Meta Ad Library: The Free Competitive Radar Most Advertisers Never Turn On

Meta Ad Library is the most underused competitive intelligence tool in paid media — and it's free. It shows every active ad a brand runs on Facebook, Instagram, and Messenger: the creative, the format, the platforms, and the date each ad started running. Most marketers open it, check one competitor's ad, and close the tab. In this article I'll show you the systematic workflow we use at Mintec — a weekly radar that turns a transparency archive into a creative intelligence feed, plus the honest limits of what it can and can't tell you in 2026.

What's actually inside the library

Meta launched the Ad Library in 2019 as a transparency archive. Anyone can search by advertiser, keyword, or country and see every active ad from any Page. No login needed, no paywall, no "contact sales."

What you get for each ad:

• The creative itself — image, video, carousel, or copy
• The format (feed, stories, reels, messenger, right column)
• The date it started running ("active since")
• Which platforms it's live on and whether it's still active

What you don't get:

• Spend. The library has never shown real budgets. Political and issue ads show broad spend buckets; commercial ads show nothing.
• Audiences and results. No targeting, no reach, no ROAS. The library answers "what are they showing?" — never "what's working?" directly.
• Platform performance. You have to infer that from behavior.

That last point is the key. The Ad Library is not a results dashboard. It's a behavioral archive: what competitors choose to keep running, for how long, in what combinations. And behavior is exactly what a competitive radar should track.

Why most competitor research fails

The standard approach is "spy on the competitor." You pick your top rival, watch their ads, and try to copy what's new. There are two problems with that.

First, single-competitor obsession creates survivor bias. Your biggest rival is not the market. The brand that's quietly growing in an adjacent category is often the one with the better playbook — and the one nobody watches.

Second, copying the new is copying the unproven. Advertisers see the shiny new campaign a competitor launched this week and assume it works. But most new creatives fail — Meta's own data suggests creative quality is the top performance driver, yet the average testing cycle under Andromeda burns concepts out in weeks, not months. What's still running after 60 or 90 days is the evidence. The Ad Library's active-since date is the closest thing to a free "this is a winner" flag in paid media.

So our first rule at Mintec: stop spying on rivals, start mapping the category.

The category radar: our weekly workflow

We run this on client accounts every week. It takes 45 minutes per account once the structure is set up.

Step 1 — Sample the category, not just the competitor. Pick 8 to 12 advertisers: 3-4 direct competitors, 3-4 brands in adjacent categories serving the same audience, and 2-3 "aspirational" brands from other markets (for LatAm clients, that usually means US or European DTC brands spending at a level your market can't match yet). Add one variable nobody thinks about: brands that disappeared. A Page that stopped posting ads is also information.

Step 2 — Log signals, not ads. Every week we record four things per advertiser: which creatives are still active (unchanged = scaled), which are new (unknown = testing), what formats dominate, and any new markets or languages. One row per advertiser, one tab per week. This becomes the "active since" timeline that the one-off spy never sees.

Step 3 — Cluster by angle, not by brand. After a month you'll see patterns the individual ads hide: every category leader is running a problem-solution hook in feed, a testimonial in stories, and an offer-led angle in reels. When you cluster by angle, you're no longer watching competitors — you're watching the category's consensus. And consensus is where the cheap insights live.

Step 4 — Write briefs, not reports. Each cluster becomes a test hypothesis for our client's account: "Three category leaders are rotating comparison creatives weekly; we have none active — brief two variations for the next creative cycle." This is the step that turns research into ROAS, and it's the one almost nobody does.

Reading the signals like a media buyer

Once the log exists, these are the signals we actually trade on:

Long-running ads are expensive proof. Meta doesn't keep unprofitable ads alive at scale. An ad that has run for 90+ days across the same placements is a competitor telling you their winner, in public, for free. Log it, study its structure, and brief a distinct variation — never a copy, because the Creative Similarity Score makes near-copies self-defeating.

Format shifts predict budget shifts. When a category leader stops launching new feed ads and starts launching reels and stories formats, it usually means they're following placement cost changes. We saw this in mid-2026 when Meta CPMs were climbing and brands quietly rotated formats; our own Meta CPM analysis documents the cost side of that move.

New placement adoption shows up early. If your category is still ignoring a placement with dramatically cheaper CPMs, the radar will show you the moment the first competitor tests it — which is your signal to test before the window closes, not after everyone's in.

Language and geography expansions. A brand that starts running Spanish-language creative for the first time is telling you they found demand in your market. That's a lead-gen signal for you — and a head start on the brief.

The 2026 catch: the API got gated

The browser version of the Ad Library is still free and open. But if you want to build this radar at scale — pulling hundreds of ads per week into a spreadsheet or dashboard — the Facebook Ads Library API situation changed. As of 2026, Meta requires identity verification and an access token to call the ads_archive endpoint. That's a real workflow change for agencies that built internal tools on it: the age of anonymous scraping is over, and automated research now lives behind Meta's authorization flow.

The limitations haven't disappeared either. Independent analyses of the API point to the same walls: the free tier is rate-limited to roughly 200 requests per hour, commercial ad coverage in the EU is narrower than in the US, and spend data stays bucketed or absent. None of that breaks a weekly manual workflow — 200 calls a week is still 10,000 ads a year — but it means the "build a scraper and forget it" playbook is dead. You either do the verification dance or you run the radar manually. For most teams, manual is fine: the discipline of logging 12 advertisers a week is worth more than a dashboard that updates itself.

Where the radar feeds into

The output of this workflow is never a report. It's input for the testing machine:

• The 3-phase testing framework we run under Andromeda starts with structured idea generation — the radar is our idea source.
• The creative angle matrix for Meta and TikTok tells us which quadrants to fill; the radar tells us which quadrants the category is filling.
• Cluster data tells us how many distinct concepts to brief, which connects directly to the similarity-score discipline that stops us from feeding Meta near-duplicates.

In other words: the Ad Library isn't a spying tool. It's the cheapest research department a paid media team can hire — if you treat it like a system, not a tab you open once a quarter.

The one thing I'd change in most accounts: block 45 minutes on Monday, build the 12-brand sample, and log for four weeks before drawing a single conclusion. The first three weeks are noise. The fourth week is where the pattern appears — and where the next creative cycle starts writing itself.