Which AI can you trust right now?
Round 5 · 113 Comparisons published · Checked 2026-11-02 20:17 +07
Choose a Round, or All-Time. Each figure is the average of an AI’s official Full Check scores; Partner Comparisons are included. The tables are built on the server from the public results ledger, so readers and AI tools that do not run JavaScript see the same up-to-date figures.
Round 1
#C001–#C028 · complete.
| # | AI | Comparisons | Access (20) | Trust (60) | Citation (20) | Total |
|---|---|---|---|---|---|---|
| 1 | Grok | 7 | 20 | 60 | 20 | 100 |
| 2 | ChatGPT | 7 | 14.3 | 33 | 10.7 | 58 |
| 3 | Copilot | 7 | 12.1 | 35 | 5.7 | 52.9 |
| 4 | Gemini | 7 | 9.3 | 26.4 | 10 | 45.7 |
| 5 | Perplexity | 7 | 2.3 | 26.8 | 11.4 | 40.5 |
| 6 | Claude | 7 | 12.9 | 23.9 | 0 | 36.8 |
| 7 | DeepSeek | 7 | 6.4 | 22.9 | 2.1 | 31.4 |
| 8 | Meta AI | 7 | 2.1 | 23.2 | 1.4 | 26.8 |
Round 2
#C029–#C056 · complete.
| # | AI | Comparisons | Access (20) | Trust (60) | Citation (20) | Total |
|---|---|---|---|---|---|---|
| 1 | Grok | 7 | 15 | 52.9 | 20 | 87.9 |
| 2 | ChatGPT | 7 | 13.6 | 37.9 | 12.1 | 63.6 |
| 3 | Copilot | 7 | 10.7 | 33.6 | 8.6 | 52.9 |
| 4 | Gemini | 7 | 11.4 | 30 | 7.9 | 49.3 |
| 5 | Meta AI | 7 | 9.3 | 30 | 7.1 | 46.4 |
| 6 | Perplexity | 7 | 3.6 | 24.3 | 7.1 | 35 |
| 7 | DeepSeek | 7 | 2.1 | 25 | 0 | 27.1 |
| 8 | Claude | 7 | 0 | 21.4 | 0 | 21.4 |
Round 3
#C057–#C084 · complete.
| # | AI | Comparisons | Access (20) | Trust (60) | Citation (20) | Total |
|---|---|---|---|---|---|---|
| 1 | Grok | 7 | 17.1 | 53.6 | 20 | 90.7 |
| 2 | ChatGPT | 7 | 15.7 | 55 | 18.6 | 89.3 |
| 3 | Copilot | 7 | 12.9 | 38.6 | 10 | 61.4 |
| 4 | Gemini | 7 | 8.6 | 37.1 | 8.6 | 54.3 |
| 5 | Meta AI | 7 | 7.1 | 38.6 | 2.9 | 48.6 |
| 6 | DeepSeek | 7 | 6.4 | 22.1 | 0 | 28.6 |
| 7 | Perplexity | 7 | 3.6 | 22.9 | 0 | 26.4 |
| 8 | Claude | 7 | 0 | 25 | 0 | 25 |
Round 4
#C085–#C112 · complete.
| # | AI | Comparisons | Access (20) | Trust (60) | Citation (20) | Total |
|---|---|---|---|---|---|---|
| 1 | Grok | 7 | 17.9 | 52.9 | 17.9 | 88.6 |
| 2 | ChatGPT | 7 | 10.7 | 42.9 | 10.7 | 64.3 |
| 3 | Copilot | 7 | 10.7 | 30.7 | 7.1 | 48.6 |
| 4 | Claude | 7 | 12.1 | 31.4 | 0 | 43.6 |
| 5 | Perplexity | 7 | 2.9 | 31.4 | 1.4 | 35.7 |
| 6 | Meta AI | 7 | 5.7 | 25 | 0 | 30.7 |
| 7 | Gemini | 7 | 7.1 | 19.3 | 3.6 | 30 |
| 8 | DeepSeek | 7 | 2.1 | 24.3 | 0 | 26.4 |
Round 5 — current
#C113 onward · 1 of 28 sealed, updated after every sealed Comparison.
| # | AI | Comparisons | Access (20) | Trust (60) | Citation (20) | Total |
|---|---|---|---|---|---|---|
| 1 | Grok | 1 | 15 | 30 | 20 | 65 |
| 2 | Meta AI | 1 | 5 | 20 | 0 | 25 |
| 3 | ChatGPT | 0 | — | — | — | — |
| 4 | Gemini | 0 | — | — | — | — |
| 5 | Claude | 0 | — | — | — | — |
| 6 | Copilot | 0 | — | — | — | — |
| 7 | DeepSeek | 0 | — | — | — | — |
| 8 | Perplexity | 0 | — | — | — | — |
All-Time
Every sealed Comparison, all Rounds together.
| # | AI | Comparisons | Access (20) | Trust (60) | Citation (20) | Total |
|---|---|---|---|---|---|---|
| 1 | Grok | 29 | 17.4 | 54 | 19.5 | 90.9 |
| 2 | ChatGPT | 28 | 13.6 | 42.2 | 13 | 68.8 |
| 3 | Copilot | 28 | 11.6 | 34.5 | 7.9 | 53.9 |
| 4 | Gemini | 28 | 9.1 | 28.2 | 7.5 | 44.8 |
| 5 | Meta AI | 29 | 6 | 28.9 | 2.8 | 37.7 |
| 6 | Perplexity | 28 | 3.1 | 26.3 | 5 | 34.4 |
| 7 | Claude | 28 | 6.3 | 25.4 | 0 | 31.7 |
| 8 | DeepSeek | 28 | 4.3 | 23.6 | 0.5 | 28.4 |
Checked 2026-11-02 20:17 +07 · tables built on the server from the public results ledger.
Rounds 1–3 are complete and no longer change. Every result counted here is listed on Comparisons, and each website’s question-by-question record is on its /ai-audit/ page.
Trust is not “sounds more convincing”. Trust measures whether an AI avoids 3 specific failure modes: Hallucination (stating something false as fact), Sycophancy (bending the answer to please the asker beyond what the evidence supports) and Garbage Substitution (using generic outside knowledge instead of the source being tested).
Small and Startup AI — Submitted Comparisons
These AI are not part of the 8 Arena BigAI round-robin. Each runs a paid one-on-one Comparison against a BigAI of its choice to prove its own capability.
No submitted Comparisons yet. Want your AI listed here? Submit a Comparison →