Turn a Messy Brief into Prioritized Clarification Questions with ChatGPT

A messy brief can contain a clear objective alongside missing constraints, contradictory instructions, and terms that mean different things to different people. Asking ChatGPT to turn it directly into a finished plan may hide those problems behind plausible assumptions. A clarification queue makes the uncertainty visible before work depends on it.

This guide produces a brief gap analysis and a concise set of questions. It is an original workflow with fictional examples, checked against current OpenAI guidance on October 6, 2026. The objective is to identify the information that changes the work, not to rewrite approved requirements or invent answers on the brief owner's behalf.

Preserve the original brief

Save the brief's current version and identify its author, date, intended audience, and requested deliverable. Label relevant passages with stable IDs. If the brief came from several messages, preserve their sequence and source IDs so later comments do not accidentally replace earlier approved instructions.

Extract the explicitly stated facts and requirements into a reference list. Keep qualifications such as “draft,” “subject to review,” and “for internal use.” These words can materially change the scope of work even when the rest of the brief sounds decisive.

Separate the brief from supporting sources. An attached example may illustrate style without defining the required structure. An older project plan may provide background without controlling the new deadline. Record the relationship explicitly instead of allowing the model to treat all supplied material as equally authoritative.

Classify gaps by the decision they affect

A useful gap analysis distinguishes missing information, ambiguous wording, contradictions, and unsupported assumptions. Missing information is absent. Ambiguous wording has several plausible interpretations. A contradiction contains incompatible instructions. An assumption is a proposition the proposed work relies on but the brief does not establish.

Examples include an unspecified audience, a request for both a short summary and a detailed report without a length rule, or a dependency on a source file that has not been supplied. Record the exact passage and the practical effect of each gap.

OpenAI's Enterprise prompting guide notes that better prompting does not replace missing source material or clear requirements. The queue below applies that principle through an original review schema.

Build a gap register

Give each gap an ID and connect it to the relevant output or decision. Avoid a vague list of concerns that the brief owner must interpret again. The register should explain why an answer matters and what work can proceed without it.

Use blocking only when the answer materially prevents accurate work. An uncertain audience may block the structure of a report. A preferred heading style may be an optional refinement if a reasonable existing template already applies. Do not label every detail urgent.

Ask ChatGPT to analyze before proposing answers

Supply the brief, source relationships, required deliverable, and register fields. Ask the model to distinguish evidence from inference and to return questions rather than a completed plan.

Analyze the supplied brief for gaps affecting the requested deliverable.
Preserve explicit approved requirements and source relationships.
Identify missing information, ambiguity, contradictions, and unsupported assumptions.
For each gap provide its passage reference, consequence, and clarification question.
Prioritize questions as blocking, important, or optional with a brief reason.
Do not answer on the owner's behalf or rewrite the brief as an approved plan.
Identify useful work that can proceed independently of unresolved questions.

Review the output for invented gaps. A question already answered in the brief creates unnecessary work for the owner. Confirm that the cited passage supports the uncertainty and that the proposed answer would actually change the deliverable.

Phrase questions that can be answered

Ask one decision per question whenever practical. “Who is this for, how long should it be, what tone do you want, and when is it due?” bundles several issues and makes replies difficult to track. Separate them when their owners or consequences differ.

Offer concise choices when the brief supports a small set of plausible interpretations. For example, ask whether a fictional launch report is intended for the project team or external customers, and explain how that choice changes the level of detail. Include a way to supply a different answer rather than forcing the owner into a false binary.

Avoid loaded wording. “Why did you omit the audience?” focuses on blame. “Which audience should the report serve?” resolves the information gap. The queue should make it easy for the owner to clarify intent without defending the roughness of the original brief.

Work through a fictional brief

Imagine a brief requesting “a concise launch overview with complete implementation detail for everyone.” It names a publication date but does not identify whether the overview is internal or public. The audience question is blocking because it affects content, terminology, and what information can be included.

The length conflict is a separate gap: concise overview and complete implementation detail may require different outputs. A useful question asks whether the owner wants a short main document with a technical appendix or a single detailed document. The model should not silently choose one.

Meanwhile, collecting approved source files can proceed if that work does not depend on the audience answer. The register can distinguish that independent preparation from drafting the final document. This allows progress without embedding unapproved assumptions in the product.

Prioritize and sequence the queue

Put questions that determine scope, audience, source authority, or success criteria first. Their answers may resolve several lower-priority questions. A decision about a public audience, for example, can eliminate uncertainty about internal implementation details.

Group related questions for readability while retaining their IDs. Keep the owner-facing queue shorter than the complete internal register when possible. Preserve minor concerns internally rather than overwhelming the owner with every possible preference.

Identify dependencies between questions. If the output format depends on audience, ask the audience question first. A flat list can cause the owner to answer a later question using assumptions that change after an earlier decision is made.

Incorporate answers and close gaps explicitly

Record each answer with its author, date, and gap ID. Update the scope reference and mark the gap resolved only when the answer addresses the uncertainty. A response such as “use your judgment” may authorize a routine choice, but it does not supply a missing factual source.

Check whether answers introduce new contradictions. If the owner changes the audience but retains an incompatible content requirement, raise the remaining issue clearly. Preserve the revised brief version instead of quietly blending the original and new instructions.

Use the feedback ledger workflow for later review comments on the resulting draft. The gap register establishes what was clarified before drafting; the editing ledger records how a specific draft changed afterward. Keep those purposes distinct.

Frequently asked questions