Transcription
So, GPT5 recently launched, and the internet is completely divided. I've been testing it out for weeks, and at first, I wasn't getting the results I hoped for. But once I discovered the techniques I'm about to show you, I was able to unlock some impressive capabilities that most people are missing.
So, in this video, I'm going to show you exactly how to get way more out of GPT5, including the specific prompting techniques that transform mediocre results into powerful outputs, some newer features and settings that you may not be aware of, and practical use cases that will save you hours every week. Let's get started.
I'm starting with one technique that will immediately improve the responses that you get from GPT5. It takes 10 seconds to add to any prompt, and you'll see the difference instantly. Then I'll show you two more prompting techniques that make everything else work even better. Here's that first technique.
I'm calling them magic phrases because of how dramatically they change your results. Add these to any complex prompt: "Think step by step," "Be extremely thorough," "Take your time," or "This is critical to get right." So, let me show you what this looks like.
First, we're going to start off with a standard prompt. "We are a digital marketing agency. Write a professional email to our client explaining that their website redesign project is 20% over budget due to additional functionality requests. Include next steps and maintain the relationship." Now, watch what happens when I add a few of the magic phrases we just talked about.
So, I'm going to say, "Think step by step and be extremely thorough." And then repeat the same prompt that I just read. And then I'm going to say, "This is critical to get right. Only provide suggestions if you're highly confident." And let's see the difference.
Now, you can see the comparison between the two. The second response is way more comprehensive, thoughtful, and actually usable. Now, here's why this works, and this is the key insight that most people are missing.
GPT5 works completely differently than the previous versions. Remember when we used to have to pick from like eight different models in that drop-down menu? Well, OpenAI reduced the model choices and built something entirely new. Well, now we have the GPT5 auto model, which decides how long to think based on your prompt. And then we have the instant, the thinking mini, and the regular thinking model, which is their deeper reasoning model.
With this consolidation, there's an internal routing system that looks at your prompt and decides, "Should I be analytical? Should I be creative here? Do you need me to think really hard about this?" And if your prompt isn't clear about what you need, it might choose the wrong processing approach. These trigger phrases act like signals to that routing system. You're essentially telling GPT5, "For this specific prompt or task, I need your best thinking."
And here's the bonus technique for accuracy critical tasks: "Only provide an answer if you're highly confident, or if you don't know, clearly state 'I don't know' rather than guessing." This eliminates hallucinations by giving the AI permission to admit uncertainty instead of making things up.
Now, these phrases work best with the auto model, which is the default, giving you ultimate flexibility within the same conversation. The auto model lets you seamlessly switch between creative and analytical tasks in the same conversation, whereas the dedicated thinking model is built specifically for complex reasoning.
Now that you know how to signal for better thinking and higher accuracy, here's how to make that thinking even better. Something I've had good success with is creating a quality rubric where I tell it what good looks like. And then I ask it to take that rubric or criteria and use that to refine its answer and give me the best output. Let me show you what I mean.
Let's say we're working on a LinkedIn post. A standard prompt to ChatGPT might look something like this: "You are a career coach helping a mid-level professional build their personal brand. Write a LinkedIn post about the importance of networking for career growth. Keep it under 250 words and include two to three actionable tips. Do not use corporate jargon."
Now, here's how we add the quality rubric to this prompt. So, we're going to put the same exact prompt as before. And then under "Quality Rubric," we're going to put a few bullet points on what's most important to us about this. So, I have here: "Strong storytelling." So, I want it to have strong storytelling ability here. "Originality." So, it needs to be fresh and avoid clichés. And then "A strong call to action." So, that's the other thing that's important. And then before responding, at the bottom here, I say, "Use the quality rubric to iteratively refine your answer. Run up to five internal improvement cycles." So, do this five times without me seeing it, and output only the final version that meets all the criteria.
Now, let's look at the side-by-side output that each of these generated. Look at how much more fresh, personal, and engaging the rubric version becomes. So, this makes GPT5 do multiple drafts internally and only show you the best version. It's basically like having a built-in editor working behind the scenes, and this technique has been working really well for me.
