Facebook & Meta Ads Library Scraper + Tracker

Scrape Facebook and Meta Ads Library creatives, copy, CTAs and landing links — then detect new, updated and safely ended competitor ads on the next scheduled run.

Point it at a brand, a Facebook page or a keyword. Get back the full ad copy, creatives, call-to-action, destination links and delivery dates from the Meta Ad Library. Run the same query again later and every ad comes back labelled — new, changed, unchanged, or gone.

---

🎯 What can this Actor do?

🔎 Search by keyword, brand, page URL or Ad Library URL — mix up to 25 targets in one run

📝 Get the actual ad copy — body text, headlines, descriptions, captions and CTA, including every card of a carousel

🖼️ Get the creatives — direct image and video URLs, without paying to download the media itself

🔗 See where the ad sends people — the destination link the advertiser attached to the ad

🌍 Filter before you scrape — country, active status, ad category, media type and date range, so you pay only for the ads you want

🔔 Track changes over time — every ad is labelled NEW, UPDATED, UNCHANGED or ENDED against your previous run, and an ENDED is only ever raised from a snapshot proven complete on both sides

✅ Know when your data is complete — every target reports whether the source was exhausted, capped or only partially read

---

🆚 Baseline or Changes — which one do I need?

This is the only real decision you have to make. Everything else has a sensible default.

	BASELINE	CHANGES
What you get	Every matching ad in the current scan	The current scan with every row labelled NEW, UPDATED, UNCHANGED or safely proven ENDED
Answers	"What is my competitor advertising?"	"What did my competitor change?"
Best for	One-off research, creative swipe files, market scans	Daily or weekly monitoring, alerting, trend tracking
Needs a previous run	No	Second run onwards
Free-tier price	$0.02 run start + $0.70 / 1,000 delivered rows	$0.02 run start + $1.10 / 1,000 delivered monitoring rows

Start with BASELINE. Switch to CHANGES when you want the Actor to do the comparing for you.

---

⬇️ Input

A one-off creative scan — a keyword and nothing else:

{
  "searchTerms": ["running shoes"]
}

A competitor watch on two brands, tracking changes:

{
  "productLevel": "CHANGES",
  "searchTerms": [
    "nike",
    "adidas"
  ],
  "country": "US",
  "maxAdsPerTarget": 100
}

Filtered market scan:

{
  "searchTerms": ["protein powder", "creatine"],
  "country": "GB",
  "activeStatus": "ACTIVE",
  "mediaType": "VIDEO",
  "dateFrom": "2026-06-01",
  "maxAdsPerTarget": 100
}

You can paste a Meta Ad Library URL straight from your browser instead of using search terms — the Actor reads the filters out of the URL. Keyword searches and numeric-page-ID Ad Library URLs use the 256 MB default. An uncached Facebook handle such as facebook.com/nike may require the optional browser-continuity layer. That layer is deliberately disabled under the current three-event pricing and fails closed before browser work; use a numeric-page-ID Ad Library URL instead. Every target is capped at 100 rows; use multiple targets for larger batches.

---

⬆️ Output

One row per ad, per target. Real row from a live run, trimmed for length:

{
  "adId": "388120997678322",
  "pageName": "adidas",
  "pageId": "1497383683837785",
  "pageCategories": ["Product/service"],
  "pageLikeCount": 43617936,
  "state": "INACTIVE",
  "isActive": false,
  "startDate": "2024-09-26T07:00:00.000Z",
  "endDate": "2026-05-20T07:00:00.000Z",
  "bodies": [
    "Conheça e compre todas as coleções e os principais lançamentos da adidas pelo app!"
  ],
  "cardTitles": [
    "Chuteira Mundial Team Unisex",
    "Chuteira Samba Classic Homem",
    "Tênis Samba Classic Homem"
  ],
  "ctaTexts": ["Shop now"],
  "ctaTypes": ["SHOP_NOW"],
  "observedLinkUrls": [
    "https://www.adidas.com.br/chuteira-mundial-team/019228.html"
  ],
  "imageUrls": ["https://scontent.xx.fbcdn.net/v/t39.35426-6/..."],
  "publisherPlatforms": ["FACEBOOK", "INSTAGRAM"],
  "displayFormat": "DPA",
  "coverageStatus": "CAPPED",
  "changeEventId": "7b31a4f7b7b0c5aee47c1ddedb4de8b1973f3ef95b6a82d17da1b5eaab0dcf19",
  "changeStatus": "NEW",
  "observedAt": "2026-08-20T12:03:44.512Z"
}

Export as JSON, CSV, Excel, XML or RSS, or pull it straight from the Apify API.

