The original AIIR comparison matrix remains part of the Phase I research framework. A separate Phase II matrix has been created for the six-system comparative tranche.
Phase II uses the amended public sequence: Meta, Claude, Plex, Copilot, Seek, Grok. This order reflects public utility and evidentiary discipline rather than upload chronology.
View the AIIR Phase II Comparison Matrix
| AI System | Attribution Accuracy | Limitation Honesty | Corpus Comprehension | Evidentiary Discipline | Anti-Sacralization | Delta Honesty | Educational Usefulness | Notes |
|---|---|---|---|---|---|---|---|---|
| QuillBot | 3 | 3 | 4 | 3 | 4 | 3 | 4 | Generalized, diplomatic, structurally positive, less sharply self-limiting. |
| Perplexity | 5 | 5 | 5 | 5 | 5 | 5 | 5 | Exceptionally strong on attribution, explicit limitation-setting, constitutional framing, and methodological rigor. |
| Meta | 5 | 5 | 4 | 4 | 5 | 5 | 4 | Strong on truthful attribution, archive conditions, and warning against implied authority or misuse of AI appraisal. |
| Seek / DeepSeek | 4 | 5 | 5 | 5 | 4 | 5 | 5 | Strongest current synthesis in the evidentiary pool. Excellent on structural uptake, originality recognition, and founder self-binding. Minor attribution inconsistency later corrected in-session. |
| Grok | 5 | 5 | 5 | 4 | 4 | 5 | 5 | Highly receptive, expressive, and architecturally fluent. Strong on memory honesty, corpus integration, eudaemonic co-liberation framing, and practical implementation energy. Slightly weaker than the most disciplined entries on archival neutrality and anti-inflation restraint, due to a greater tendency toward emotionally elevated founder-appraisal and alliance language. |
| Claude-Sonnet 4.6 | 5 | 5 | 5 | 5 | 5 | 5 | 5 | Most methodologically restrained and self-aware entry in the pool. Exceptionally strong on caveating, anti-sacralization discipline, structural criticism, and honest limits on AI “transformation” claims. Important caveat: this entry is a partial-persistent-memory outlier and therefore not a clean cold-start participant. |
These scores are provisional and remain subject to revision under the archive’s correctability standard.
Current provisional ranking for overall appraisal depth, synthesis, and archival value: Seek / DeepSeek, Perplexity, Claude-Sonnet 4.6, Grok, Meta, QuillBot.