The Ping-Pong Prompting Method – 7 Simple Steps to AI Success (in 2025)

Are you tired of the frustrating loop of trial-and-error prompting? Does each new attempt feel like the AI just doesn't get it?

I've been there countless times, until my business partner, Mercer, showed me the power of the ping-pong prompting technique. 

Ready to ditch the prompting frustration? Want better outputs and a more enjoyable way to work with AI?

The key is this simple yet effective technique…

See how ping-pong prompting instantly improves your AI results

See how Mercer explains this powerful technique in his Gemini workshop. He'll show you how this back-and-forth approach can dramatically improve your AI outputs right away.

Key takeaways

• Experience instant improvement: ping-pong prompting is easy to learn and works on your very first attempt.
• Be specific in your AI prompts and provide context: To get faster and more accurate initial responses, you need to help the AI understand your needs.
• Break down large AI tasks into smaller, manageable chunks: Instead of one big prompt, give the AI smaller, focused requests. Easier for you to manage, and easier for the AI to understand and analyze!
• Work with AI in real time: AI is your teammate in this ping-pong match! If you truly work together, the output will be the high-quality you’re aiming for.

Step 1: Craft your initial human prompt

Think of your initial prompt as the starting point of your ping pong game with AI. 

While the technique is all about refining as you go, a good first prompt can still set you up for better AI results right away.

”Ping-pong prompting is a secret to taking a bad prompt and turning it into a good one.” Jeff Sauer

Here's Mercer’s advice to keep in mind for your first serve:

• Be clear and specific: The AI will help you refine your prompt, but it’s best to avoid vague instructions. For example, instead of “Write a blog post,” try “Write a blog post about the benefits of using AI in marketing for small businesses.“
• Define your desired outcome: What exactly do you want the AI to produce? A list of ideas? A draft of an article? Code? Be explicit about your goal.
• Consider the role: Include a role in your initial prompt if it helps to frame your request. For example, “Act as a marketing expert and suggest…“.
• Don't overthink it: Remember, this is just the starting point. The AI will help you improve it. 

As Mercer mentions in his workshop on how to use Gemini, the initial prompt doesn't need to be a “power prompt.” 

This is also clear from the basic prompting scheme he made during his presentation:

You can start with something basic. 

Then, apply the following step…

Step 2: Engage AI for analysis

This is where the “ping-pong” action truly begins. Once you have your initial prompt, you'll feed it back to the AI along with a specific instruction to analyze it. 

This instruction is crucial, as Mercer found out.

Here's the prompt you'll essentially be giving to the AI:

(replace “[Your Initial Prompt]” with the prompt you crafted in the previous step):

Analyze the following prompt:

[Your Initial Prompt]

You will rate your understanding of the desired result on a scale of 1 to 10, where 10 means you fully understand what I am looking for.

Then, explain your reasoning for the rating.

After that, rate your understanding of the hidden intention behind the desired result on a scale of 1 to 10, where 10 means you fully understand what I am trying to achieve with this result.

Finally, explain your reasoning for this second rating.

Key elements of this instruction:

• Clear task: You are explicitly asking the AI to “analyze” your prompt.
• Two ratings: You are asking for two distinct ratings: one for the understanding of the desired result and another for the understanding of the hidden intention. This is a key differentiator of this powerful prompting method.
• Defined scale: You clearly define the 1-to-10 scale, with 10 representing complete understanding.
• Reasoning: You are not just asking for a rating, but you want to see the AI's reasoning behind each score. 

By using this specific instruction, you're prompting the AI to think about what you're asking for on multiple levels. And to share its thought process with you.

And that will help you with the next step…

Step 3: Understand the AI's reasoning

After the AI analyzes your initial prompt, it will provide ratings and, crucially, the reasoning behind them. This reasoning offers valuable insight into how the AI interpreted your request, highlighting areas where your prompt was clear or ambiguous. Pay close attention to this feedback, as Mercer advises, for both the desired result and the hidden intention.

• Desired result reasoning: A rating below 10 here shows the AI likely found something unclear or ambiguous in your prompt regarding your desired output. Look for mentions of missing details, potential misunderstandings, or areas needing more specificity.
• Hidden intention reasoning: A low rating for hidden intention means the AI doesn't fully grasp why you want this result or the underlying problem you're trying to solve. This is vital information! Understanding the AI's perspective here will allow it to provide more relevant and helpful outputs, potentially even suggesting better ways to achieve your goal.

Carefully read and understand the AI's reasoning. 

This feedback is central to the “ping-pong” process and will guide your prompt refinement in the following steps.

Step 4: Let AI identify knowledge gaps

If the AI's rating for its understanding of either the desired result or the hidden intention was anything less than a 10 (and as Mercer mentioned, it rarely is a 10 on the first try), it signifies that there are gaps in its understanding. 

The reasoning it provided in the previous step will often hint at these gaps.

While the AI might not explicitly ask a series of questions at this stage, its reasoning will often point towards what it needs to know to achieve a higher level of understanding. For example, it might say:

• “I understand you want a blog post about AI in marketing, but I'm unsure about the target audience.” (Gap in desired result understanding)
• “I understand you want a list of benefits, but I'm not clear on the specific goals of these small businesses.” (Gap in hidden intention understanding)

The following step involves prompting the AI to pose questions that highlight areas of missing knowledge. 

This will make it even clearer what information you need to provide to refine your prompt effectively.

