What the Meta Ad Library is (and is not)
The Meta Ad Library launched in 2018 as the Political Ad Archive and expanded in 2019 to cover all active ads across Facebook and Instagram. It is the richest public record of advertising ever assembled. It is also radically misunderstood. Most of the guides you will find are either thin swipe-file listicles or Meta's own help-page prose. Neither tells you what you can actually do with it as a researcher.
Political and social-issue ads get a full historical record in most countries. Commercial ads outside the EU are visible only while active and for roughly thirty days after. Inside the EU, the Digital Services Act Article 39[1] changed this in February 2024 for very large online platforms: all ads are retained for at least one year with full targeting parameters and reach by member state.
The Ad Library is the single largest open corpus of creative performance signal in existence. Ninety percent of the people who try to use it miss the value.
What you can actually pull
Per ad, the API returns the creative asset URL, the advertiser page, a spend range (not exact spend), an impression range (not exact impressions), demographic splits, active dates, and for EU political ads, targeting parameters and per-member-state reach.
What you cannot pull: exact spend, exact impressions, click data, conversion data, or individual targeting parameters (except under EU DSA for political ads). The Ad Library is an input disclosure record, not a performance dashboard. It tells you what ran, not how well it ran.
Getting API access
The Ad Library API requires identity verification, a Facebook developer app, and an ad account. Plan for two to three days of back-and-forth on identity verification, and expect roughly 200 requests per hour per access token as the historical rate limit. Aggressive queries get throttled. Batch by advertiser or date window, not by creative ID, to stay inside limits.
Common error patterns: rate-limit rejections without helpful error messages, inconsistent field returns across regions, and occasional reverification demands. None of this is unique to Meta, and none of it is catastrophic. Build a retry layer with exponential backoff and log every rejection.
How serious researchers use it
NYU Ad Observatory. Damon McCoy and Laura Edelson at the NYU Cybersecurity for Democracy lab[2] used an Ad Observer browser extension to collect ads seen by volunteer users, supplementing the public API. In August 2021 Meta disabled the researchers' accounts, prompting a public condemnation from FTC Acting Director Samuel Levine. The story is important because it showed the limits of the public API for accountability research.
Mozilla Ad Observer. The continuation extension lives on for Firefox and Chrome.
Who Targets Me. UK-based, founded by Sam Jeffers in 2017. Tracks political ad targeting in more than thirty countries.
AI Forensics. Paris-based research org founded in 2022. Their 2024 "No Embargo in Sight" report[3] documented Russian disinformation in the EU Ad Library. A 2023 audit study found Meta approved 75 percent of test EU election ads that contained disallowed content.
Academic papers powered by the Ad Library
Edelson et al. 2020. IEEE Security & Privacy[4]. Security analysis of the Ad Library. Documented that roughly 10 percent of political ads are missing from the public API at any given time, and that advertiser verification is inconsistent.
Silva et al. 2020. WWW[5]. Facebook Ads Monitor during the Brazilian elections.
Sosnovik and Goga 2021. WWW[6]. Roughly 55 percent of political ads escape Meta's automated classifier at the time of the study.
Ribeiro et al. 2019. FAccT[7]. Microtargeting of socially divisive ads.
These papers are the right starting point for anyone planning serious research on the Ad Library. They also document the honest limits you will run into.
The third-party swipe-file ecosystem
Several third-party products wrap the public API with search, tagging, and alerts. Foreplay ($59 to $299/month) is the most polished for creative teams. AdSpyder, Social Ads Spy, PowerAdSpy, AdLibrary.io, and Admixer Inspector occupy similar mid-market space with different interfaces and tagging depth. Sensor Tower (Pathmatics) and AdBeat resell cross-platform creative data including Meta, Google, and connected TV.
None of these adds new data. They add organizational and analytical layers over data you could, in principle, pull yourself. For most creative teams the time savings is worth the subscription. For academic research, go to the API directly.
Limitations you must plan around
- Missing ads. Edelson 2020 documented roughly 10 percent of political ads are absent from the API at any moment.
- Range-only disclosure. Spend and impressions are banded, not exact. Plan statistical analysis around banded variables, not point estimates.
- Short commercial retention outside the EU. Non-political commercial ads age out roughly thirty days after they stop running.
- Uneven advertiser verification. Page ownership can be ambiguous, especially for political and issue advertisers.
- Rate limits. Heavy research projects hit ceilings fast.
Five practical workflows
1. Competitor creative analysis. Pull active ads for competitor pages, download creative, tag manually or with VidMob/CreativeX (see the creative intelligence landscape).
2. Political spend tracking. Journalists, election monitors, and accountability researchers. The EU DSA expansion makes this materially richer for European elections.
3. Category benchmarking. Pair Ad Library creative density with paid-social benchmarks (see benchmarks) to compare against category medians.
4. Creative attribute labeling at scale. Feed Ad Library creative into a tagging pipeline to train internal models. Respect rate limits and Terms of Service.
5. Validation corpus for behavioral prediction models. This is our use case at OpenAffect. We use public Ad Library data to calibrate our predictions against real-world performance signals (see the calibration page). The Ad Library is the closest thing to open ground truth that the category has.
References
- 1EU Digital Services Act package.
- 2NYU Cybersecurity for Democracy.
- 3AI Forensics research.
- 4Edelson et al. Security analysis of the Meta Ad Library. IEEE S&P 2020.
- 5Silva et al. Facebook Ads Monitor. WWW 2020.
- 6Sosnovik and Goga. WWW 2021.
- 7Ribeiro et al. FAccT 2019.
- 8Meta Ad Library.
- 9Meta Ad Library API docs.
- 10Mozilla Foundation.
- 11Who Targets Me.
- 12Foreplay.