Public ads.
Read straight
from GraphQL.
Meta’s Ad Library is open to anyone, but the data behind it comes through an internal GraphQL interface. I read that interface directly: no login, no API key, any country.
This run: 48 ads in 5 requests and 13 seconds, with no proxy and no API key.
How I read it.
MetaAdsCollector · meta-ads-collector 1.7.0- Page or keywordI start from an advertiser’s Facebook Page ID or from keywords, in one country, several, or all of them, for active ads, ended ads, or both.
- GraphQL, page by pageThe Ad Library’s own GraphQL interface answers each query with a page of ads and a cursor. I follow the cursor until the cap is reached or Meta has no more. The actor asks for 30 ads per call, the page size the Ad Library’s interface uses, and runs one query at a time, because Meta rate-limits concurrent traffic.
- One record per adEach ad becomes a typed record: its Ad Library ID, the page, delivery start and stop, platforms, display format, and every creative’s title, body, caption, link and call to action.
- Kept once, watched for new adsAn ad is kept once, however many queries return it. Run again later, the collector reports the ads it has not observed before, and long collections can run as background jobs.
- JSON, JSONL or CSVResults export to the format the next tool needs.
- CLI, Python, MCPThe same collector runs from the command line, as a Python API, or as an MCP server, so an AI assistant can search the Ad Library itself. On the Apify Store it runs as the Meta Ad Library Scraper: keyword search, advertiser audits by Page ID, and multi-country sweeps.
Nike’s ads, as recorded.
Nike’s public ads in the Meta Ad LibraryRecorded real run · 30 October 2026 · Nike’s public ads, US, active
Real public ads from the Meta Ad Library, collected by the author with his own tool, no login, no API key. Not live; status and dates are as of the run date.
Recorded, not live: I collected these ads on 30 October 2026 and kept them exactly as the collector returned them, so you can see what one run gives you.
Creatives are shown as text only. I left out image and video URLs because Meta’s signed URLs expire; each ad links to its own Ad Library page instead.
Still running means the ad’s stop date is on or after the run date, 30 October 2026; ended means it is before. Meta marked all 48 ads active at collection.
Loading the recorded run…
What I asked for.
These are the actor’s real input fields. I made this recorded run with the pip library meta-ads-collector 1.7.0, the same collector behind the actor, so the last column shows the equivalent values I used: page Nike, country US, active ads, 48 results.
modeKeyword search, an advertiser audit by Page ID, or the same keyword swept across many countries.This run: advertiser, the equivalent of a page searchqueriesOne or more keywords for keyword search and multi-country sweeps.This run: not setpageIdsFacebook Page IDs to audit in advertiser mode.This run: 15087023444 (Nike)countriesTwo-letter country codes, or ALL for every country. The one required field.This run: USactiveStatusActive ads only, ended ads only, or all.This run: activeadTypeEvery ad, or only Meta’s regulated categories (political and issue, employment, housing, credit), which carry spend and impressions disclosure.This run: not setmediaTypeAll media, image, video or meme.This run: not setpublisherPlatformsFacebook, Instagram, Messenger or Audience Network; empty means all.This run: not setcontentLanguagesLanguage codes of the ad copy; empty means all.This run: not setmaxAdsPerQueryA cap for each keyword, country and filter combination; the actor pages until it is reached or Meta has no more.This run: 48pageSizeAds asked for per GraphQL call; 30 matches the Ad Library’s own interface.This run: not setproxyConfigurationNeeded for cloud runs, because Meta rate-limits datacenter IPs.This run: none; I ran it locallymaxConcurrencyKept for backward compatibility with no effect: queries run one at a time, because Meta’s endpoint rate-limits concurrent traffic.This run: not set
Run it yourself.
Two routes to the same collector: the hosted actor, or the open-source package on your own machine.
On the Apify Store
Keyword search, advertiser audits and multi-country sweeps, with no setup. Pay per event.
22 users · Apify Store figures, 30 Oct 2026
Meta Ad Library Scraper on the Apify StoreAs a Python package
The CLI, the Python API, and an MCP server so an AI assistant can query the Ad Library for you. No Meta API key.
MetaAdsCollector on GitHubWhat this page does not show
- A single dayI recorded this run on 30 October 2026. Status and dates are as Meta reported them that day; the ads may have changed since.
- Text, not mediaCreatives appear as text only. Image and video URLs are signed by Meta and expire within days, so I left them out and link each ad to its Ad Library page.
- No personal fieldsI also left out bylines and the funding entity, payer and beneficiary fields, because they can name individuals, plus the bulk of the raw API fields.