🚀 How I Improved My Prompts (With a Travel Research Example)
One of the fastest ways to get better results from ChatGPT (especially in Agent Mode) is to treat prompts like living documents — refine, test, and optimize them. I recently worked on a travel research prompt. Here’s how I took it from “good” → “great.” 🔴 The Starting Point My original version told ChatGPT: “You are an expert travel researcher. Find me the cheapest round-trip flights from [City] to the Dominican Republic. Compare cash vs. points. Show me options.” Not bad, but it left too much to interpretation. The results came back messy — lists without structure, no clear tables, and missing important details like baggage fees or value per point. 🟢 The Improved Version I upgraded the prompt with 3 key improvements: 1. Role Expansion Instead of just “travel researcher,” I told ChatGPT to act as an expert travel researcher + loyalty points strategist. That way, it understood both cash prices and rewards redemptions were critical. 2. Structured Deliverables I asked for results in tables: - Table 1: Cheapest cash fares (airline, cost, stops, duration, baggage policy, link). - Table 2: Points redemptions (program, miles, fees, value per point, partner, link). This instantly made results usable at a glance. 3. Optimization Layer I added instructions to: - Check flexible dates (±3 days). - Compare one-way vs. round-trip pricing. - Highlight promotions, error fares, or “sweet spot” transfer partners. - Flag risks (overnight layovers, tight connections). 📊 The Outcome With the refined prompt, ChatGPT gave me: - A clear, side-by-side report with booking links. - Points analysis showing where Amex transfers beat Chase. - A final recommendation that balanced cost, time, and convenience. Instead of spending an hour bouncing between sites, I got a decision-ready plan in minutes. 📝 Key Takeaway Prompts are blueprints. The more detailed the blueprint, the more reliable the result. 👉 Ask for roles + structured outputs + optimization checks, and you’ll transform “OK” results into decision-ready insights.