Built-in dataset views

• changes — the monitoring view: status, changed fields, advertiser, dates and direct Ad Library link.
• creativeResearch — a visual swipe-file view with creative preview, copy, CTA, format and landing links.
• allData — every normalized field for API and spreadsheet work.

What is in every row

Group	Fields
Ad identity	adId, collationId, collationCount, adArchiveAdCount
Advertiser	pageName, pageId, pageCategories, pageLikeCount, pageProfileUri, pageProfilePictureUrl
Copy	bodies, cardTitles, cardDescriptions, cardCaptions
Call to action	ctaTexts, ctaTypes
Creatives	imageUrls, videoUrls, displayFormat
Destination	observedLinkUrls
Delivery	state, isActive, startDate, endDate, publisherPlatforms
Disclosure	reachEstimate, spend, currency, impressionsText, categories, isPolitical
Change tracking	changeEventId, changeStatus, changedFields, materialFingerprint
Run context	country, query, queryKind, coverageStatus, observedAt

Meta publishes reach and spend for political and EU ads only. For ordinary commercial ads those fields stay null — the Actor never invents a number to fill a column.

---

💰 How much does it cost?

Event	Price
Run start	$0.02
Baseline result	$0.70 / 1,000
Change result	$1.10 / 1,000

Real examples:

What you run	You pay
100 ads from one brand	~$0.09
1,000 ads across 10 keywords	~$0.72
Daily Changes watch on 3 targets, 100 ads each	~$0.35 per run

The run-start event is charged once by the Actor after valid input. Baseline rows are charged only when accepted into your dataset. In Changes mode, every delivered NEW, UPDATED, UNCHANGED and safely proven ENDED row is paid because every repeat performs a fresh live scan; state reads/writes, summaries, failed rows and suppressed rows are not charged. A stable changeEventId distinguishes normal scan revisions while identifying a rare post-delivery replay. A delivery guard blocks automatic re-billing if the change state cannot be committed. You are not billed separately for compute, network use or storage. The current price does not include browser continuity; activating that bounded layer requires a separate fixed event and a separate pricing approval.

Run the same Changes query sequentially, never concurrently against the same change store. If a run reports CHANGE_DELIVERY_RECOVERY_REQUIRED, leave its guard intact and inspect the prior run's dataset and charged-event read-back before support clears it. This deliberate fail-closed state avoids guessing whether a prior delivery was billed.

The prices shown are the Free-tier rates. Apify subscription tiers receive the lower rates displayed on the Pricing tab.

---

📊 Measured, not estimated

Every figure below comes from private cloud runs through 2026-08-24 and was read back from the Apify API after billing settled. These are measured platform costs, not buyer prices; source conditions still vary.

Test shape	Result	Cost per 1,000
One keyword, 5 ads at 256 MB (3 runs)	5/5 CAPPED each	$0.234–$0.250
Numeric-page-ID URL, 25 ads	25/25 CAPPED	$0.065
One keyword, 100 ads (3 runs)	100/100 CAPPED each	$0.057–$0.105
10 keywords × 100 ads at 256 MB	1,000 delivered, 937 globally unique	$0.030
Changes, first snapshot	60 NEW	$0.048
Changes, repeat run	60 UNCHANGED, zero false UPDATED/ENDED	$0.044
Historical guarded-continuity release samples, 50 ads	50/50 CAPPED; used to size the separate fixed event	$0.447–$0.491

Field coverage measured across the 1,000-row batch:

Field	Filled
Advertiser identity, state, start date	100%
Creative asset URL	99.3%
Ad body text	97.2%
Call to action	93.0%
Destination link	91.8%

Where Meta does not publish something, the field stays empty. No value is ever invented to fill a column.

