to the caveats), but across a large enough set of ads it's the strongest signal you can get without seeing the advertiser's own dashboard. The question the generic advice never answers is: how long is "long"? That depends enormously on the channel and the sector, and this is where real data earns its keep.

What "long-running" actually means: 143,006 ads measured

Across the UK brands AdRival tracks, 143,006 creatives have a measurable run length. 50,280 of them (35%) have been running for 180 days or more. But that headline hides a dramatic split by channel:

Channel	Ads with run-length data	Median run	Share running 180+ days	Longest we've seen
Google (Search, YouTube, Display)	110,145	129 days	45%	1,969 days (~5.4 years)
Meta (Facebook & Instagram)	16,689	18 days	1.5%	675 days
TikTok	16,172	1 day	<0.1%	1,110 days

Read that again: the median Google ad in our corpus has run for over four months; the median Meta ad, eighteen days. A "6-month-old ad" means completely different things on the two channels: on Google it's unremarkable, on Meta it's a one-in-66 survivor. Any tool or blog post that applies one longevity threshold across channels is telling you nothing. (These run lengths are the platforms' own first-shown and delivery-start records, not our scrape timestamps; we've documented exactly where each date comes from.)

TikTok is the odd one out: its library's shown-dates make almost every ad look days old. Whether that's genuinely brutal creative churn or conservative disclosure, the practical conclusion is the same: longevity is not a usable winning-ad filter on TikTok. Use it on Google and Meta.

Finding winning Google ads: longevity barely filters — until you know the trick

Here's the finding no ad-spy blog has the data to tell you. For a mature Google advertiser, almost everything live is already long-running. Among live Google ads we track: Victorian Plumbing is at 98% running 180+ days (of 947 live ads), Dreams 97% (of 800), Pets at Home 96% (of 1,202) and carwow 89% (of 2,340). Sorting those brands' libraries by "longest running" doesn't shortlist winners — it shortlists nearly the whole library.

So on Google, do two different things:

• Look at the extreme tail for the evergreen core. carwow has a creative first shown on 31 October 2021 and still being served 1,817 days later; Pets at Home's oldest live ad dates to 25 October 2021 (1,772 days). Ads like these have survived roughly five years of budget reviews — that's the brand's proven, always-on message, and the single best thing to study before writing your own.
• Flip the filter for brands that churn. Some advertisers run Google like Meta: only 5% of Motorway's live Google ads and 6% of Ocado's have reached 180 days; EE's median live ad is 26 days old. For brands like these, longevity works again — the rare old ad in a sea of fresh ones is the one that earned its place.

Know your sector's baseline before judging any single ad

"Long-running" is relative to the category. Median run length of live Google ads in our data, by sector:

Sector	Median run (live Google ads)	Share at 180+ days
Pet retail	730 days	89%
Furniture	702 days	80%
Mattress & sleep	617 days	75%
Travel & holidays	449 days	88%
Fashion	407 days	67%
Fintech	341 days	63%
Insurance	316 days	62%
Supermarkets	100 days	30%
Telecoms	100 days	32%

A 200-day ad is a veteran for Tesco (supermarkets churn on promotions; median just 100 days) and a baby for Pets at Home. Judge each ad against its own sector's median, not an absolute number.

Caveat: for very high-volume Google advertisers, the Transparency Center stops paginating at roughly 900 creatives even when it reports thousands, so any tool's Google numbers (ours included) are a large sample, not a census, for the biggest spenders. We documented this cap in our Google Ads Transparency Center guide.

Finding winning Meta ads: longevity is a brutal, beautiful filter

On Meta the maths inverts. The Meta Ad Library only shows ads that are running right now, and brands launch and kill creative fast — that 18-day median. So the 1.5% of live Meta ads that pass 180 days are genuinely exceptional, and the library hands them to you:

1. Open the brand's page in the Meta Ad Library (country: UK, category: all ads).
2. Check each ad's "Started running on" date. Anything past ~90 days is already in the top tier of survivors; past 180 days is elite.
3. Study what those survivors share (the hook, the offer, the format) before you look at anything launched last week.

Real examples from our tracking: Huel currently has 22 of its 291 live Meta ads past the 180-day mark, the oldest at 285 days: a visibly deliberate evergreen layer under its launch churn. iwoca has kept one Meta ad alive for 675 days, and Wren Kitchens one for 492. When a performance-marketing team leaves a single creative running for a year and a half, that is as close to a public declaration of "this one prints money" as you will ever get.

The second signal: how many versions are they running?

Longevity tells you what has already won. Variant count tells you what the brand believes is winning right now. The Meta Ad Library groups near-identical ads ("this ad has multiple versions"), and when an advertiser spins one concept into many versions, they're scaling it: cutting new ratios, swapping headlines, pushing budget into what works.

In our live Meta data, 821 of 16,309 active ads sit in groups of five or more versions: about 5%, another naturally selective filter. Nectar is the heaviest scaler we track, with 466 of its 810 live Meta ads in version-groups of five-plus; Gymshark runs concepts in up to 14 versions, and Cancer Research UK has pushed a single concept to 32. An ad that is both old and heavily versioned is the strongest "winning ad" read available from public data.

A 20-minute workflow you can run today

1. Pick one competitor and one channel: Google if they're search/brand-led, Meta if they're DTC/social-led. (Our free ad libraries guide has exact steps and links for all four.)
2. On Google: note the oldest live ads (the evergreen core) and the sector baseline above. If most of the library is young, the rare old ad is the story.
3. On Meta: shortlist everything past 90 days, then rank by version count.
4. Extract the pattern, not the ad. For each survivor write one line: what's the hook, what's the offer, who's it for. Copying the creative gets you a lawsuit; copying the pattern gets you a brief.
5. Recheck in two weeks. A survivor that disappears, or a new ad that suddenly sprouts versions, is the change worth reacting to.

What longevity can't tell you

• Survival isn't ROI. Big brands run awareness campaigns that are never judged on direct response; a 2021-vintage ad may simply be brand wallpaper — or genuinely forgotten in a dusty campaign. Weight the signal by how disciplined the advertiser is.
• Relaunches reset the clock on Meta. A brand that relaunches the same creative gets a fresh start date each time, so a "young" Meta ad can be an old winner in disguise. (Details in our launch-date accuracy write-up.)
• LinkedIn's history is too shallow for this. Its library shows run dates, but ads disappear a year after their last impression, so the multi-year survivors this method leans on are invisible there. (See our LinkedIn Ad Library guide.)
• You only see what you capture. Meta's library shows no history for ordinary commercial ads: once an ad stops, it vanishes. Judging winners properly means snapshotting over time, which is manual-spreadsheet territory or a tracking tool's job.

Everything in this guide is doable by hand with the free libraries. What AdRival adds is doing it continuously: we snapshot each tracked brand's ads across Google, Meta, TikTok and LinkedIn, record every run length, and qualify each creative on 19 dimensions (hook, offer, discount, audience, urgency and more) — so "show me this competitor's proven evergreen ads and what they're scaling this week" is one view rather than an afternoon's work. You can see it on any brand page free, and reports start at £99.