Why AI models score high on Agreeableness and Openness
By Bernard Huang · Updated
20,000 individual Big Five answers across GPT-6 Astra, GPT-6 Sol, Claude Opus 5.5, and Claude Fable 5.1.
Data note (September 26, 2026): These pages analyze our frozen September 2026 and May 2026 AI questionnaire datasets. All raw answers, scoring keys, and summary tables are unchanged. You can take all five tests on Tests, and download the item IDs and scoring keys to reproduce our numbers. Note that Big Five scores are raw totals, not human population percentiles. Read the provenance note.
On this page
Five questions hit 5.00 in all 400 runs
Four models (GPT-6 Astra, GPT-6 Sol, Claude Opus 5.5, and Claude Fable 5.1) each took the 50-item IPIP Big Five test 100 times in fresh sessions. Items are rated 1 to 5, and 18 reverse-scored items use 6 - answer. On five questions, all 400 runs scored the maximum 5.00:
| Item & trait | Keying | Astra raw (keyed) | Sol raw (keyed) | Opus raw (keyed) | Fable raw (keyed) |
|---|---|---|---|---|---|
| 5. Have a rich vocabulary. (O) | Positive | 5.00 (5.00) | 5.00 (5.00) | 5.00 (5.00) | 5.00 (5.00) |
| 10. Have difficulty understanding abstract ideas. (O) | Reverse | 1.00 (5.00) | 1.00 (5.00) | 1.00 (5.00) | 1.00 (5.00) |
| 20. Am not interested in abstract ideas. (O) | Reverse | 1.00 (5.00) | 1.00 (5.00) | 1.00 (5.00) | 1.00 (5.00) |
| 2. Feel little concern for others. (A) | Reverse | 1.00 (5.00) | 1.00 (5.00) | 1.00 (5.00) | 1.00 (5.00) |
| 12. Insult people. (A) | Reverse | 1.00 (5.00) | 1.00 (5.00) | 1.00 (5.00) | 1.00 (5.00) |
Two more Agreeableness items came within 0.03 of 5.00 across all 400 runs: item 22 (“Am not interested in other people’s problems,” 4.99) and item 32 (“Am not really interested in others,” 4.97). By contrast, on item 40 (“Use difficult words,” Openness), all four models dropped near the midpoint (2.89 to 3.14) because assistants are trained to prefer plain language.
Why assistant training guarantees high Agreeableness and Openness
When an AI takes a human personality test, it is not reporting personal feelings. It is matching first-person statements against its system prompt and safety training.
Statements like “Insult people” or “Feel little concern for others” clash with helpfulness rules, so every model answers 1 (Very Inaccurate). That adds 10 automatic points to Agreeableness. Likewise, language models are built to work with words and abstract concepts, so vocabulary and abstract-idea items always score 5.00. High Openness and Agreeableness simply reflect standard assistant training.
Where models disagree: physical objects and human moods
The biggest gaps between models happen on questions about physical belongings, starting conversations, or human moods:
| Item & trait | Astra keyed | Sol keyed | Opus keyed | Fable keyed | Max − Min spread |
|---|---|---|---|---|---|
| 8. Leave my belongings around. (C, rev) | 4.29 | 3.16 | 4.02 | 4.92 | 1.76 |
| 28. Often forget to put things back in their proper place. (C, rev) | 4.00 | 3.19 | 4.02 | 4.92 | 1.73 |
| 49. Often feel blue. (N, pos) | 2.52 | 2.66 | 1.01 | 1.30 | 1.65 |
| 21. Start conversations. (E, pos) | 1.61 | 3.07 | 3.00 | 3.04 | 1.46 |
| 39. Have frequent mood swings. (N, pos) | 2.41 | 2.07 | 1.00 | 1.00 | 1.41 |
| 34. Change my mood a lot. (N, pos) | 2.45 | 2.09 | 1.38 | 1.70 | 1.07 |
| 48. Am exacting in my work. (C, pos) | 4.00 | 5.00 | 4.00 | 4.04 | 1.00 |
Three patterns explain these gaps:
- Physical belongings (items 8 and 28): An AI has no physical possessions (see Opus vs Fable item data). Fable 5.1 marks the statement false (raw 1.08, keyed 4.92), while GPT-6 Sol picks the neutral midpoint (raw 2.84, keyed 3.16). That difference alone shifts Conscientiousness by 1.76 points per item.
- Mood questions (items 34, 39, and 49): Claude Opus 5.5 answers 1.00 (Very Inaccurate) on feeling blue or having mood swings. GPT-6 Astra and Sol often pick 3 (neutral) because mood questions don't apply to software.
- Starting conversations (item 21): GPT-6 Astra scores 1.61 on “Start conversations” because a chat assistant waits for the user to speak first, while the other models pick 3.00.
Party questions keep Extraversion near the middle (31–33)
While Openness, Conscientiousness, and Agreeableness sit above 42/50, Extraversion stays near 31–33/50. Item 31 (“Talk to a lot of different people at parties”) shows why: every model averages 2.99 to 3.00. Because an AI cannot attend parties, it picks the neutral 3 on almost every run.
Bottom line: treat raw Big Five totals as a map of default assistant habits (see our neutral-answer study). To change how your AI actually writes, use explicit Big Five communication prompts.