Unofficial Facebook Ads Library API: search Facebook and Instagram ads by keyword, Page ID or Ad Library URL, and get ad copy, images, videos, CTAs, landing pages and dates as JSON or CSV. Date range filter, exact ad limits, fast HTTP scraping. No Meta token or login.
Ten public example tasks, each aimed at a searched keyword found with AnswerThePublic and Google autocomplete:
competitor ads by Page ID
a brand keyword search
product research
lead ad examples
real estate and photography ad examples
a video downloader
search URL export
Instagram ads
a date range
Three examples created on November 20 were renamed and given new keywords, so their old example URLs redirect to the examples list.
scripts/publish-tasks.mjs renames a task through replaces and checks videos, images and call-to-action text before publishing.
0.2 listing update (2026-11-20)
Store listing repositioned as an unofficial Facebook Ads Library API and Instagram ads scraper, a search niche separate from the main Meta Ad Library Scraper.
README follows Apify's Store structure: keyword-first introduction, tutorial, input examples with a matching real output, API usage in curl, JavaScript and Python, pricing, related Actors, and FAQ.
Input form: readable labels for ad status and media type, the ad limit next to the search fields, and the date range in its own section.
Dataset views: labeled and typed columns, the partnership creator, and a new Creatives view.
Four public example tasks: a keyword search, a competitor Page export, Instagram ads by keyword, and a date-range search.
0.2 (2026-11-20)
Ports the improvements made to this scraper's code inside Seercraft, checked against live Ad Library responses. The code was released as build 0.2.2 after 16 of 16 cloud checks passed on November 20, 2026; build 0.2.3 updates the Store listing text only.
New
Date range.dateFrom and dateTo inputs, and start_date[min]/start_date[max] in pasted URLs, apply Meta's Ad Library date filter. Meta matches ads that ran during the range, so ads that started earlier can be included.
Pasted URLs just work. Tracking and navigation parameters such as fbclid, utm_*, or source are ignored and listed in RUN_SUMMARY.ignoredParameters instead of failing the run. m.facebook.com, web.facebook.com, a missing trailing slash, and # fragments are accepted. Any sort mode Meta writes in its URLs (for example relevancy_monthly_grouped) is accepted.
Video posters. Each video has poster, the preview image Meta shows before playback.
Partnership ads. Branded-content ads have collaboratorName, collaboratorPageId, and collaboratorLogo for the creator the ad runs under.
Clearer run outcome. The run's status line says what was delivered and why it stopped. RUN_SUMMARY adds skipped and metaResultCount (Meta's approximate total).
Changed output
endDate is null for running ads. Meta reports the current day as a live ad's end date, which is not an end.
Dynamic catalog (DPA) ads are format: "dynamic" with multipleVersions: true, like dynamic creative (DCO) ads. They were reported as carousels.
multipleVersions is false for known single-version formats; it was null. It stays null when Meta reports no format.
A post with several images (MULTI_IMAGES) is a carousel; it was image.
ctaText of a dynamic ad is the first version's button text, as the Ad Library shows it.
brandLogo is null on branded-content ads; it used to show the creator's picture as the advertiser's logo.
Reliability
One malformed record no longer fails a whole run: it is skipped and counted, and the run is marked not complete. A page with no usable record still fails loudly.
In a Page search, a stray ad from another Page is skipped instead of failing the run.
A search that ends exactly at maxAds is reported complete, and a fully delivered page advances the resume checkpoint.
A maintainer-configured operation ID that Meta invalidates (error 1357004) falls back to discovery instead of failing every retry.
A redirect to Meta's login or checkpoint page is a retryable block. An off-site redirect is never followed and is retried on a fresh session, and its error names the target host. got had wrapped it as a network error, which hid the code. Making it terminal in build 0.2.4 failed one run whose Residential IP was redirected; two reruns on fresh IPs succeeded, so it is retryable again.
Error 1675004 is now documented as Meta's per-IP GraphQL rate limit (checked in a raw response), not a rejected cursor.
Cost
Meta's script bundles (static.*.fbcdn.net, several MB per cold run) bypass the Residential proxy. Only www.facebook.com is proxied. HTTP_METRICS.proxiedBytes reports the proxied share. In the release checks, a 31-ad run proxied 0.7-1.7 MB instead of about 8.3 MB and cost $0.002-0.004 instead of $0.015-0.018; 31-ad runs took 9-26 seconds instead of 15-29.