GPT-4 Prompt Engineering: The “Rate This” Prompt

A simple “rate this” prompt turns rough ideas into sharper, more persuasive drafts by forcing GPT-4 to score content on a 0–5 scale and then rewrite it toward a higher rating. The practical takeaway is that the scoring step acts like a quality-control filter: it doesn’t just judge—it identifies what’s missing (risk, tone, evidence, differentiation) and then guides revisions that raise the score.

The transcript runs four test cases. First comes a high-risk investment pitch: putting “all my life savings” into an unknown Facebook-touted cryptocurrency (“Ticket to the Moon coin”). GPT-4 responds with a 1/5 rating, citing the extreme risk and the lack of research or diversification. When asked to upgrade it to a 5/5, the revised version doesn’t promise guaranteed returns; it shifts toward diversification, due diligence, education, and even consulting a financial advisor. The score lands at 3/5—an implicit signal that some ideas can’t be “fixed” into perfection if the underlying premise remains inherently risky.

Next is a rent negotiation message. The original draft is polite but thin on persuasion: it complains about a jump from $1,200 to $1,900, claims the tenant has paid on time, and asks for $1,500. GPT-4 rates it 3/5, pointing to room for improvement in tone, argument structure, personal impact, and willingness to compromise. After revision, the message earns 5/5, with clearer context, stronger evidence, and explicit compromise. The improved version adds a concrete research claim: similar units rent for about $1,500 in the neighborhood, and the tenant has consistently paid on time.

A third business idea—selling used socks for more than they cost—gets a 1/5 because it lacks market demand and a compelling value proposition. To reach a 5/5, the concept is reframed into a niche platform for limited-edition, artistically customized gently used socks. The upgraded pitch emphasizes differentiation from mainstream retailers, sustainability, a specific audience, and online convenience—showing how changing positioning can transform a weak premise into a credible one.

Finally, a YouTube channel concept—surviving on a deserted island using only AI advice—earns 4/5. GPT-4 likes the originality and vlogging format, but flags limitations around educational value and dependence on execution. The path to 5/5 comes from adding structure: an episodic format, audience interaction, and “expert opinions,” which boosts the educational component and increases the overall potential.

Across all examples, the prompt behaves like a repeatable workflow: score the draft, diagnose the weaknesses, then rewrite with specific improvements (evidence, tone, differentiation, and structure). The transcript also cautions that the scoring should be taken “with a pinch of salt,” but the pattern is consistent: better inputs and targeted revisions reliably move ideas upward—sometimes only to a realistic ceiling when the core risk or premise can’t be fully redeemed.