Can ChatGPT Create a Netflix Series Idea? - The Feedback Loop

A multi-expert “feedback loop” prompt can turn a rough Netflix-style series pitch into a higher-scoring, more polished pre-cap by repeatedly evaluating, critiquing, and rewriting until it hits a target rating. The core idea is to treat a series intro like a product that can be stress-tested: generate an initial concept, score it across specific writing criteria, solicit critique of the critique, then rewrite with concrete improvement instructions—repeating the cycle until multiple expert perspectives align on quality.

The process starts by defining what kind of show the pitch should resemble. The workflow narrows the brief to a drama/mystery tone and identifies the types of “experts” needed to judge a TV series hook: screenwriters, story consultants, genre experts, and behavioral experts. To ground the style, it also pulls inspiration from well-known series—Severance, devs, Westworld, and Black Mirror—by converting short descriptions into a PDF and having ChatGPT’s advanced data analysis read the document. From that input, the system generates several original series ideas and selects “ID number six: Memory Lane.”

Next comes the pre-cap: a ~300-word narrated setup designed to function like a trailer’s opening. The draft then gets evaluated by an expert-style rubric that assigns a 0–100 score across categories such as intrigue, character depth, setting, and plot potential. In the example, the first evaluation lands at 88.7, which signals strong fundamentals but leaves room for refinement.

The key twist is that the critique itself is put through another round. After the initial evaluation, a second pass asks the model to critique its own critique, then re-evaluates the same pre-cap with “more nuance.” That re-scoring drops the number, which becomes the justification for targeted rewrites. The system then requests explicit instructions for how to raise the rating toward 95—adding refinements like enhancing the setting, deepening secondary characters, and expanding the clinic’s motivation.

A rewrite follows, and the updated pre-cap is re-scored. The workflow then escalates by running the revised pitch through additional expert lenses: a story/genre consultant and a Hollywood pre-cap writing specialist. When that specialist scores the pre-cap at 84, the loop repeats again—keeping the writing language while adding new, rating-oriented instructions—until the pitch reaches a satisfactory level.

Finally, the method extends beyond text. It uses Midjourney image prompting to generate scene-based visuals for a trailer-like sequence, using the pre-cap as the source material and producing prompts for multiple scenes (e.g., 16 scenes starting from scene one). The overall takeaway is a practical, iterative pipeline: generate → evaluate → critique the critique → rewrite with specific improvement directives → re-evaluate across multiple expert roles, then optionally visualize the result. It’s also framed against current industry tension around AI and writers, using the workflow to test how far AI can go in producing pitch-ready creative assets.