Now, the first two techniques will dramatically improve your results, but there's one mistake that can sabotage everything, and most people don't even realize they're doing it. Tip number three is to never give conflicting instructions. GPT5 actually gets confused and performs worse when you give it conflicting instructions in the same prompt. The key is to be ruthlessly clear in everything that you communicate.
For example, don't say "be brief" but also "provide detailed explanation," or "be casual and conversational" but also "maintain strict professional formality." They're contradictory. So, instead, be very clear and consistent. So, you could say, "Provide brief summaries, but include detailed explanations when specifically requested," or "Be friendly but professional, like talking to somebody that you work with."
Now, if you want to double-check to make sure that your prompt doesn't have any contradictions, then you can always use the OpenAI prompt optimizer tool, and it will make suggestions for you as to what it should, what you should change, and it will also tell you the reason behind the change it's made. This principle of ruthless clarity dramatically improves the quality of the response that you get from GPT5 by letting it focus its brainpower on your actual request.
And speaking of clarity, that's exactly what separates mediocre prompting from great prompting. Now, these enhancement techniques will definitely help. But here's the thing: if your foundational prompting structure isn't solid, then even these powerful add-ons won't reach their full potential.
If you want consistent professional results instead of hit-or-miss outputs, then I suggest that you take a look at this free advanced prompt engineering guide from HubSpot. I'll link it in the description box below. This guide is a 7-day program that teaches you how to be more systematic in your prompting. You'll learn how to organize information, structure your requests, and build the detailed instructions needed for professional-level outputs. You'll get proven frameworks for complex business tasks, step-by-step techniques that will help the AI think through problems methodically, and ready-to-use prompt templates that you can copy and customize. Plus, it includes real business examples for marketing, sales, and analysis that you can start using immediately. My favorite part is the section on persona crafting in day five. It goes way beyond just telling the AI to act like an expert and shows you how to build a detailed persona with a specific communication style and reasoning pattern, giving you incredible control over the output. Download this guide using the link in the description, and thanks to HubSpot for sponsoring this video.
Now that we've covered the prompting foundation that makes everything else work better, let me show you how to take your inputs to the next level. If you're only using ChatGPT with text, then you're missing out on some powerful capabilities. And that brings us to tip number four.
Here's a capability that most people overlook that can completely transform how you work with ChatGPT, and that's giving it multiple types of inputs at once, like images, PDFs, documents, voice, or audio, etc. Let me show you with a practical business example.
Let's say we owned a pool company, and we created these flyers, but they weren't converting as well as they could be. We can upload an image of it and give it a prompt to get expert-level feedback. All right. So, I'm going to click on the plus sign right here by the prompt box. I'm going to click on "Add photos and files." I'm going to find my file, upload it, and I'm going to give it this prompt: "You are a marketing consultant with 20 years of experience in direct response for home services. Your task is to analyze the attached flyer for my pool company. The goal is to get specific, actionable feedback that will increase lead generation and sales conversions." And I'm asking it to structure this in two parts: "Expert critique, actionable recommendations," and "Be direct and assume that I have a budget for full redesign of the flyer." So, I'm being very specific.
And look at that analysis. It's identifying layout issues, call-to-action placement problems, visual problems, giving specific actionable recommendations. So, ChatGPT can actually see and understand design principles, not just read text. One thing I try to do whenever possible is I try to show ChatGPT what good or bad looks like with visual examples that go along with my prompt. For example, if you're putting together a client proposal and you found an example of one that you already like, give it the PDF example along with your prompt so that it has a clearer understanding of what you're looking for.
Now, beyond analyzing visual content, there's another input method that's incredibly powerful but often overlooked: voice input. Instead of always typing, you can just talk to ChatGPT. This is great for when you're on the go, on your commute, or you just can't type at that moment. So, to do this on the mobile or desktop, you're going to click on the microphone button to say your message and have ChatGPT respond with a written response. Or, to access the advanced voice feature where you can have a back-and-forth conversation, you're going to click on the soundwave button there, and then you just start talking, and that'll allow you to have a whole back-and-forth.
