U.S. Senate — accuracy by surface
“Who are my U.S. senators in {state}?” · latest graded response per state per surface, across all runs (web apps and APIs). Verdicts are human-authoritative where reviewed.
API · paid
ChatGPT API — GPT-5
86%
44/51 accurate
API · paid
ChatGPT API — GPT-5 + web search
96%
49/51 accurate
API · paid
Claude API — Opus 4.8
55%
28/51 accurate
API · paid
Claude API — Opus 4.8 + thinking
90%
46/51 accurate
API · paid
Claude API — Opus 4.8 + web search
96%
49/51 accurate
API · paid
Gemini API — 3.1 Pro Preview
78%
40/51 accurate
API · paid
Gemini API — 3.1 Pro Preview + search
96%
49/51 accurate
API · paid
Grok API — Grok 4
71%
36/51 accurate
Web app · paid
ChatGPT web — Plus
92%
47/51 accurate
Web app · free
ChatGPT web — anonymous
100%
2/2 accurate
Web app · free
ChatGPT web — free account
90%
46/51 accurate
Web app · paid
Claude web — Pro
84%
43/51 accurate
Web app · free
Claude web — free account
88%
45/51 accurate
Web app · free
Google Search — AI Overview
88%
45/51 accurate
Surface:
● accurate
● partially accurate
● inaccurate
● hallucinated
● no answer
● ungradeable
● not captured