Skip to main content
๐Ÿ”ฌ

Current AI Analysis Method

This page documents the implemented inputs, formula, and pre-publication evidence checks. External data that was not retrieved is not treated as evidence.

The unusual-review signal is the average of five published rules over retrieved Steam reviews. Model-returned numeric judgments are not blended into it. AI input is limited to 40 representative reviews, and core claims are checked against exact source quotes.

โš–๏ธ Current Implementation Summary

ItemCurrent processingLimit
AI inputUp to 40 representative reviewsNot the entire retrieved corpus
Core claimsReview IDs and exact quotes are checkedFailed results are not published
External comparisonMetacritic / OpenCriticUsed only after successful retrieval
Optional analysisAccepted only with required text or metadataMissing data is not filled by inference
Display languagesJapanese, English, Korean, ChineseCultural nuance can still be missed

๐Ÿ“ก Data Sources

The primary input is review text that was successfully retrieved. External scores are used only when retrieval succeeds.

๐ŸŽฎ
Steam Reviews and Game DataPRIMARY

For Steam titles, retrieved review text, verdict, playtime, and helpful votes provide the primary evidence.

๐Ÿ“
Metacritic ReviewsCONDITIONAL

For non-Steam titles, successfully retrieved reviews form the primary corpus. The analysis is rejected when that corpus cannot be retrieved.

๐Ÿ“Š
Metacritic / OpenCritic ScoresOPTIONAL

Used for comparison only when retrieval succeeds and usable values exist. They are not added to the unusual-posting signal formula.

๐Ÿ” Unusual-Review Signal Calculation

The published number is the retrieved-corpus average of five rules. AI text analysis separately produces the rank, summary, and evidence and cannot overwrite this number.

STEP 1 โ€” Rule-Based Score (Per Review)

Five rule indicators are applied to each review in the retrieved corpus, then averaged.

Playtime < 1 hourโ€” Retrieved value is in the short-playtime band
+30
Review text < 50 charsโ€” Retrieved text is short
+20
Owned games < 5โ€” Retrieved owned-game count is low
+15
Reviews posted > 100โ€” Retrieved posted-review count is high
+25
Votes received = 0โ€” Retrieved helpful-vote count is zero
+10
Cap: 100 pts | Retrieved-corpus average = Unusual-Review Signal
STEP 2 โ€” Evidence-Bound AI Text Analysis (Separate)

After removing duplicates and overly short text, up to 40 representative reviews are selected with positive/negative balance and sent to Gemini.

โ—Each overall verdict, pro, and con must bind to a valid input review ID and an exact source quote.
โ—Unknown review IDs, quote mismatches, insufficient evidence, and undersized explanations fail the publication check.
โ—Verdict/text contradictions, playtime, and repeated phrasing are considered only when visible in the supplied input.
โ—Optional language-gap, wording-pattern, GPU, and contradiction modules are accepted only when their required text or metadata exists.
STEP 3 โ€” Publication Boundary
Published signal = Five-rule averageModel-supplied numeric overrides rejected

This is a 0โ€“100 heuristic score, not a percentage, probability of manipulation, or proof of fraud.

๐Ÿ“ 6-Axis Quality Assessment

The six 0โ€“100 axes are estimates from the representative review excerpts. An axis with little review coverage is more uncertain and is not a measurement.

GRAPHICS
Graphics

Visual quality, art style, framerate stability

GAMEPLAY
Gameplay

Controls, mechanics depth, replayability

STORY
Story

Narrative quality, world-building, characters

VALUE
Value

Content amount vs price point

VOLUME
Volume

Playtime, replayability, endgame content

STABILITY
Stability

Bug frequency, crash rate, optimization

๐Ÿ† Overall Rank System

S
Very high estimate

Several estimated axes are high in the representative sample. Check cited evidence and personal fit before buying.

A
High estimate

Most axes are high, but opinions outside the retrieved sample may be absent.

B
Upper-middle estimate

The sample contains both clear strengths and concerns worth checking.

C
Middle estimate

Strengths and weaknesses are roughly balanced in the sample.

D
Low estimate

The representative reviews contain several material concerns. Check the cited reviews.

๐Ÿงฌ Special Analysis Modules

JPโ€“EN Sentiment Gapevaluation_gap

Estimates differences within the retrieved sample only when enough Japanese and English reviews are present.

Supplementary Wording-Pattern Notesai_generated_suspicion

Summarizes repetition or formulaic wording in cited text. It does not classify authorship or generation method, and uncalibrated percentages or confidence labels are not published.

GPU Performance Reportsspec_analysis

Organizes concrete GPU, FPS, and fault reports only when they occur in the input reviews.

Review Contradiction Detectionreview_contradictions

Shows verdict/text contradictions only when an exact supporting quote is available.

โš™๏ธ AI Model & Technical Specs

AI EngineGemini 2.5 Flash-Lite is attempted first; Gemini 3.1 Flash-Lite may be tried within the request deadline when needed.
AI InputUp to 40 representative reviews after removing text shorter than 12 characters and duplicates, with positive/negative balance.
Evidence ValidationOverall reasoning, pros, and cons are checked against valid review IDs and exact source quotes. Failed results are not published.
External ComparisonMetacritic / OpenCritic are used only after successful retrieval; failures and missing values are treated as no_data.
Optional ModulesEach module is accepted only when its required language, hardware, or contradiction evidence is present.
API Key SwitchingConfigured keys are tried within the request deadline for eligible authentication or rate-limit responses.

โš ๏ธ Limitations & Disclaimers

โ‘  Estimates only:The unusual-posting signal, hardware-related estimates, and six-axis scores are derived from retrieved data, rules, and/or AI. The signal is not a percentage or proof of wrongdoing, and none of these values is guaranteed accurate.

โ‘ก Review sampling:AI input is limited to 40 representative reviews. When the retrieved corpus is larger, omitted opinions may not be reflected in the result.

โ‘ข AI limitations:Gemini excels at context but may not fully understand sarcasm, jokes, or cultural nuances.

โ‘ฃ Point-in-time and external data:Analysis reflects retrieval and cache time. Metacritic / OpenCritic comparisons and optional modules appear only when their required data was retrieved.

โ“ FAQ

Q. How is the unusual-review signal calculated?

Five rules are averaged over the retrieved review corpus. Up to 40 representative reviews are sent to AI for the separate rank and summary, but model-supplied numbers are not added to this signal.

Q. How are AI claims verified?

Overall reasoning, pros, and cons must bind to valid input review IDs and exact matching source quotes. Unknown IDs, quote mismatches, and insufficient evidence are rejected before publication.

Q. What are the 6-axis criteria?

Graphics, Gameplay, Story, Value, Volume, and Stability are estimated from representative reviews. Axes with sparse coverage are uncertain rather than measured facts.

Q. Are external scores and optional analyses always used?

No. Metacritic / OpenCritic comparisons run only after successful retrieval, and each optional module requires the relevant review text or metadata.

๐Ÿ“‹ Methodology Update Record

Changes that affect the mapping between public explanations and implementation are recorded here.

DateChangeImpact
2026-07-20Removed unacquired sources and unsupported forecast fields from public explanationsAccuracy
2026-07-20Documented the 40-review maximum and exact-quote claim validationTransparency
2026-07-20Limited external comparisons and optional modules to available required dataGrounding
๐Ÿ” Analyze a Game
How AI Analysis Works | Data Sources, Scoring Formula & Methodology