Hello everyone. So as you know deep research is now a feature in many AI models and advanced research capabilities are now built directly on top of tools like chat GPT's deep research Gemini enabling users to generate evidencebacked insights analyze complex topics and even extract information from visuals all with a single prompt. But these tools won't reach their full potential unless you know how to prompt them properly. No matter what you're researching, the quality of your prompt determines the depth, the accuracy, and reliability of the answers you receive. In this video, I'm going to be sharing with you 10 essential prompt hacks, plus one bonus one that will help you unlock the power of deep research and ensure that you get the best possible results. And not only that, but I'll leave you the full prompt structure in the end so that you can use that as the basis for your prompt. So, let's get started. So, for our first hack, we need to set a clear research goal. And this is really important because just like a well- definfined research project, the clearer your question, the sharper your answer is going to be. So an example of a bad prompt would be what are examples of AI failures? And the reason why this is a bad prompt is that it's too vague. There's no context or time frame and it doesn't guide the type of insight that we need. So a much better prompt would look something like this. investigate why AIdriven turnaround strategies failed in large corporations post 2018 and identify core strategic missteps. And the reason this is so much better is that we have a focus on a specific scenario. So we're talking here about turnaround strategies where it sets a time frame. So it's telling deep research to focus mainly on articles and what's happening post 2018. So overall this narrows the scope. It encourages critical thinking and then it produces the more relevant and usable insights. So for our next hack, we're going to be focusing on adding context, constraints, and keywords. And this is really important because AI always performs better when it knows what boundaries to follow and what to focus on. So I'm going to add to my prompt and I'm now going to say, investigate why AIdriven turnaround strategies failed in large corporations post 2018. focus specifically on Fortune 1000 companies undergoing digital transformation using AI whose only sources from 2018 to 2025 such as analyst reports, peer-reviewed journals, and strategy case studies. Prioritize insights related to strategic misalignment, organizational resistance, AI overdependence, and execution risk. Now, let's focus on some of the elements of this prompt. We've now narrowed the context. So, now it's focusing on Fortune 1000 companies and digital transformation. We've added some important constraints such as limited time frame and the source types. And now we have really important keywords that we wanted to focus on such as the analysis on recurring failure themes. We wanted to look at strategic misalignment, organizational resistance, AI overdependence and execution risk. Now this focus on keywords is extremely important and it helps direct the whole deep research around these concepts. It will extract other information as well, but it will make sure that it's focusing on the key areas that are relevant to us. Hack number three, you need to ask for direct quotes and citations. So, I'm going to add part to the prompt here that says, for each claim, include a direct quote and APA citation. This is really important because direct quotes they boost credibility helped with fact-checking and it will help you straight away to identify whether the AI is hallucinating the information or whether it's actually extracted that information from an actual source and this save you a huge amount of time in ensuring that the information that you have in your deep research report is accurate and up to-ate. So, on to hack number four, and this is surprisingly one that is always overlooked because people are in a hurry to start their deep research, and it's basically to suggest a format that you want your output to be in. Spend two or three minutes in the beginning to really think through the way you want your report to look like and then specify the format. This will save you hours of work later on. And this could be something as simple as adding this to your prompt. Organize the output as follows. Context and overview. Illustrative case studies. So have it mention case study 1, case study 2, case study 3. Underlying causes of failure or challenge, key insights and strategic takeaways. Then recommendations and citations and source references. And in that way you get their output of the deep research in exactly the same style and format that you want. Next hack is extremely important for prompting in general. And this is basically setting the role and the tone for your prompt. So for example in our deep prearchet prompt we can say something like you are a strategy consultant with over 20 years experience use a formal professional tone and this is really important because the tone not only affects the depth but also critically the terminology that deep research will come back with. So if I specify here that I'm a strategy consultant I want this to be more of a professional tone. It will match the tone that is required for executive readers not blog readers and also not academic writing. So again, an important point to keep in mind, adding a prompt that says after reviewing each case, critically analyze the assumptions or decisions that influence the outcome. Use language such as despite initial expectations the results suggest. A key oversight was this reflects a misalignment between an underlying assumption that proved flawed was and this raises questions about the validity of aim to uncover not just what happened but why it happened and what deeper implications it reveals. And this takes your deep research to a completely different level where the model is now not only describing what happens but also critically evaluating those assumptions, the results, looking at opposing arguments and bringing in different elements that will really enhance the final output of your research study. Now for the next hack, we needed to look at depth as complexity as deeper analysis and not just quantity of analysis. In this example, I'd add to my prompt for each company, go beyond surface level outcomes, break down leadership, mindset, internal resistant, market response, and long-term impact. And what I'm doing here is I'm forcing it to get multi-dimensional insights from these concepts and not just summaries, which a lot of the models tend to do when they come back with their results. Now, another really important hack is to extract insights from non-extual data. So just adding a simple prompt that says use a variety of data sources such as visuals, tables, strategy maps, and performance dashboards or whatever data sources are relevant to your research. And then interpret what these materials reveal about decision-m patterns, performance gaps, and failure trends. Just adding this line will prompt it to look at other key data sources. And this is really important because sometimes the key findings are found in images, in tables or charts. So to make sure that you have that in-depth research and that you're covering all the data sources, adding this to your prompt will really enhance the output that you get. Now this text hack is really relevant when you want to direct the research and enhance it by giving it something that you're already working on. So most of the deep research models now allow you to upload from your computer or connect to your Google Drive. So definitely use this hack when you have a direction for the research and you want to enhance or update your knowledge in that area. And now the final hack is iterative refinement. Even deep research rarely comes back with an excellent first draft or an excellent first output. And sometimes I've had the first draft come back with really poor missing elements. And by asking it to focus on the parts that are missing and then conducting the deep research again, I've had much better outputs. As we know that now in AI for example, the deep research limit has increased. So don't be afraid of writing in all your comments and missing elements that came back from the first draft and then resending that for a second deep research. you will see the enhanced response and how it will now start to extract even deeper information that are relevant to your suggestions. Now, the bonus hack before I even start my deep research, what I always like to do is use other models to help me refine prompt and to really focus my research before I actually start using it for deep research. So, for example, if my question has more of a reasoning element that involves justification or different viewpoints, I would often start with 04 mini or 04 mini high. Or if I'm looking for general suggestions about how to structure my prompt, I would start with chat GPT40 just to lay down the full parts of the prompt and how I want to refine it. And then once I have my final prompt, then I would then add that to deep research. So, for example, I'd come here and I'd say, I want to conduct deep research on examples of AI failures, write a full detail prompt that will help me get the best results, include all the essential elements such as context, output, critical elements, and so on. And you can add the other features as well. So, we will enter that. Okay. And now you can see that the prompt is now detailed. It's now asking it to conduct a deep evidence-based analysis. It's given it the context, the goal, the sources to use, include nonextual insights, key themes to explore. Here it's added optional but powerful, the output format because my chat GPT now remembers what I've been working on. And it's added these elements. You can get that from the prompt that I will leave you below. And also start by adding the role as we said in the prompt. And so we can play around with this prompt until we have the full correct structure. And you can see how this is so much more powerful. So, I hope you found these tips and hacks useful. Let me know in the comments below if there are other hacks that you're using in your deep research prompt that you have found useful. And I hope to see you in the next video.