Alternative Social Media Trend Monitor & OSINT

Monitor public conversations across 4chan, Gab, GETTR, Telegram, and Pastebin and turn fresh posts into an evidence-linked OSINT dataset. This Actor detects related narratives, calculates transparent trend scores, reports source coverage, and preserves links to the public evidence behind every result.

Use it for alternative social media monitoring, emerging-narrative discovery, watchlist research, trust and safety triage, brand-risk analysis, and structured inputs for analyst or AI workflows.

Each run is independent. The Actor starts fresh, never reads previous runs, and does not build a cross-run profile or monitoring history.

Use cases

• Discover emerging narratives in a bounded sample of current public posts.
• Monitor phrases, hashtags, domains, brands, people, or issues with watchlist queries.
• Compare source breadth across 4chan, Gab, GETTR, Telegram, and Pastebin.
• Rank narratives using volume, engagement, recency, and within-run momentum signals.
• Export normalized evidence with source URLs for investigation and verification.
• Feed structured narrative intelligence into dashboards, reports, alerts, or AI agents.

Quick start

The default run is deliberately small and cost-controlled. It checks up to five current threads from 4chan's /g/ board, does not use AI summaries, and normally completes in well under five minutes.

{
  "mode": "discover",
  "sources": ["4chan"],
  "fourchanBoards": ["g"],
  "searchDepth": "quick",
  "includeAiSummaries": false
}

For a cross-platform watchlist, provide public accounts or channels for sources that require targets:

{
  "mode": "watchlist",
  "queries": ["example topic", "example.org"],
  "sources": ["gab", "gettr", "telegram", "pastebin"],
  "gabTargets": ["public_username"],
  "gettrUsernames": ["public_username"],
  "telegramChannels": ["public_channel"],
  "searchDepth": "quick",
  "includeEvidenceRows": true
}

Input options

Field	Description
mode	discover groups related records from the fresh sample. watchlist keeps evidence matching your queries.
queries	One to ten phrases, hashtags, or domains. Required in watchlist mode.
sources	Select any combination of 4chan, gab, gettr, telegram, and pastebin.
fourchanBoards	Public 4chan board names without slashes. Defaults to g.
gabTargets	Public Gab usernames, profile URLs, or post URLs. Required when collecting from Gab.
gettrUsernames	Public GETTR usernames. The @ prefix is optional.
telegramChannels	Public Telegram channel handles or t.me URLs. Private channels are not supported.
lookbackHours	Keep records within the last 1–168 hours when timestamps are available.
searchDepth	quick, balanced, or deep, capped at 10, 50, or 200 normalized records per source.
includeEvidenceRows	Add normalized evidence rows to the dataset. Narrative rows always include evidence IDs.
includeAiSummaries	Use optional AI-assisted narrative labels and summaries when available. Cluster membership remains deterministic.
Sources that require a username, URL, or channel are skipped when no target is supplied. Pastebin always uses fresh collection; cumulative indexing and search-engine discovery are disabled.

The Console keeps expert safety controls out of the main form. API clients can optionally set strictMode to fail on a source error or maxUpstreamChargeUsd to adjust the default $0.05 source-collection ceiling. Both retain safe defaults when omitted.

Dataset output

The default dataset contains three recordType values:

• narrative — detected narrative, summary, trend score, score coverage, keywords, contributing sources, and evidence IDs.
• evidence — normalized public source record with text, timestamp, engagement, source URL, and narrative ID. Returned when includeEvidenceRows is enabled.
• source_status — collection status, counts, reported upstream cost, and safe error details for each selected source.

Dataset views are convenient column layouts, not row filters. API clients should filter on recordType when they need only narratives, evidence, or source status.

Narrative example

{
  "recordType": "narrative",
  "rank": 1,
  "narrativeId": "nar_7dfb68f58c9dc94a",
  "title": "solar · battery · market",
  "summary": "4 fresh evidence items across 2 sources. Activity spans 2026-09-16T10:00:00.000Z to 2026-09-16T11:00:00.000Z. Largest source contribution: gab (3 records).",
  "trendScore": 71.4,
  "scoreCoveragePercent": 90,
  "evidenceCount": 4,
  "sourceCount": 2,
  "sources": ["gab", "gettr"],
  "evidenceIds": ["ev_12b...", "ev_91c..."],
  "evidencePreview": [
    {
      "evidenceId": "ev_12b...",
      "source": "gab",
      "sourceUrl": "https://gab.com/example/posts/123",
      "publishedAt": "2026-09-16T11:00:00.000Z",
      "textSnippet": "Public evidence excerpt supporting this narrative."
    }
  ],
  "keywords": ["solar", "battery", "market"],
  "matchedQueries": [],
  "firstSeenAt": "2026-09-16T10:00:00.000Z",
  "lastSeenAt": "2026-09-16T11:00:00.000Z",
  "sourceBreakdown": {"gab": 3, "gettr": 1},
  "aiEnhanced": false,
  "freshRunOnly": true
}

