ThemysThemys/Docs

How Attention Levels Work

Understand attention levels, clause categories, severity levels, and how review guidance is generated.

Attention Levels

Every scan assigns a qualitative attention level. The label is meant to help you decide where to focus your review, not to make a legal risk determination.

LabelColorMeaning
StandardGreenTypical clauses with lower review priority
Worth reviewingAmberSome clauses are worth reading before agreeing
Needs attentionRedClauses deserve careful review before accepting

Treat every attention level as review guidance, not legal advice or an authoritative legal risk determination.

Severity Levels

Flagged clauses use severity cues to explain why something was highlighted:

  • Info (blue): Standard clause that's worth noting but not necessarily alarming.
  • Caution (amber): Clause that could affect you in specific situations.
  • Warning (red): Clause that gives the service significant power over your data, content, or legal rights.

Categories

Clauses are classified into six categories, each targeting a different area of concern:

CategoryWhat It Covers
Data PrivacyData collection, sharing with third parties, tracking, retention policies
ArbitrationMandatory arbitration, class action waivers, jury trial waivers
LiabilityLimitation of liability, indemnification, warranty disclaimers
Content/IPContent licensing, IP ownership transfer, broad usage grants
BillingAuto-renewal, hidden fees, unilateral price changes, cancellation hurdles
AccountUnilateral termination, account suspension, data deletion on termination

How to Use the Guidance

Start with clauses marked Warning, then read findings that involve the categories most important to you. Use the full extraction to find the relevant section, and verify important findings against the original legal document.

Example

Consider a ToS with two major findings:

  • Mandatory arbitration with class action waiver in the Arbitration category
  • Broad data sharing with third parties in the Data Privacy category

Those signals may produce an overall Worth reviewing attention level, with the arbitration and data-sharing clauses highlighted first.

Color Coding Reference

Results use consistent color coding throughout the web app and the planned

extension (coming soon):

Attention LevelBadge Color
StandardGreen
Worth reviewingAmber
Needs attentionRed

These colors appear in attention badges, category breakdowns, and the overlay UI to provide at-a-glance review guidance.