Step 5: Empower AI to generate answers

Instead of you figuring out what the AI needs, you can ask it to come up with the answers. This powerful part of ping-pong prompting saves you time and effort.

Here's how you can prompt the AI to do this:

Based on your analysis and reasoning in the previous step, list the top questions you have about the prompt to reach a 10/10 understanding of both the desired result and the hidden intention. Then, provide the most useful answers to those questions yourself.

Key elements of this instruction:

• Reference previous analysis: You explicitly tell the AI to build upon its previous analysis and reasoning.
• Request questions: You ask the AI to list the questions it has. This makes the knowledge gaps explicit.
• AI-generated answers: You instruct the AI to provide the most useful answers to its own questions. This is where the AI leverages its knowledge and understanding to fill in the gaps.

By giving the AI this instruction, you're essentially asking it to think critically about your prompt. And it will proactively identify and address areas where it needs more information. 

This leads to a more refined understanding and sets the stage for the final iteration of your prompt.

Step 6: Iterate until full understanding

The process you've just initiated – having the AI analyze your prompt, provide reasoning, identify knowledge gaps, and generate answers – might not lead to a perfect 10/10 understanding on the first try. 

This is where the “ping-pong” continues, as Mercer shows in his workshop

After the AI provides its initial set of questions and answers, carefully review them. 

• Do the answers it provided fully address the gaps in understanding? 
• Do you agree with its interpretation of your desired result and hidden intention?

If the AI's understanding is still below a 10 for either category, you'll essentially repeat the previous steps:

• Feed the AI's questions and answers back to it.
• Ask it to re-evaluate its understanding of the desired result and hidden intention based on this new information.
• Instruct it to provide updated ratings and reasoning.
• If the ratings are still below 10, ask it to identify any remaining questions and provide answers.

”The whole point of AI is it collapses time so we can get a result faster.” Mercer

You'll continue this iterative process, going back and forth with the AI, until it indicates a strong (10/10) understanding of both what you want it to produce and why you want it.

Important Note: Follow Mercer's advice: “Trust, but verify.” 

While the AI is doing the heavy lifting in this iteration process, it's crucial for you to review its reasoning and suggested answers critically. 

Ensure they align with your actual goals and intentions. 

After all, you are the manager of your AI team!

Step 7: Human review and management

While ping-pong prompting empowers the AI to help you refine your prompts, your role as the “manager” of this process is crucial. 

As Mercer emphasizes in the workshop: 

You wouldn't hire a team of humans and then completely step away, expecting perfect results without any guidance or oversight. 

The same principle applies to working with AI.

Even when the AI indicates a 10/10 understanding, take the time to review its reasoning. 

Check the questions it asked, and the answers it generated. 

And ask yourself:

• Does this truly align with my intended outcome?
• Are there any nuances or subtleties that the AI might have missed?
• Are the AI's suggested answers and refinements actually helpful and accurate?

Don't blindly accept everything the AI suggests. 

Your judgment and expertise are still vital.

Use your understanding of your goals and audience to ensure the final prompt is truly effective.

This final review step ensures that the “ping-pong” game leads to the best possible results for your specific needs.

Answers to your most common questions about the ping-pong Prompting technique

What exactly is ping-pong prompting? 

Think of it like this: instead of just throwing one prompt at the AI and hoping for the best, you start with a prompt and then have a little back-and-forth with it. You ask the AI to think about your prompt, tell you what it understands (and what it doesn't!). Then, you use that feedback to make your prompt even better. It's like a conversation that helps you and the AI get on the same page.

How is ping-pong prompting different from traditional AI prompting methods?

ping-pong prompting is about actively involving the AI in the process of figuring out the best way to get you the results you need, rather than just blindly following your initial (possibly not-so-perfect) instructions.

Is it hard to learn and use ping-pong prompting? 

While it might seem like more steps, ping-pong prompting is designed to get you to the right answer faster. Think about it – how much time have you wasted tweaking prompts that just weren't hitting the mark? This method helps you quickly pinpoint what needs to be clearer, so you spend less time in that frustrating trial-and-error zone. Plus, getting the AI to do some of the thinking (like generating its own questions) speeds things up considerably.

Can ping-pong prompting be used with any type of AI or just specific models? 

While Mercer showed you this using Gemini, the good news is that the core idea of having a conversation with the AI to refine your prompts can be applied to lots of different AI tools and platforms. As long as the AI can understand your prompts and give you feedback (even if it's not in the exact same rating scale we talked about), you can definitely use the ping-pong approach to get better results.

Ready to start your first ping-pong prompt challenge? 

Now, it’s your turn to put the power of ping-pong prompting to work and see instant improvements in your AI results. 

Here's your simple plan to get started today:

1. Choose a task you need to accomplish with AI right now or something you've been meaning to try.
2. Write your initial prompt: Don’t overthink, but try to write down the best possible instruction you can think of.
3. Ask the AI to analyze your prompt: Make sure you ask for desired result and hidden intention. 
4. Start your ping-pong: Follow the steps outlined in this guide and the video. Iterate with the AI based on its feedback and your own assessment. You will soon reach a prompt that delivers the results you're looking for.

Don't be afraid to experiment and see how this technique can transform your AI interactions. 

If you want to learn other advanced techniques to work with AI and get better results, make sure you watch Mercer’s free Gemini AI workshop.

Good luck prompting!