How to Build a Swipe File from a Competitor's Meta Ads in 5 Minutes

A swipe file is only as good as what's in it. Most people fill theirs with whatever caught their eye scrolling — a slick video here, a clever headline there. The result is a folder of ads you liked, not ads that worked. Those are very different things, and only one of them helps you write a brief.

The fix is to stop collecting impressions and start collecting evidence. Below is a five-minute workflow that builds a high-signal swipe file from any competitor's Meta ads — pulling the ones that have proven themselves, grouping them so the patterns jump out, and exporting the whole thing so it's actually usable.

Why the manual gallery-save approach fails

The standard move is to open Meta's Ad Library, search a brand, and start saving screenshots into a folder or a tool like Foreplay. It works, technically. But it has three problems baked in:

The Ad Library shows you every ad a brand is running — live tests, dead-on-arrival creatives, and proven evergreens all in the same grid, with no way to sort by what's working. So the manual swipe file inherits that noise. You end up curating on vibes.

A good swipe file inverts this. You want the few ads a brand has bet on — the ones it keeps paying to run — and you want them organized by the lever they pull.

The signal that separates winners from noise: days running

Here's the one number that does the heavy lifting. Meta doesn't hand you a competitor's CTR, CPC, or ROAS — those live inside the advertiser's account and nobody outside it can see them. (If a tool shows you a competitor's exact ROAS, it made the number up.) But there's a signal you can verify: how long an ad has been continuously running.

Brands are ruthless about spend. They do not keep paying to serve an ad that loses money. So an ad that's been live for 100+ days has, by definition, survived the only test that matters — it's still profitable enough to keep funding. Run-time is the cleanest performance proxy you get from the outside.

The catch: Meta's API returns no history. Ask it today and it tells you an ad is active; it won't tell you the ad has been active since January. That history has to be built by snapshotting the library daily and watching what persists. AdWhispr does exactly that, which is why "longest-running" is a question you can actually answer instead of guess at.

So the rule for your swipe file is simple: rank by days running, then keep the top of the distribution. Not one ad — read the spread. A brand with fifteen ads all past 90 days has a deep, stable strategy worth studying. A brand with one 200-day outlier and a pile of week-old tests is telling you something different.

The 5-minute build

This is the whole workflow. You can run it by chatting at adwhispr.com or with AdWhispr connected to Claude as an MCP server (npx adwhispr-mcp-server config, or OAuth at https://adwhispr.com/api/mcp).

Step 1 — Add the brand (about 30 seconds)

Paste the competitor's Facebook page or Ad Library URL. The brand's entire Meta ad library gets ingested and snapshotted. Start with one direct competitor — depth beats breadth for a swipe file.

"Add Ridge Wallet and pull their stats."

Step 2 — Pull the proven winners, not everything (1 minute)

Don't ask for "their ads." Ask for the ones that have earned their place:

"Show me Ridge's 15 longest-running active ads, with days running and format for each."

This is the move the manual approach can't make. Instead of scrolling a grid of 80 mixed-quality creatives, you get a ranked shortlist of what the brand has committed budget to over time. That shortlist is your swipe file's spine.

Step 3 — Group by hook and format (1 minute)

A flat list of fifteen winners is better than a screenshot folder, but the value is in the pattern. Ask AdWhispr to classify them — it tags every ad by hook, format, tone, and offer:

"Group those winners by hook and format. Which combinations show up most?"

Now structure appears. You might find their proven creatives cluster into three buckets — say, founder-talking-to-camera UGC, problem-agitation static images, and a "ditch the bulky wallet" comparison angle. That grouping tells you which formats survive for this brand, which is exactly what a brief needs.

Step 4 — Export it (30 seconds)

A swipe file you can't share is a private hobby. Ask for a competitive brief, which leads with the derived-intelligence panel — the longevity curve, engagement-verified reach, and how fast they ship new creative — then lists the winning ads grouped by angle:

"Generate a brief on Ridge's winning creative, exportable as PDF and Markdown."

Drop the Markdown into your team's doc, or hand the PDF to a client. That's a finished, evidence-backed swipe file — built in roughly the time it'd take to save five screenshots by hand.

Optional: turn the swipe file into your own creative

A swipe file's real payoff is the next ad you write. Once you've spotted a winning pattern, you can ask AdWhispr to clone_ad from a specific verified winner — an image ad becomes a fresh image in your brand's identity; a video ad becomes a scene-by-scene script and shot list. It's always original copy and visuals, always grounded in a real ad that's proven it can run, and it cites the source so you know which winner it's built on. The swipe file stops being a museum and becomes a starting line.

What you've actually built

The difference between this and a folder of saved ads isn't the tool — it's the selection criterion. By ranking on days running, you let the competitor's own spend decisions filter their creative for you. Every ad in your file is there because the brand keeps paying for it, not because it looked nice at 11 p.m.

That's a swipe file you can defend in a brief, hand to a freelancer, or reverse-engineer into your next test. Want to go deeper on reading run-time distributions and engagement-verified reach? More teardowns live on the AdWhispr blog.

Build your first evidence-backed swipe file free at adwhispr.com — one brand, no card.