The run also provides these key-value-store records:

• OUTPUT — effective settings, top delivered narratives, source results, warnings, timing, billing, and consolidated costs.
• UPSTREAM_RUNS — source collection provenance, status, counts, and reported costs.
• AI_USAGE — optional AI model, token usage, reported cost, and fallback information.

In OUTPUT, analyzedEvidenceCount reports all fresh records used for analysis. deliveredEvidenceRowCount reports how many separate evidence rows were actually written, and datasetComposition provides the delivered narrative, evidence, source-status, and total row counts. This keeps partial pay-per-event runs unambiguous.

How trend scoring works

Each score uses only signals available in the current run:

• source breadth: 30%
• evidence volume: 25%
• engagement percentile within each source: 20%
• recency inside the requested lookback: 15%
• momentum inside the current run's time window: 10%

scoreCoveragePercent shows how much of the scoring model was available. Missing timestamps or engagement values lower coverage rather than silently becoming zero. Momentum compares newer and older portions of the same run window; it does not compare against previous runs.

Pricing and cost controls

The Actor uses pay-per-event pricing:

• $0.01 per delivered narrative result
• $0.00005 Actor start charge
• no separate charge for evidence or source-status rows

Apify's maximum run-cost setting is respected. If the remaining limit cannot cover every generated narrative, the Actor delivers only paid narratives and their matching evidence, then reports the result as partial. Optional AI work is limited to narratives that can be delivered.

Upstream source collection is separately protected by a default $0.05 ceiling and a stricter operator maximum. The inexpensive default uses one source, quick depth, no full evidence rows, and no AI summary.

Python API example

import os
from apify_client import ApifyClient

client = ApifyClient(os.environ["APIFY_TOKEN"])

run = client.actor(
    "thescrapelab/alt-social-narrative-trend-intelligence"
).call(
    run_input={
        "mode": "discover",
        "sources": ["4chan"],
        "fourchanBoards": ["g"],
        "searchDepth": "quick",
        "includeEvidenceRows": True,
        "includeAiSummaries": False,
    },
    max_total_charge_usd=0.10,
)

dataset = client.dataset(run["defaultDatasetId"])
for item in dataset.iterate_items():
    if item["recordType"] == "narrative":
        print(item["title"], item["trendScore"], item["sources"])

You can also run the Actor from Apify Console and download results as JSON, CSV, Excel, XML, RSS, or JSONL from the Dataset tab.

Privacy and responsible use

• The Actor processes public source material only.
• Every run starts fresh and does not consult previous Actor runs.
• It creates no named dataset, cache, task, schedule, webhook, or background monitor.
• Optional AI sends bounded excerpts of public evidence—up to eight 350-character excerpts for each of at most five deliverable narratives—to OpenRouter for labeling and summarization.
• AI cannot merge or split deterministic clusters and must cite existing evidence IDs.
• If AI is unavailable, deterministic narrative results are still returned.

Use source URLs and evidence rows to verify important findings. Follow applicable laws, platform terms, privacy obligations, and your organization's policies when using public data.

Limitations

• Results represent the bounded public records collected in one run, not population-level prevalence.
• Public sources may be unavailable, incomplete, deleted, rate-limited, or change their interfaces.
• Engagement values differ by platform, so engagement ranking is calculated within each source.
• Records without usable timestamps can remain eligible but reduce score coverage.
• Discover mode finds patterns only within the selected sources and search depth.
• The Actor does not identify coordinated behavior, intent, truthfulness, or causation.

Troubleshooting and support

• A skipped source usually needs a public username, URL, or Telegram channel.
• An empty watchlist result means the bounded fresh sample did not contain the query. Add relevant targets or increase search depth deliberately.
• Partial results remain usable. Review source_status rows and UPSTREAM_RUNS before retrying.
• If AI was requested but not used, review AI_USAGE; deterministic narratives remain available.

For reproducible support, include the Actor run ID, selected sources, non-sensitive input settings, and the affected source_status row. Never include API tokens or private credentials.