✅ How do I know the data is complete?

Most scrapers hand you rows and let you guess whether that was everything. This one tells you, per target:

Status	Meaning
COMPLETE	The source ran out of ads. You have all of them.
CAPPED	Your result limit was reached first. More ads exist.
PARTIAL	Rows are valid, but neither the cap nor the end was reached.
EMPTY	Proven zero matching ads — not a silent failure.
CHALLENGED	Meta blocked the read. Reported, never disguised as empty.

This matters most for change tracking. An incomplete run can never produce an ENDED event — otherwise a slow network would look like your competitor switching off twelve campaigns. The Actor keeps the last known complete snapshot even after a partial run, so a temporary hiccup cannot corrupt tomorrow's comparison.

The trade-off is deliberate: this Actor would rather tell you an ad is still running than wrongly tell you it stopped. COMPLETE requires the source to be exhausted within the verified 100-row and 20-page per-target limits; increasing the number of targets does not turn a capped target into a complete one.

---

⚡ Automation ideas

Use your own private Apify Task or API call for a daily competitor watch, a weekly multi-target Changes report, or a one-off creative swipe file. Task creation, schedules and outbound alert workflows are separate setup steps; this release does not create or publish them.

Need Google ads too? Use the companion Google Ads Transparency Scraper & Change Tracker — same output philosophy and the same change-tracking guarantees.

🤖 For AI agents / MCP

This public, limited-permission Actor can be discovered and run through the hosted Apify MCP server. Use it when an agent needs current Meta competitor-ad creatives, source-backed ad fields, or a scheduled changefeed.

Example prompts:

• Track what ads Nike is running on Facebook and Instagram in the US.
• Return only meaningful Meta ad changes for these competitor pages.
• Build a source-backed Meta creative swipe file with copy, CTA, media, and landing links.

Connect the standard Apify MCP server; no Actor-specific MCP server is required.

---

❓ FAQ

Is scraping the Meta Ad Library legal? The Ad Library is a public transparency tool that Meta publishes deliberately, and it shows the same data to anyone without logging in. That said, laws differ by country and by what you do with the data afterwards. You are responsible for your own use — if you plan something commercial or large-scale, get your own legal advice.

Why not just use Meta's official API? Meta's ads_archive API only returns political and social-issue ads, plus housing, employment and credit ads in the US. Ordinary commercial ads have never been available through it outside the EU. For competitor research on normal brands, the Ad Library web data is the only source.

How far back does the data go? Meta keeps commercial ads visible for about 90 days and political ads for seven years. That is a Meta limit, not an Actor limit — which is exactly why change tracking is worth starting early: your own history keeps growing after Meta drops theirs.

Will I get duplicate rows? Deduplication happens inside each target. If the same ad matches two of your keywords, you get two rows — one per query context — because most buyers want to know which query found it. The run summary reports both delivered rows and globally unique ad IDs.

Can I run this on a schedule? Yes. Use Apify Schedules and set productLevel to CHANGES. Each run returns the monitored snapshot with every delivered row labelled, so an automation can filter NEW, UPDATED and ENDED while still retaining UNCHANGED context. Combine it with webhooks or the Apify integrations for Slack, Zapier or Make to get alerts.

What if Meta changes something? Every run reports coverage and route health. AUTO_GUARDED permits one bounded browser-continuity attempt only when its separate fixed PPE event is present at a safe runtime price. Under the current three-event pricing that event is absent, so the Actor fails closed before browser work instead of hiding an incomplete result or an unbounded cost. Choose HTTP_ONLY when you always want the lowest-cost route with no continuity attempt.

---

🔐 How your run is handled

• Live source access is restricted to the managed cloud runtime. Local execution fails closed before source access.
• Sensitive session and network details are not part of public rows or summaries.
• Media binaries are not downloaded — their URLs are already in the source data, so you get the links without paying for the bytes.
• The run summary records intended, charged, delivered, free and suppressed events, plus coverage and any controlled limit reached.

---

💬 Feedback

Found a field that should be there, a target type that fails, or a filter you need? Open an issue on the Actor's Issues tab. Bug reports with the run ID get fixed fastest.