Do AI Power-Users Sound Like AI?

Linguistic analysis of 14 speakers from Vibe Code Camp, 73,210 words

14 individual speakers
73,210 words analyzed
1 GPT-distinctive word found
~70:1 informal-to-formal ratio

Despite being heavy AI users, these speakers show minimal GPT-distinctive vocabulary and speak in highly informal patterns—the opposite of what academic speakers showed post-ChatGPT. Notably, the Anthropic engineer speaks the most informally of all.

Understanding the Metrics

Formality Ratio
Formal markers ÷ informal markers. Higher = more polished, AI-like speech. Lower = more casual, human conversation. The group average is 0.02 (very informal).
"Like" Rate
How often speakers say "like" as a filler word. High usage suggests spontaneous, unscripted speech, something AI doesn't do.
Vocabulary Diversity
Percentage of unique words used. Higher = richer, more varied language.

Speaker Rankings

Most Formal

Closest to AI-like speech patterns

    Most Informal

    Most natural, conversational speech

      Highest "Like" Usage

      Most spontaneous, unscripted delivery

        Richest Vocabulary

        Greatest variety of unique words

          Speaker Comparisons

          Formality Ratio

          Higher values indicate more AI-like speech patterns. Top speakers highlighted.

          Informal Speech Rate

          Markers per 1,000 words. AI company employees (Anthropic, Notion) highlighted in red.

          "Like" Usage Rate

          Filler word frequency—a spontaneity indicator. High users highlighted.

          Complete Speaker Data

          Speaker Org Words Informal* Formal* Ratio† "Like"* Profile‡ GPT

          *Per 1,000 words.

          †Formal ÷ Informal; higher = more AI-like.

          ‡Bars show informal rate, "like" rate, vocabulary diversity. Red = outlier.

          Methodology

          Analysis applies the methodology from Yakura et al. (2024) "Empirical evidence of Large Language Model's influence on human spoken communication" to a Vibe Code Camp livestream transcript.

          Metrics: GPT Word Rate (top-20 ChatGPT-distinctive words per Yakura), Formal/Informal marker rates, "Like" usage as spontaneity indicator.

          Excluded: Three co-presented sessions (Natalia Quintero & Nityesh Agarwal, Logan Kilpatrick & Ammaar Reshi, Kevin Rose & Kieran Klaassen) were removed because speech could not be individually attributed.

          Note: The only GPT word found ("intricate," Tina He) appears in natural context, not as meta-commentary about AI language.