The LEGIT Engine™

How a Legit Score works

A Legit Score is a 0–100 confidence opinion on how real and responsive the audience behind an account is. It always comes with a confidence band, a tier and the top reasons behind it, each with a fix.

Engine version 1.0.0-dev. Every score is stamped with the engine version, sample sizes and data sources, so it can be re-explained later.

The seven-stage pipeline

1Acquire
Pull the public profile, recent posts (views, likes, comments, shares, timestamps), recent comments and a statistically sound sample of the audience.
2Classify
Place every sampled follower and engager into one of six Audience Cohorts, as a probability rather than a hard label.
3Measure
Compute around 40 audience signals across the five LEGIT pillars.
4Normalise
Turn each signal into a 0–100 Peer Percentile against accounts of the same platform and size, using the Legit Benchmark Graph™.
5Compose
Combine the pillars with fixed weights, then apply Integrity Modifiers so a single clear pattern can't be averaged away.
6Calibrate
Move the score towards caution when evidence is thin, producing the final score and its confidence band.
7Explain
Rank what moved the score and show the top 3–5 Legit Reason Codes™, each with a fix.

The five LEGIT pillars

Each pillar answers one question about the audience. The weights are fixed and public.

Pillar	Question it answers	Weight
LListener Authenticity	Is the audience made of real people?	30%
EEngagement Integrity	Does the engagement behave like real people?	25%
GGrowth Genuineness	Was the audience grown honestly?	20%
IInteraction Network	Is the audience reachable and in the right place?	15%
TTraffic Truth	Do the views look like real watching?	10%

Pillars are compared against peer accounts of the same platform and size. If a pillar can’t be measured for an account (for example, no view data), the remaining weights are rescaled rather than guessing.

The six Audience Cohorts

Each sampled follower and engager is classified with a probability, not a hard label. No individual is named or published; only the overall mix is reported.

C1Active Real
Real people who engage.
C2Passive Real
Real people who are quiet.
C3Mass Followers
Real people who follow so many accounts they rarely see posts. Counted against reach quality, never against authenticity.
C4Notables
Creators and brands in the audience.
C5Suspicious
Accounts that look inactive or low quality.
C6Automated
Accounts showing strong patterns common to automated accounts or purchased engagement.

Confidence, not certainty

Scores are built from samples, so every score carries a confidence band (for example, 78 ±6). Small samples can’t produce a top score by luck, and new or small accounts aren’t marked down just for having little data. When evidence is thin the band widens and the tier is capped.

Each score also shows its Data Depth: Public Sample (public data and samples only), Deep Sample (a larger sample from a deep audit) or Connected Data (the creator connected their account through the official platform login).

Tiers

Score	Tier	What it tells a sponsor
85–100	Exceptional	Premium sponsorships; expect full real-audience delivery.
70–84	Verified Commercial	Standard campaigns.
55–69	Moderate Variance	Pay on performance or discount the rate; use tracked links.
40–54	Low Confidence	Performance-only deals.
0–39	Critical Variance	Not recommended for paid campaigns.

Legit Reason Codes™

Every score lists the top 3–5 reasons behind it. Reasons describe patterns in the audience, never accusations, and each comes with a fix. Some examples:

LA-01

Elevated share of mass followers

Fix: Leave follow-for-follow and giveaway loops; mass followers rarely see your posts.

LA-02

High share of incomplete profiles in the audience sample

Fix: Remove inactive followers using the platform's tools.

EI-01

Low comment variety / repetitive comments

Fix: Leave engagement groups and giveaway loops; ask specific questions in captions to invite real replies.

EI-02

Engagement pattern unusual for this size

Fix: Avoid engagement services; consistent organic posting brings the pattern back in line over time.

GG-01

Unusual follower growth acceleration

Fix: If a spike came from press or a collaboration, add that context through the dispute process.

IN-01

Clustered early interactions

Fix: Leave engagement pods; the same group engaging first on every post lowers confidence.

IN-02

Audience location doesn't match the content's market

Fix: Create for your core market and connect your account to show real audience geography.

TT-01

View patterns stall near common payout thresholds

Fix: Use only organic distribution and connect your account to verify view analytics.

TT-02

Views rising without matching engagement

Fix: Avoid view-boosting services and connect your account to verify view analytics.

Format-Neutral Scoring™

We score the audience, never the creator’s format. Human, faceless and AI-generated creators are scored on the same footing. Having no face on camera, a synthetic voice or AI visuals is never an input. Only audience behaviour counts.

Disputing a score

Disputes are free and never depend on a subscription. There are four routes:

Re-audit: We re-score the account automatically with a fresh, non-overlapping sample.
Connect Data: You connect your account through the official platform login and we re-score it on your real stats.
Context: You explain something we can't see, such as press coverage or a collaboration.
Analyst review: A person checks the evidence and can exclude activity aimed at your account that you didn't cause.

We aim to acknowledge a dispute within 72 hours and resolve it within 14 days. While a dispute is open, the score shows “Under review”. Open a dispute or opt out.

What we keep private

Like a credit bureau’s scorecard, some details stay private so the score can’t be gamed: individual signal thresholds, modifier triggers and factors, classifier models, benchmark data, and sampling seeds and schedules. These change only with a new engine version.