Every active ad on Meta's platforms is public. Not a leak, not a scrape: a searchable archive Meta publishes deliberately, covering every advertiser, for free, with no login.
Your competitors are paying money to say specific things to specific people, and you can read all of it. Almost nobody does, and the ones who do mostly browse rather than research.
Here's how to make it produce a decision.
What it is and where
The Meta Ad Library covers ads currently running across Facebook, Instagram and Meta's other surfaces. Search by advertiser name or by keyword, filter by country.
TikTok has an equivalent in the Creative Center, though it works differently: it's a Top Ads browse tool organised by industry rather than a per-advertiser search. Useful for format inspiration, much weaker for tracking a specific competitor.
What it can and can't tell you
Be precise about this, because it's the difference between research and a pleasant hour of scrolling.
It can tell you:
- Exactly which creatives are live right now
- How long each has been running: the single most valuable field in the whole tool
- How many variants of an ad are running simultaneously
- The copy, the hook, the offer, the call to action, verbatim
- When a campaign started, which brackets launches and seasonal pushes
It cannot tell you:
- Spend, impressions, click-through, or conversions
- Whether any of it is working, in commercial terms
- Who the ads are targeted at
- What they tested and killed before this
That last one matters more than people realise. You're seeing survivors, not experiments. The library shows you the end of a process whose middle is invisible.
The method
1 · Search the competitor, then sort by what's oldest
Search each competitor from your watchlist. Note the total count of active ads. That alone tells you roughly how seriously they're investing.
Then look for what's been running longest. Those are the ads that earned their place. A brand's newest ad is a hypothesis; their oldest is a conclusion.
2 · Count the variants
If you see the same creative with twelve copy variations, they're testing. If you see one creative running alone for months, they've stopped testing and settled.
Both are informative. Heavy variant testing means a category still figuring out its message. A single long-running creative means somebody found something and is now just buying reach.
3 · Break each survivor into four parts
Don't describe the ad. Decompose it. Four fields, and the same four every time so the analysis aggregates:
| Field | The question |
|---|---|
| Hook | What are the first three seconds doing? Curiosity, problem, social proof, authority, urgency, story, stat? |
| Format | UGC, talking head, demo, testimonial, founder, voiceover-over-b-roll, text-on-screen, static? |
| Offer | What's the actual proposition? Discount, trial, guarantee, information, nothing? |
| Proof | What makes it credible? Testimonial, statistic, before-and-after, press, demonstration, authority? Or nothing at all? |
Recording these as fixed categories rather than prose is what turns twenty ads into a finding. Prose descriptions don't aggregate; tags do.
4 · Aggregate across the set
Once you've tagged fifteen or twenty ads across your competitors, count.
If eleven of twenty long-running ads use a problem-first hook and only two use aspiration, the category has learned something about its buyers. That's a genuine signal, and it cost you nothing.
Mind the sample. Under about three examples of a pattern, you have an anecdote. Acumin's own Ad Research applies exactly that floor (a pattern needs at least three categorised captures before it will call anything dominant) for the same reason the rest of the product does: a "dominant format" derived from two ads is a confident-sounding accident.
5 · Look for what nobody is doing
The most valuable output isn't the pattern. It's the gap.
If every competitor runs polished studio product demos and nobody runs a founder talking to camera, that's either an opportunity or a constraint. Be honest about which. Sometimes nobody does a thing because it doesn't work in the category, and sometimes because nobody's tried.
The way to tell them apart is cheaply: run it as a small experiment rather than a quarter's commitment.
The mistakes
Reading polish as performance. An expensive-looking ad tells you about budget, not results. The scrappy UGC ad that's been running nine months is beating it.
Copying the ad instead of the structure. The specific creative belongs to their brand, their proof, their audience. What transfers is the structural choice: problem-first hook, testimonial as the proof, a guarantee as the offer. Copy the skeleton, not the skin.
Confusing "running" with "winning". Correlated, not identical. New ads run because they're new. Large advertisers carry losers longer.
One session, then never again. Ad libraries change weekly. A single browse is a snapshot; the value is in the change over time: who started, who stopped, what got dropped.
Forgetting the funnel. A retargeting ad and a cold-traffic ad look different for good reasons. Judging one by the other's standards produces nonsense. See TOF/MOF/BOF.
The evidence tier, stated plainly
This is tier 1: public data anyone can see. It's a strong tier-1 source, because unlike organic view counts it carries an implicit budget signal in the run duration. But it's still the outside of someone's operation.
It's excellent for generating hypotheses. It cannot validate them. The validation is your own tier 3 data, from running something and measuring it.
Where Acumin fits
Ad Research is built around this workflow rather than around a feed, and the reason is worth stating: Acumin cannot pull competitor ads into the app. Meta's ads_archive API is restricted to political and issue ads; TikTok's Commercial Content API is researcher-only. Any product showing you a browsable in-app inventory of competitor ads is either scraping or showing you something else.
So the surface does what's actually possible and honest:
- Seeded deep links: one-click searches into the public Ad Library, pre-filled from your real category and your saved competitors, plus a TikTok Creative Center entry.
- Captures: you save an ad you found, and break it down into those four fields.
- Patterns: deterministic counts across your captures, with the three-capture floor before anything is called dominant. Arithmetic over your own tags, not an AI reading trends into six examples.
- Build to campaign: a captured pattern seeds a Concept Read or an Ad Lab campaign, so the research ends in something you make.
The capture step is manual by necessity, and that turns out to be a feature: tagging an ad by hand is what makes you actually look at it.
How to use this tomorrow
Take your three closest competitors. Search each in the Ad Library, and for each one write down the single ad that's been running longest, plus its hook type and its proof type.
Nine minutes, three data points. If two of the three share a hook type, you've found the thing to test next.
Related: Which posts deserve budget is the same question pointed at your own work. Building a competitor watchlist is how to make this a habit rather than a one-off.