See if your tweet will land — before you hit post.
Paste a draft. Get a quality score, the specific reasons it might get throttled, and a predicted view chart so you know whether the algorithm will push it. No vibe-check from a chatbot — every number is measurable and shows its work.
no signup required for first 3 scores
Free, no signup. Same scoring engine the paid app uses.
What different tweets actually score
Six real tweets, six real scores from the live engine. Click any card to see the full breakdown — or paste your own above.
Most local businesses don't have a lead problem. They have a follow-up speed problem. The first three minutes after a missed call decide whether you book or lose the job. What's your floor?
Three problems with how people pre-screen tweets today
Vibe-check tools are unfalsifiable
"This tweet looks good" from a chatbot is not a measurement. You can't backtest it, can't calibrate it, can't tell if it's wrong. TweetSim emits 18 action probabilities, composite dimensions, and a view-growth curve — every number is comparable post-to-post and verifiable against your real metrics.
"Engagement" isn't one number
A tweet that gets 50 likes and 0 replies is different from one that gets 10 likes and 30 replies. Twitter's algorithm weighs them differently. TweetSim breaks engagement into the actual phoenix-style action set (favorite, reply, retweet, profile_click, follow_author, dwell, share, …) plus the negative actions (mute, block, report) that throttle reach.
Nobody shows you the trajectory
Total likes after 24h is the wrong question — by hour 24 the algorithm has already decided whether to push your tweet. TweetSim predicts impressions over time (5min · 30min · 1h · 24h) so you see whether your tweet will hit the velocity threshold the "For You" algo gates on.
Four scoring layers, transparent
Open-source engine. Read the code, fork it, prove the math wrong. We publish the methodology and the calibration scorecard.
Phoenix-style action scorer
Predicted probability for each of 18 user actions, weighted by Twitter's leaked algorithm coefficients. Positive actions (favorite, reply, retweet, profile_click, follow_author, dwell, click) sum upward; negative ones (not_interested, mute, block, report) penalize sharply.
Composite quality scorer
10 dimensions on a 1-10 scale: niche_consistency, reply_trigger_score, dwell_time_potential, profile_click_potential, bookmark_potential, clarity, sales_relevance, annoy_risk, hook_strength, contrarian_edge. Combined into a single composite score.
Hard-block gate
Concrete reasons to reject: emoji, em-dash, AI-tell phrases (25 phrases including "in today's fast-paced", "harness the power"), engagement bait ("RT if you agree"), more than 1 hashtag, over 280 chars, link-only posts.
View-growth curve (Phase 1)
Gompertz-shaped impression trajectory with 80% confidence intervals at t+5min · 30min · 1h · 6h · 24h. Surfaces velocity score (views in first 30min / ceiling) — the single number that predicts whether the algorithm will keep pushing the tweet.
We launched today. No fake testimonials.
You won't find a logo wall on this site. The simulator works — that's verifiable in 30 seconds in the demo above. The engine is open source and runs publicly on every score call. As we collect real customer data, this section will fill in with attributable, real-name case studies.
In the meantime: read the methodology, run examples through the live API, fork the engine. The product is its own proof.
- ·130 passing tests across the engine, gate, and live API
- ·Open-source engine — twitter_scoring v0.3.0, MIT license
- ·Founder dogfood — Isaiah scores his own posts before publishing
- ·Live calibration scorecard — coming as soon as enough metrics accumulate
Pricing
Free tier covers casual use. Paid tiers unlock the persona system, calibration history, and the full audience panel.
Free
- ·20 scores / month
- ·5 generations / month
- ·Cold-start view curve
- ·No persistence
Builder
- ·500 scores / month
- ·100 generations / month
- ·30-day history
- ·Bring-your-own-persona
Pro
- ·Unlimited scores
- ·Audience panel (6 archetypes)
- ·Calibration scorecard
- ·Trained view-curve model
- ·Persona ingest from X / GitHub / Medium
Free tier is live. Paid tiers ship in v1 — see the roadmap post.
Score a draft now
No signup, no credit card. Paste a tweet, see the breakdown.
Score a tweet →