Rhetorically Governed Co-Authoring (RGC) is a developing framework for structuring AI-assisted writing without treating fluent output as a substitute for rhetorical judgment. It assigns AI a defined and visible role while keeping the writer responsible for purpose, boundaries, evaluation, revision, attribution, and transfer.
From invisible delegation to governed collaboration
The framework begins with a distinction between two arrangements of rhetorical labor. In a ghostwriter model, AI performs substantial writing work invisibly and the writer may accept the result without adequately planning, monitoring, comparing, revising, or evaluating it. Fluency can then conceal weakened authorship because the finished prose does not show where judgment occurred—or whether it occurred at all.
In a co-author model, AI performs a bounded role inside a process the writer can explain. “Co-author” describes distributed rhetorical labor; it does not claim AI personhood, legal authorship, or equal agency. The human writer remains responsible for the rhetorical situation and for every decision that follows.
Anchor: establish the human rhetorical situation
Anchor comes before generation. The writer names the situation, purpose, audience, criteria, and stakes of the work. The writer also defines what role—if any—AI may play, identifies boundaries, and decides what documentation the context requires.
This phase matters because a system cannot preserve a purpose the writer has not articulated. Anchoring creates a basis for later comparison: the writer can evaluate an option against declared criteria instead of accepting it because it sounds polished.
- Identify purpose, audience, genre, and constraints.
- Define standards for a successful response.
- Assign AI a limited role rather than open-ended authority.
- Name prohibited uses and documentation expectations.
Field: generate bounded options
Field treats AI output as a field of possibilities rather than an answer. The system may help produce alternatives, questions, examples, organizational options, or revision possibilities within the role established during Anchor. Multiple options make comparison possible and reduce the pressure to treat the first fluent response as authoritative.
The goal is not maximum generation. It is purposeful optioning: enough variation to support judgment while preserving the writer’s responsibility to notice assumptions, omissions, standardization, and rhetorical consequences.
Filter: monitor, decide, revise, and account
Filter is where the writer compares generated material with the original situation and criteria. The writer selects, rejects, combines, rewrites, verifies, and attributes. Reflection makes the decision process visible and supports transfer: the writer identifies what changed, why it changed, and what judgment can carry into later work.
The full rhetorical loop is: Situation → Criteria → Role and Boundaries → Options → Comparison → Selection and Revision → Attribution → Reflection → Transfer.
A framework for learning, not a prompt formula
RGC is designed to make intellectual checkpoints visible. In a classroom, students can show their rhetorical plan, compare multiple possibilities, explain rejections, document revisions, and reflect on what they learned. In professional settings, the same structure can make responsibility and review clearer when AI assists with communication.
The framework remains in development. Current work focuses on clarifying its conceptual boundaries, translating it into classroom protocols and workshop materials, and testing how it can preserve accessibility without turning appropriate assistance into invisible delegation.