5 Easy Ways to Increase ChatGPT Prompt Results Tenfold (Scientists Say)
Even if you’ve been using ChatGPT for a while, you’re probably making mistakes. There’s still a lot of work to be done. You fall into common prompting mistakes, create content that sounds way too much like ChatGPT, or ignore the essential components of the perfect prompt. Even the smartest entrepreneurs don’t always get it right. Get back to basics with these five simple prompting tips to get maximum results in the shortest amount of time.
The research paper Plan-and-Solve Prompting: Improving Zero-Shot Chain-of-Thought Reasoning by Large Language Models was prepared by several academics from Singapore Management University and others. In addition to specific prompt additions used to create superprompts, the paper provides five simple tips that no one should overlook.
Applying these tips helps large language models (LLMs) like ChatGPT, Claude, and Gemini handle multi-step reasoning tasks more effectively, allowing entrepreneurs to use them more effectively.
ChatGPT Tips for Better Prompt Results: 5 Science-Based Methods
Customize prompts for different tasks
Just as you would instruct different members of your team in different ways depending on the task at hand, you should tailor your prompts to different tasks by adjusting the level of detail and specific instructions. You wouldn’t record someone’s Loom conversation to discuss a critical and urgent business decision. You also wouldn’t call a meeting about a simple accounting question. Apply the same mindset to ChatGPT’s prompt.
You don’t need a superprompt for every request. Tailor the prompt to the business task at hand. Don’t make it unnecessarily complicated. For example, if you want ChatGPT to evaluate your landing page, include best practices, explain your audience, ask it to take on a role, and provide the feedback you want. On the other hand, for a simple task like “Find a synonym,” a one-line instruction might be enough.
Integrate practical examples
ChatGPT doesn’t live in the real world. It processes over 300 billion words in its training data to make predictions, and sometimes gets it wrong. Help your enthusiastic AI co-star. Use real-world business scenarios in your prompts to make the answers more relevant and practical. Help it understand what you’re trying to achieve.
Use real-world examples to illustrate the problem so that ChatGPT gets the mood it wants no matter what is asked. Include square brackets with the context it needs to know, for example, “My company offers [explain your offer] for [explain your audience] and I am facing challenges, including [describe them].” Get closer to the perfect answer the first time.
Divide tasks
You wouldn’t give a human team member a bunch of confusing instructions for the week, with no context or sense of priorities. You also wouldn’t demand that they do everything at once. Apply the same mindset to ChatGPT. Instead of sending a huge prompt with lots of sections and confusing instructions, take it one step at a time. Ask for the first result, then take the result and ask for more.
By breaking complex tasks down into smaller phases, each with its own instructions and goals, ChatGPT can better handle them. For example, if you request a blog post, ask for heading options first before choosing the best one and asking for a summary. Provide your summary changes before asking for section headings, and refine those before requesting each section to be written. This is better than requesting a blog post all at once, and means you have more control over the outcome.
Combine human and ChatGPT efforts
Contrary to what the AI hype would have you believe, ChatGPT cannot solve all your problems. At least not yet. It hallucinates, it exaggerates. It comes out with random stuff that is not customer-focused and would do more harm than good to your business. Don’t forget the human touch. Use ChatGPT for the initial steps and have humans review and validate the final results. This combines the best of both worlds.
The document recommends creating an initial solution and having it reviewed by a human expert. Don’t move from prompt to publishing without taking this step. Publishing ChatGPT-generated content without proper review will damage your brand. Watch for the telltale signs, carefully check every word, and only use sentences you would actually say. Check facts, check meaning, and maintain your integrity.
Continue to improve
The document acknowledges that ChatGPT is constantly evolving. Even between model numbers, results can be completely different a few weeks apart. Keep learning as the platform evolves. Read the studies, be highly critical of the results, and continually refine your prompts based on feedback and performance.
Get the best results with ChatGPT using the same approach you use in business. What worked like a charm in the past may not work today. Experiment, learn, iterate, improve. Even try the same prompts with alternative LLMs. Don’t settle for your tried-and-true prompts; review them regularly.
Back to basics with ChatGPT: quick improvements for better results
By following these simple tips, entrepreneurs can get better results from ChatGPT, especially on complex tasks. Customize prompts based on desired outcome, incorporate real-world examples, break tasks down, combine machines and humans, and constantly improve. Simple steps that will help ChatGPT understand and solve problems more accurately, for smarter decisions and better business results. Stop wasting time on suboptimal prompts. Develop this skill and constantly improve. You’ll need it.