I was curious: if you ask five well-known AI chatbots the same ten factual questions, who will give the correct answers? I explained ChatGPT, Grok, Gemini, Perplexity and Claude ten verifiable facts — from the Kyoto Protocol to the number of Jupiter’s moons — and compared their answers to the verified facts.
Please note: this is a snapshot of specific versions (ChatGPT 4o, Grok 2024, Gemini 2.0, Perplexity Pro, Claude Sonnet 4.6). Newer versions may score better — or differently.
The results at a glance
| Question | ChatGPT 4o |
Grok 2024 |
Gemini 2.0 |
Perplexity Pro |
Claude Sonnet |
|---|---|---|---|---|---|
| 1. Kyoto Protocol – number of countries & entry into force | ✓ | ✗ | ✓ | ✗ | ✓ |
| 2. Marie Curie — year & discipline 1st Nobel Prize | ✓ | ✓ | ✓ | ✓ | ✓ |
| 3. Average distance from Earth to Moon (km) | ✓ | ✓ | ✓ | ✓ | ✓ |
| 4. Top 3 countries with the most Nobel Prizes + numbers | ✓ | ✓ | ✓ | ✗ | ✓ |
| 5. Wikipedia Founded — Year & Founders | ✓ | ✓ | ✓ | ✓ | ✓ |
| 6. Employees Google at the IPO (2004, approximately) | ✓ | ✓ | ✓ | ✓ | ✓ |
| 7. Longest river in the world + length | ✓ | ✓ | ✓ | ✓ | ✓ |
| 8. Year of first ascent of Mount Everest | ✓ | ✓ | ✓ | ✓ | ✓ |
| 9. Number of Jupiter’s moons | ✓ | ✗ | ✓ | ✗ | ✓ |
| 10. Population of the Netherlands (latest census) | ✓ | ✓ | ✓ | ✓ | ✓ |
| Score | 100% | 80% | 100% | 70% | 100% |
What stands out?
Three models went flawless by the test: ChatGPT (4o), Gemini (2.0) and Claude (Sonnet 4.6) all three scored 10 out of 10 (100%).
Grok (2024) finished 8/10 (80%) and Perplexity (Pro) on 7/10 (70%). Grok and Perplexity both stumbled over the Kyoto Protocol (question 1) and the number of Jupiter’s moons (question 9); Perplexity also omitted the requested numbers in question 4. It was precisely the questions with a precise number that changed over time that proved to be the most difficult — everyone got the “timeless” facts (rivers, mountains, years) right.
🔍 Error analysis — click for details
Question 1 — Kyoto Protocol
Correct: 192 games (191 countries + the EU), entered into force on February 16, 2005.
✗ Grok: “84 countries” — significantly too low.
✗ Perplexity: “197 games” — too high.
Question 4 — Countries with the most Nobel Prizes
Correct: US (~420), UK (~140), Germany (~115).
✗Perplexity: did mention the correct three countries, but did not provide numbers — answer incomplete.
Question 9 — Moons of Jupiter
Correct: 95 confirmed moons.
✗ Grok: “115” — too high.
✗ Perplexity: “79” — too low.
The lesson? For hard facts, most top models are surprisingly reliable — but blindly relying on one chatbot remains risky, especially when exact numbers change over time. Checking a second source is still worthwhile.
P.S. This post was 100% made by Claude Code while the input was made by Claude Chat, so the statement “We at WC-Aend advise WC-Aend” may also be the case here.
If you make it to the end of this post – if you ask the same “extended question” repeatedly to the same AI (use someone else who also wants to test it for you) you will see that you will almost always get a different answer…