AI Article Review

Problem

Many AI technical or applied articles look complete before submission, yet they may have vague titles, weak ledes, missing core questions, shallow mechanism explanations, evidence disconnected from claims, or financial-risk sections that avoid underlying mechanisms. This skill reviews drafts intended for mainstream financial, business, or industry media, turning subjective concerns into reviewable issues and actionable fixes.

How It Works

The review follows a standards-first workflow:
- Identify the title, target media, research type, estimated length, and input source; for Word files, convert the document to readable text and flag possible image, OCR, or revision loss.
- Infer the core questions a qualified article should answer based on technical, applied, or hybrid research, including capability boundaries, implementation conditions, ROI, risks, and the tension between regulatory explainability and non-linear AI behavior.
- Check problem awareness, mechanism explanation, argument closure, and restrained expression against internal style notes, mainstream media preferences, and general review frameworks.
- Score Clarity 30%, Depth 30%, Evidence 15%, and Writing 25%; the title is evaluated separately and excluded from the total score.
- Classify issues as P0, P1, or P2, with location, reason, and concrete revision for each item.

Boundaries

It is suited for pre-publication review, structure optimization, depth evaluation, and proofreading, especially for AI industry analysis, applied deployments, technical roadmaps, and financial, credit, or risk-control topics. The skill does not invent facts, cases, data, or citations; when evidence is missing, it recommends what kind of source should be added. If the target outlet differs from mainstream business media or the user asks for a full rewrite, those limits and goals should be stated.