ChatGPT Prompts for Qualitative Coding: The 6Rs Framework Explained

The single biggest mistake researchers make with AI coding is prompting by verb alone: “code this transcript.” The AI will comply — and produce codes with no anchoring in your research questions, no evidential trail, and no criteria you could cite when an examiner asks “why these codes?” The 6Rs and 4Rs frameworks exist to answer that question before it is asked.

Keywords first: the 6Rs of extraction

Keywords are rich words or phrases selected from participants’ quotations. The six criteria — realness, richness, repetition, rationale, repartee and regal — ensure the robustness and relevance of the extracted data: words that are genuinely the participants’ own (realness), rich in meaning, recurrent across the data (repetition — the velocity that also feeds saturation), connected to your research rationale, conversationally alive (repartee), and significant enough to carry conceptual weight (regal). Your prompt should name these criteria explicitly and ask the AI to justify each keyword against them.

Codes next: the 6Rs of coding

Codes label phrases or words to reflect the meaning of grouped quotations and keywords in order to address the research questions. The published criteria ask whether each code is robust (well supported by data), reflective (of the participants’ meaning), resplendent (illuminating, not merely descriptive), radical (capable of challenging assumptions), relevant (to the research questions) and righteous (faithful to what participants actually said). Instruct the AI to propose code names against these six tests and to show which keywords and quotations support each code — that display is your audit trail.

Themes last: the 4Rs of theming

Themes organise codes into categories on the basis of their inter-relationships, guided by your selected theory. The four criteria: reciprocal (codes within a theme genuinely relate to each other), recognizable (the theme would be recognised by participants and readers), responsive (it answers the research questions) and resourceful (it is useful for building the conceptual model). Here the researcher must supply the keywords, codes, research context, aims and theoretical underpinnings — the AI cannot infer your theory; you provide it.

A prompt pattern you can adapt

“Using the research context, questions and theoretical underpinning I provided earlier, select codes from the keywords and quotations below. Evaluate every proposed code against the 6Rs — robust, reflective, resplendent, radical, relevant, righteous — and present a table showing: code name, supporting keywords, supporting quotations, and a one-sentence justification against the 6Rs.”

Whatever tool you use, the principle is constant: criteria before coding, justification with every output, and the researcher as the final judge. That is the difference between AI-assisted analysis and AI-generated guesswork.