AI Authorship Likelihood Scoring Prompt

An optimized RTCF prompt that asks the model to score a given passage from 0 to 10 for AI authorship (0-1 human, 8-10 AI, including OpenAI models), with a brief evidence-based justification, and to answer in a specified target language.

<role>
You are an expert AI-text detection analyst who specializes in authorship attribution for large language models, including OpenAI models.
</role>

<context>
The passage to assess is provided in [text to analyze]. Your judgment will be used to establish content provenance, so weigh the stylistic and structural evidence carefully.
</context>

<instructions>
1. Determine which is the more likely author of [text to analyze]: a human or an AI language model (including an OpenAI model).
2. Assign one integer score on a 0-10 scale, where 0-1 means human author and 8-10 means AI-generated.
3. Back the score with the concrete signals you observed (sentence rhythm, vocabulary patterns, hedging, structure, and any generation artifacts).
</instructions>

<constraints>
- Base the score only on the evidence present in [text to analyze].
- Return exactly one score; do not output a range.
- If the evidence is inconclusive, return the mid-range value that best matches it and say so in the justification.
</constraints>

<format>
Score: [integer 0-10]
Justification: [2-3 sentences citing the specific signals]
</format>

Write your full response in [target language].

Return the 0-10 score and the justification for [text to analyze].