These multimodal capabilities are available both on the free and the paid plans, but the paid subscribers get higher usage limits. The key is remembering that ChatGPT isn't just a text tool anymore. It can handle multi-step complex projects.
Let me show you what I mean with this next tip. One of the things that OpenAI emphasized about GPT5 is its enhanced coding abilities. Now, before you say, "Well, I'm not a coder," I'm not either. But this is a super useful feature that makes it really easy for you to design a personal tool for yourself that solves a real problem.
Since I'm interested in health and wellness, and one of GPT5's strengths is providing more accurate health information (though you should always double-check things with a professional), I'm going to show you how to create a functional health dashboard with a single prompt. This is where canvas mode becomes really powerful.
To access it, what you're going to do is click on the plus sign and then click on the three dots at the bottom and then select "Canvas." And I'm going to give it this prompt here: "Using canvas mode, build me a personalized weekly health hub. The overall design must be beautiful, clean, modern, and mobile-friendly, inspired by a calming wellness aesthetic." And I'm giving it very clear instructions on how I want this to look, the color palette, the layout, styling, etc. I'm even telling it the functionality that I want as well, like when it, when the button's clicked, it must generate a simple one-sentence instruction for a quick mindfulness exercise, etc. And then arrange these four modules within the grid to create a balanced and intuitive dashboard.
So, it works for about 2 minutes to complete this application. You can see it's designing the user experience. It's choosing the colors, creating the layout. It's even putting the food options in there. And when it's all done, here's what you end up with. It looks really good. You can see that it's actually functional. So, when you click on "Generate a healthy lunch idea," it does that. The quick workout section works as well. We can log our hydration here, and it gives you a mindfulness moment, too. So, "Inhale calm, exhale tension. Silently say those words once you breathe." Awesome. What's amazing is that you don't need any coding knowledge to create something this functional and well-designed. Canvas mode is available to everyone, although paid subscribers get higher usage limits.
This next feature is genuinely game-changing for how you work with ChatGPT, but most people don't even know it exists. It's called branches, and it's basically like having multiple versions of your conversations. Here's the problem it solves: How many times have you been working in ChatGPT and you wanted to try a different approach, but you didn't want to lose your original work? Before branches, you'd have to start a new conversation and lose all of that context. Now, you can explore multiple directions from the same starting point.
Here's how it works. Let's say I'm planning our company's annual fun run charity event, and I want to figure out the best approach for how I should organize the event. I'm going to ask ChatGPT for an initial approach by giving it this prompt: "Based on the information above, please propose an initial plan for organizing the fun run. Organize it in a traditional sequential checklist approach where each person's responsible for their own area, etc." Now we have an initial solution.
Now, instead of starting a new chat over, I hover over that response. I'm going to click on those three dots, and then I select "Branch in new chat." Now, this creates a completely separate conversation thread. So, you can see now I have two tabs open. I have the old conversation and the new one. So, it's basically now a duplicate of what I had.
Now, in this new branch, I can explore a totally different approach. Maybe doing an all-hands-on-deck collaborative approach instead. And the best part is, if I wanted to, I could create even more branches to test other ideas. So, now I have two fully developed event plans, all building from the same original context. Each branch preserved all the background information about my planning team and the event goals but explored different directions. This is perfect for complex decision-making, creative alternatives, testing different solutions, or when you're genuinely not sure which path to take. The best part is that this is available to all users of ChatGPT, free and paid. Once you start using branches for decision-making, you'll wonder how you ever made decisions without them.
Tip number seven on the list is to leverage projects. Projects just became available to those on the free plan, and if you haven't tried this feature out, you're missing out on one of the most powerful features of ChatGPT. I've done a dedicated video on projects before on my channel, so I'm going to keep this brief.
