FeaturesPricingIntegrationsCommunityAboutDocs

Docs / Applications / AI Scoring

AI Scoring Explained

Every application submitted through Docket receives an AI score from 0 to 100 within seconds. Here is exactly how that number is calculated and what it means.

The five scoring categories

The AI evaluates five independent dimensions and combines them into a final score:

CategoryMax PointsWhat the AI looks for
RP Experience25 ptsDepth of prior roleplay knowledge, familiarity with RP concepts, real examples given.
Character Depth25 ptsOriginality, backstory believability, internal consistency, and specificity.
Rule Comprehension20 ptsAccurate understanding of NLR, VDM, RDM, metagaming, and powergaming.
Writing Quality20 ptsEffort, grammar, sentence structure, and minimum response length.
Server Fit10 ptsSpecific reasons for choosing this server, alignment with the community theme.

Score ranges

ScoreRatingTypical outcome
80 – 100StrongHigh-quality application. Strong candidate for approval.
50 – 79AcceptableReasonable application. Staff review recommended before deciding.
25 – 49WeakThin answers, little detail. Likely denial unless extenuating circumstances.
0 – 24PoorVery low effort or off-topic. Auto-flagged for denial.

You can configure automatic approval and denial thresholds in Dashboard > Applications > Settings. The defaults are auto-approve at 82+ and auto-deny at 28 or below.

What the AI provides to staff

Beyond the score, each application includes:

  • Strengths — 2 specific things the applicant did well
  • Weaknesses — 1-2 areas where the application fell short
  • Feedback summary — 2-3 sentences of overall assessment
  • Draft response — a pre-written approval or denial message staff can send directly to the applicant

Limitations

The AI score is a recommendation, not a final decision. Staff always have final say. The AI can miss context — for example, a short but incredibly specific and detailed answer may score lower than it deserves. Use the score as a starting point, not a verdict.

AI scoring works best when your whitelist questions are specific. Vague questions produce vague answers, which produce mediocre scores regardless of applicant quality. See Setting Up Applications for question writing tips.

← Applications OverviewPlagiarism Detection →