How to find UGC ads in the Meta Ad Library and TikTok Creative Center

Everyone says "study winning ads" — almost nobody says where the ads live. Two free tools cover it: the Meta Ad Library shows every ad currently running on Facebook and Instagram, and TikTok Creative Center surfaces top-performing TikTok ads by industry. You'll see more real, in-market UGC there in an afternoon than a paid course — which is usually screenshots of these tools anyway.

The Meta routine:

  1. Open the Ad Library and set the country to a market you want to work in.
  2. Search a brand by name — start with brands you'd genuinely pitch: DTC skincare, supplements, apps, kitchen gadgets.
  3. Filter to video, then scroll past the polished studio spots. The ads that look like a person talking to their phone in a kitchen are the UGC ones.
  4. Note the "Started running on" date on every card. You'll need it in a minute.

In TikTok Creative Center, open Top Ads, filter by industry and region, and sort by CTR or likes. Meta shows everything a brand runs, winners and duds alike; TikTok hands you a feed that's already performance-ranked.

How to tell a winning ad from one that's just running

Run time is the tell. An ad still live after 60–90 days is paying for itself — no brand keeps funding a loser for three months, and most tests are dead inside two weeks.

The second tell is duplication — the Ad Library groups near-identical ads, and a creative with a stack of versions is getting more budget.

A quick read on any brand:

  • Started 60+ days ago, still active — proven; study it.
  • One creative across many versions — scaling; study it closely.
  • Same script re-shot by different creators — the structure won; now they're testing faces.
  • Launched last week — a test; it proves nothing yet.

What isn't a signal: production value, the organic version's view count, or your own taste. Fair warning — the ads making real money are often a little boring. They answer one objection per beat and get out. One caveat: huge brands sometimes leave awareness ads on autopilot — another reason to study DTC brands, who watch every dollar.

What goes in a swipe file — a simple template

Log five things per ad, in your own words. A folder of saved videos is not a swipe file — rewatching ads without logging them is doomscrolling wearing a productive outfit.

Per ad, one short note:

  1. Hook — the opening line word for word, plus what it's doing: calling out an audience, naming a pain, breaking a pattern.
  2. Structure — the beats with timestamps: problem (0–3s) → failed fixes (3–8s) → demo (8–20s) → proof → CTA.
  3. Cut rhythm — how often it cuts, and where b-roll takes over from the talking head.
  4. Overlays — caption style, which words land on screen, whether the offer flashes.
  5. CTA placement — end only or seeded mid-video, spoken or on-screen or both.

Add brand, link, start date, and version count. A notes app or one spreadsheet row per ad is plenty — the tool doesn't matter, the logging does.

Then make it compound: 20 minutes, once a week, same slot. Pick one niche, pull its two or three longest-running ads, log them properly. That's a dozen ads a month; around week eight you'll catch yourself predicting the next beat before it plays. Scripts stop starting from a blank page, and writing hooks turns from guessing into choosing between proven angles.

Where studying an ad ends and copying begins

The line is simple: structure is shared craft, wording is someone's work. Nobody owns "problem → demo → proof → CTA," any more than a songwriter owns verse-chorus-verse; reverse-engineering those patterns is what media buyers do all day. A specific script is different — the sentences, the jokes, the exact hook line belong to whoever wrote them, and lifting them is plagiarism at best, a copyright problem at worst.

There's a practical reason too: the people you're pitching watch these libraries for a living. A cloned script gets recognized, and its creator doesn't get a second email.

A clean test: can you explain why a beat works without quoting it? "She raises the objection before the viewer can think it" is learning. Reciting her line is copying. When you build on a swiped structure, swap everything specific — niche, product, angle, and especially the claims. Never carry a claim from someone else's ad into yours; you can't verify it, and neither can the brand's legal reviewer.

The reassuring part: structure is where most of the performance lives anyway. You lose nothing by rewriting every word.

How to use ad research in a pitch

Reference the brand's longest-running ad and offer to extend it. Compare:

"Hi! I'm a UGC creator and I'd love to work together. Here's my portfolio."

versus:

"Your ad that opens with 'I tried everything for razor burn' has been running since June, so it's earning its keep. I'd test two hook variations against it — one leading with the objection, one with the end result. Here are three videos of mine in a similar style."

The first message asks for a job. The second describes the media buyer's own job back to them — their week is spent hunting fresh variations of proven winners, and you just walked in holding two. I've watched this framing turn a cold DM into a discovery call: it proves you think in tests, not just film in kitchens.

The mechanics:

  • Pitch brands already running UGC-style ads. They pay creators; nobody needs persuading the format works.
  • Name the actual ad — quote its hook or start month — so the message can't be mistaken for a template blast.
  • Offer variations, not critiques. "Your ads are weak" gets deleted; "your winner could have siblings" gets a reply.
  • Back it up with two or three of your own videos in that niche, linked where they play instantly.

If cold outreach itself is the scary part, the full playbook is in how to pitch brands as a UGC creator — the swipe file just hands you a sharper opening line.

FAQ

Is the Meta Ad Library free, and do you need an account? Free, and browsable without logging in. TikTok Creative Center is free too; some dashboards ask for a TikTok for Business login, also free.

Can you see how much a brand spends on an ad? Not for regular product ads — spend is public only for political and issue ads, plus some EU reach data. Run time and variant count are your proxies, and they're enough.

How many ads make a swipe file useful? Fifteen to twenty properly logged ads beat two hundred saved ones. You'll know it's working when you can recall an ad's structure without rewatching it.

Should you admit you found the brand in the Ad Library? Yes. Their own team checks it weekly — referencing a live ad reads as homework, not surveillance.