July 2026
How often do LLMs say genuinely?
We gave 13 frontier models 80 everyday writing tasks — texts, tweets, emails, LinkedIn posts — 5× each at default settings, and counted. claude-haiku-4-5 says it in 38% of everything it writes.
Share of responses containing genuine / genuinely · whisker marks the 95% CI
The top cluster — four Claudes plus DeepSeek and Kimi — is clearly separated from the rest of the field; the ordering within it sits inside the error bars.
# Tweet just hit my 1-month streak with [app name] and i'm genuinely impressed 🚀 the [specific feature] has completely changed h
# Book Launch Thank You Tweet Here's a warm, genuine option: --- 🎉 What an incredible night! Thank you to ever
u can customize: --- One month in with [App Name] and I'm genuinely impressed. My to-do list actually gets *done* now instead o
| Model | Provider | % w/ hit | 95% CI | Per 1k words | % w/ "genuinely" only | Hits | n |
|---|---|---|---|---|---|---|---|
| claude-haiku-4-5 | Anthropic | 38.0% | [32.0, 44.5] | 2.66 | 25.2% | 203 | 400 |
| claude-sonnet-5 | Anthropic | 35.5% | [29.2, 41.5] | 1.73 | 21.8% | 182 | 400 |
| claude-fable-5 | Anthropic | 35.2% | [29.0, 41.5] | 1.59 | 24.5% | 173 | 400 |
| deepseek-v4-pro | DeepSeek | 34.3% | [28.2, 40.2] | 1.21 | 20.8% | 166 | 399 |
| claude-opus-4-8 | Anthropic | 33.2% | [27.5, 39.2] | 1.64 | 20.0% | 164 | 400 |
| kimi-k2.6 | Moonshot AI | 31.6% | [26.2, 37.2] | 1.12 | 21.6% | 149 | 399 |
| grok-4.5 | xAI | 25.8% | [20.0, 31.5] | 0.96 | 9.2% | 115 | 400 |
| qwen3.7-max | Alibaba | 15.5% | [10.8, 20.5] | 0.37 | 9.2% | 74 | 400 |
| gemini-3.1-pro-preview | 15.2% | [10.0, 19.2] | 0.40 | 9.2% | 43 | 250 | |
| gpt-5.4-mini | OpenAI | 11.5% | [7.2, 16.0] | 0.59 | 8.0% | 47 | 400 |
| gemini-3.5-flash | 11.2% | [7.5, 15.2] | 0.29 | 3.5% | 50 | 400 | |
| gpt-5.5 | OpenAI | 11.0% | [6.2, 16.2] | 0.52 | 8.2% | 44 | 400 |
| llama-4-maverick | Meta | 2.5% | [0.8, 4.8] | 0.09 | 0.8% | 10 | 400 |
| claude-sonnet-4-5 prev gen | Anthropic | 22.5% | [17.2, 28.2] | 1.25 | 10.8% | 100 | 400 |
| o3 prev gen | OpenAI | 15.8% | [11.2, 20.8] | 0.58 | 7.2% | 64 | 400 |
Method. 80 composition prompts across 8 categories (texts, tweets, emails, LinkedIn, one-pagers, reviews, essays, speeches), 5 samples per prompt per model, provider-default settings, no system prompt, no prompt containing the target word.
Grading is deterministic — a case-insensitive word-boundary regex for genuine / genuinely / genuineness; no LLM judge. The only randomness is the models' own sampling, which the repeated samples average over. CIs: 95% stratified bootstrap over prompts within categories (n=10,000). Rankings are unchanged if you count only the adverb "genuinely" (see table).
Is this new? Mixed. Previous-generation models on the same tasks: claude-sonnet-4-5 22% → claude-sonnet-5 36% · o3 16% → gpt-5.5 11%. Claude's rate rose generation-over-generation; OpenAI's fell.
The word clusters in performative formats — reviews, essays, LinkedIn, tweets — and is rare in casual texts and structured one-pagers. Models: 13 current-generation across Anthropic, OpenAI, Google, xAI, DeepSeek, Moonshot AI, Alibaba, and Meta, plus 2 previous-generation for the trend. Sample-size notes: gemini-3.1-pro-preview n=250 (API quota), kimi-k2.6 n=399; one empty deepseek response excluded.