Scoring methodology

Every score in this directory is a human judgement backed by public signals. We publish the rules so you can audit them, argue with them, and submit corrections when we get one wrong.

1. The 10-point quality bar

Ten discrete boxes, filled left to right. The palette moves from red (low quality / automated) through amber to emerald (exemplary human account). The tenth box carries a subtle glow because a perfect 10 is rare.

1Almost certainly a bot farm

2Heavy automation

3Mostly automated

4Low effort / slop

5Unrated / mixed signal

6Some human effort

7Mostly original

8Clearly original

9High quality creator

10Exemplary human account

2. AI usage verdicts

Using AI for Post

NoPartialYes

No = original human writing. Partial = AI assisted drafting, translation or lore summarising. Yes = templated or fully generated content.

Comment automation (script / bot)

YesNo

A separate, far more serious question than post writing: are the replies posted by a script or a bot?

Hard rule: comment automation is a bot, not "assistance"

Automating replies is the clearest bot signal we track, so it gets no benefit of the doubt. The moment an account is marked as automating its comments, three things are forced and cannot be overridden:

Account verdict is forced to "Full AI Bot"
Quality rating is capped at 4/10
Engagement is labelled "Botted" even if the measured ratio looks healthy

Why engagement cannot be trusted: a script that replies to 300-400 accounts a day reliably converts into 50-100 follow-backs. Those followers arrive because a bot knocked on the door, not because the account earned an audience, so a large views-to-followers ratio is manufactured rather than organic. Reading it as "Healthy" would reward the automation.

3. Engagement ratio

ratio = (average recent views ÷ followers) × 100
View counts come from X's public syndication payload. A healthy creator usually sees 2% or more of their follower base on a post; anything under 0.5% means the followers are not real people.

Healthy≥ 2%Reach is consistent with a real, responsive audience.

Low0.5% - 2%Weak organic reach. Often paired with inflated follower counts.

Botted< 0.5%Views are statistically impossible for the stated audience size.

4. Account verdict

100% HumanNo automation detected in posts or comments.

AI AssistedHuman-led, with AI drafting help. Never applied to automated commenting.

Full AI BotContent generation, replies or reach are predominantly automated. Any account with automated comments lands here.

5. Where the data comes from

01FxTwitter - public profile payload: display name, avatar URL, bio, follower and post counts, account creation date.
02X's public syndication timeline - recent posts and their view counts, which is how the engagement ratio is computed.
03The syndication follow-button widget - a numbers-only fallback for follower counts.
04Human review. No entry is published automatically - every score is set or confirmed by a reviewer.

Not affiliated with X Corp. We only use public, unauthenticated data. We never log into an X account, never pay for an API tier, and never host profile images - avatars are rendered directly from Twimg URLs stored as plain text. Spotted a mistake? Use + Snitch an Account to request a re-review.