Thread reconstruction
Stitch fragmented reply chains back into the original argument, even after deletions or rate-limit gaps. Read the conversation the way it was meant to be read.
Explore threads
Forensic X Analytics
Decode every conversation on X with clarity and precision.
TweetBlocker turns the noise of timelines, threads, and quote-posts into structured intelligence. Track influence, detect manipulation, and surface what actually matters — in minutes, not weeks.
2.4B+
Public posts analyzed every month
94%
Bot-network detection precision
37k
Verified researchers and analysts
11min
Average time from query to insight
Live X Analytics Teaser
Watch a conversation unfold as it happens.
Our streaming pipeline ingests X's public firehose the moment content surfaces — so dashboards update in seconds, not at the next refresh cycle.
Streaming now · 18,402 events / min
Trending thread: AI regulation debate
A policy thread crossed the 10K-reply threshold 14 minutes ago. TweetBlocker has already clustered 312 sub-arguments, identified 4 high-credibility voices, and flagged 2 inauthentic amplification patterns.
Open this conversationFeature Highlights
Six lenses on the same conversation.
Every post, reply, and quote-post can be read in a dozen different ways. TweetBlocker gives you the six views that actually change a decision — built for analysts, not dashboards.
Velocity & reach
Track how a post propagates across followers, lists, and communities in real time. Spot the inflection point where organic traction flips into coordinated push.
See velocity mapsBot & network detection
Multi-signal classifiers flag sock-puppets, amplification rings, and coordinated inauthentic behavior — with explainable confidence scores, not opaque black-box scores.
Read the modelSentiment drift
Measure how tone shifts across a thread, a community, or a 30-day window. Separate signal from the emotional weather of a feed.
Browse insightsInfluence mapping
Rank accounts by real audience quality, not raw follower counts. See who actually moves replies, not who shouts loudest.
Map influenceExport & API
Push clean, normalized data straight into your warehouse, BI tool, or investigation notebook. CSV, JSON, Parquet, or webhook — your choice.
View API plansWhy TweetBlocker
Built for analysts who need evidence, not vibes.
Most social tools stop at engagement counts. We treat X as a forensic dataset — every claim traceable, every score explainable, every export audit-ready.
01
Explainable by default
Every classification comes with the signals that triggered it. No "trust the model" — you see the receipts, adjust the thresholds, and defend the conclusion.
02
Streaming-first architecture
Our pipeline processes events as X publishes them, not in 15-minute batches. You see the conversation while it's still a conversation.
03
Researcher-grade provenance
Each datapoint carries its capture timestamp, source endpoint, and transformation lineage — ready for publication, litigation, or peer review.
04
Privacy-respecting by design
We analyze public posts only, hash identifiers at the edge, and never enrich with off-platform data. Compliance teams actually approve us.
How it works
From raw post to defensible insight in five steps.
No black boxes. No magic. A transparent pipeline you can inspect at every stage — and reproduce, if needed, six months from now.
STEP 01
Ingest
We tap the public X firehose and authenticated API streams, with multi-region failover.
STEP 02
Normalize
Posts, replies, quotes, and media get a canonical schema — language-tagged, de-duplicated, versioned.
STEP 03
Enrich
Bot scoring, thread reconstruction, entity extraction, and topic clustering layer on top of the raw signal.
STEP 04
Index
Everything lands in a queryable archive optimized for sub-second investigation across billions of posts.
STEP 05
Surface
Dashboards, exports, and webhooks deliver the result to the analyst, the newsroom, or the SIEM.
Methodology credibility
A model you can audit, not just trust.
TweetBlocker's detection stack has been peer-reviewed, red-teamed, and stress-tested against adversarial datasets. Here's how it performs in the wild.
Independent validation, public benchmarks.
We publish quarterly model cards, release held-out evaluation slices, and submit to the academic community for external reproduction. When a metric moves, you'll see the diff.
Our bot-classifier was independently verified at 0.94 precision / 0.91 recall across 14 languages and 6 platform dialects — including the long tail of inauthentic behavior that's hardest to detect.
0.94 P
Bot-detection precision
14+
Languages evaluated
Q1'26
Latest model card
3×
External red-teams / year
In the field
Analysts, journalists, and trust & safety teams run on TweetBlocker.
From breaking-news desks to platform-integrity teams — here's what changed when they moved from manual monitoring to forensic X analytics.
"
TweetBlocker replaced three internal tools and a graveyard of Zapier workflows. We trace a viral claim back to its origin in under fifteen minutes — used to be half a day.
"
The bot-detection explainability is the part I never thought I'd get. When we flag an account, our trust & safety committee can see exactly which signals tripped the score. That changes the legal conversation completely.
"
We integrated the API into our election-monitoring dashboard in a weekend. Real-time thread reconstruction during a primary was the single most useful feature I've deployed in five years.