Meta Content / Social-Listening Intel - Facebook + Instagram

Public-post social listening across Facebook + Instagram public Pages/posts. Search public posts by keyword/topic/hashtag/page, capture post text, engagement, timestamps, and the public author, then get the social-listening intelligence that matters: topic volume over time, sentiment + themes, top posts/pages, and cross-topic share-of-voice - in one normalized schema.

This is the commercial answer to a gated API. Meta's official social-listening path - the Meta Content Library (CrowdTangle's successor) - is application-restricted to academic and non-profit researchers. Brands, PR/comms teams, agencies, and journalists are categorically excluded. This actor delivers the equivalent public-post listening data from the logged-out public surface, no application required.

Scope - read this first. This actor is organic public-post social listening. It is deliberately distinct from two sibling actors:

• It is NOT ad-library-intel - that covers paid ads from the ad-transparency archives.
• It is NOT instagram-creator-intel - that covers creator / influencer audience analytics. If you want ads, use ad-library-intel. If you want creator metrics, use the creator actor. If you want to listen to what's being said about a topic/brand in public posts, you're in the right place.

---

What you get

A single dataset stream of normalized records, keyed by record_type:

record_type	What it is
post	One normalized public post: platform, public author, post type, text, engagement (likes/comments/shares/views), timestamp, hashtags, mentions, media URLs, permalink.
topic_rollup	Per-topic aggregation: post volume over time, sentiment split, top posts, top pages, top hashtags.
share_of_voice	Cross-topic table: each topic/brand's share of posts + engagement and its sentiment split.
access_notes	Per-platform access method, anti-bot escalation telemetry, PII posture, and responsible-use note.
platform_pending / fetch_error	Honest fail-soft notes when a logged-out surface was blocked (Meta login-walls automated clients). Never fabricated data.
monitor_digest	On scheduled runs: new posts + volume/sentiment shift since the last run.

Five dataset views are pre-defined: overview (AI-agent / analyst skim), posts (every field), topics (rollups), an MCP-compatible JSON link, and a CSV download.

---

Modes

Mode	Input	Does
keyword_listening (default)	keywords	Search public posts by topic across platforms, then aggregate (volume, sentiment, top posts/pages, share-of-voice).
page_posts	pages	Pull recent public posts for given Facebook Pages / public Instagram profiles (brand/competitor owned content).
hashtag_intel	hashtags	Pull public hashtag feeds + topic-volume / top-posts intel.

Note on keyword search: Facebook and Instagram have no logged-out keyword post search (their search UIs require login). In keyword_listening mode a keyword is therefore matched against its hashtag feed on each platform - the closest public, logged-out equivalent. For owned-content monitoring use page_posts.

Example input

{
  "mode": "keyword_listening",
  "keywords": ["sustainability", "plant based"],
  "platforms": ["facebook", "instagram"],
  "max_results_per_platform": 30,
  "include_listening_analysis": true,
  "include_topic_rollup": true,
  "pii_minimization": true
}

---

Legal-safety & responsible use

This actor follows the portfolio's gated-data playbook (Section B):

• Logged-out, public-only. No login, no account, no user cookies, no paid token. Only public Pages/posts and public profiles/hashtag feeds.
• PII minimization (on by default). Posts authored by an apparent personal individual (not a Page / business / verified account) have the author name dropped and the handle replaced with a stable opaque token (person_xxxx), and the author id stripped. Pages/business/verified accounts (the brands and orgs social listening is about) are retained. The dataset is about public discourse on a topic, not a profile of private individuals. Turn off only with a lawful basis (pii_minimization=false).
• Polite rate limits and bounded concurrency.
• Fail-soft. When a surface blocks the logged-out request, the actor emits a documented platform_pending note - it never fabricates data.
• Favorable precedent. Meta v. Bright Data (2024) held that logged-off scraping of public Meta data is not barred by Meta's terms.

Use this for aggregate brand/topic listening, reputation monitoring, PR/comms, and trend research - not for surveillance of individuals.

---

Anti-bot escalation

Meta's public surfaces are token-gated and login-wall automated clients. The actor runs a four-tier ladder per request:

1. httpx over the DATACENTER proxy (cheapest, first).
2. curl_cffi with real Chrome TLS impersonation over the RESIDENTIAL proxy (defeats JA3/TLS fingerprinting).
3. patchright / playwright stealth browser over the RESIDENTIAL proxy - loads the public page and replays its own JSON XHR (Instagram web_profile_info / tags/web_info; Facebook GraphQL). Runs on the apify/actor-python-playwright base image (headful under Xvfb).
4. Fail-soft - a documented platform_pending / fetch_error note.

For the most reliable results against managed challenges, point browser_cdp_url (or the BROWSER_CDP_URL env var) at a warm anti-detect browser; the browser tier connects over CDP and inherits its session + IP.

---

Pricing (pay-per-event)

Event	Price	What
post_record	$0.004	One normalized public post.
listening_analysis	$0.008	Sentiment + theme tagging per post.
topic_rollup	$0.010	Per-topic volume/sentiment/top-posts/SOV aggregation.
scheduled_delta_run	$0.050	One monitor-mode delta digest.

A run that returns nothing costs nothing. A 5-layer cost-control defense (pre-flight caps, demo-mode soft-fail, _RunBudget guard, in-loop budget checks, hardcoded prices) protects against runaway compute.

---

Monitor mode

Save your input as a task and put it on an Apify Schedule. On scheduled runs the actor compares against the previous run and emits a monitor_digest (new posts, volume + sentiment shift). Set monitor_webhook_url to post the digest to a Slack-compatible webhook.

---

MCP / AI-agent use

An MCP twin, mcp-meta-content-intel, exposes this actor as AI-agent tools (search_public_posts, get_page_posts, get_hashtag_posts, analyze_topic, track_topic_mentions) and is x402 (USDC on Base) + Skyfire ready for token-less agentic payments. Or wire the overview view directly via the MCP-compatible dataset link in the output.

---

Who it's for

Brands and DTC companies monitoring their reputation, PR/comms teams tracking narrative and sentiment, agencies running social-listening for clients, journalists and researchers studying public discourse - everyone locked out of Meta Content Library's academic-only gate.