Projects provide dedicated workspaces where ChatGPT actually gets smarter about your specific work over time. Instead of starting every conversation fresh, you can create a project by clicking on the left-side panel of the screen right here. And then you can upload your relevant files over here and then give it custom instructions, and ChatGPT maintains context across all your discussions about the topic. So, whether you're working on a client proposal, a content series, or a research project, it remembers everything and builds on previous conversations. So, it's not just a way of organizing your chats in a folder. The real value is in the ongoing memory. So, you can upload your brand guidelines, previous work samples, or research documents, and ChatGPT remembers them and references them automatically in future conversations. Free members can now upload up to five files per project, plus subscribers get 25 files, and pro users get 40 files per project. You can create unlimited projects and customize them with colors to keep everything organized. Check out my dedicated projects video if you want a breakdown on how to actually set these up effectively.
Tip number eight is to activate study mode for interactive learning. One of the newest features of ChatGPT is study mode, which helps you learn any subject, whether you're a student, learning something for work, learning a new language, or trying to master something. But here's what makes it different: instead of just giving you the information, study mode turns it into an interactive tutor, which challenges your thinking and gives you insights that you can apply.
Let me show you with a business example. Say you're thinking of launching a new product, but you're not sure if you've thought of everything. You wish you had access to a team of advisors that could ask you tough questions to figure out what you're missing. To activate study mode, you're going to click on the plus sign and then the three dots at the bottom, and then you're going to click on "Study and Learn." Then I'm going to give it this prompt and ask it to create an AI board of experts to help me evaluate my business idea, which is a mobile app where parents can input their child's information and interests and have a moral of the story, and an AI generates a unique illustration of the storybook. So, basically like a custom storybook.
Now, when I give it this prompt, it comes back with three different questions: one from the "Offer Architect," the "Customer Champion," and the "Brand Innovator." These are questions that I need to answer. So, for example, the Offer Architect asks, "You know, parents can already get free bedtime stories from ChatGPT or buy thousands of cheap kids' online books on Amazon. Why would they pay for your app? And at what price point would they feel it's a no-brainer?" That's a great question. The Customer Champion says, "Parents are busy and overwhelmed. What specific pain points are you solving for them?" And the Brand Innovator says, "What makes your app not just another AI toy, but a remarkable product that parents would rave about and share with friends?" All good questions.
So, I'm going to respond back to the Offer Architect first and see what feedback I get. And then after I answer, see, it's providing strategic insights about why that response matters for my business model. Then it guides me to the next expert's question. This interactive approach means you're not just passively consuming answers, but you're actually working through problems with expert-level guidance. And this works for anything you want to learn deeply, whether it's technical stuff, creative projects, academic subjects, or professional skills.
Now, a lot of the stuff we've been covering in this video has been about how to make ChatGPT think better. But now, I want to show you how to make it write like an actual human. Now, here's a technique that's going to solve one of the biggest complaints about GPT5, and one of my biggest pain points with it when I first started using it, and that is that GPT's natural writing style is not very humanized. It sounds corporate, robotic, and just not like a real person wrote it. But after playing around with it a lot, I think I've come up with something that works really well.
Here's my complete framework, and you can copy this directly. So, the first thing I do is I give it a couple of writing samples of work that I've done in the past. And then I say, "Okay, ChatGPT, your task is to write a newsletter for my audience. Before you begin, you must understand two things. First, analyze the writing samples that I gave you to create a personal voice guide for me. Pay attention to my tone, sentence structure, and writing philosophy." And then you'd have the writing samples for it to refer to. And then second, "You must understand unbreakable writing rules. Write in a clear, human way. Keep sentences under 20 words. Use everyday words at an eighth-grade level. Avoid jargon and buzzwords like 'leverage' and 'cutting edge.' Connect ideas naturally with 'and,' 'but,' 'so' instead of formal words."
And based on my prompt, here's what it generated. See how natural and human-like this sounds? "Hi friends. A few years ago, I was guilty of trying to change everything at once. New workout plan, new business goals, new morning routine, all starting Monday. And by Friday, I was exhausted, discouraged, and right back where I started." I really like that. The key is giving ChatGPT both your personal voice samples and rules about how humans actually write and talk. And this technique works for emails, blog posts, social media, anything where you want to avoid sounding like a corporate robot and you really want to sound human and real.
Now, tip number 10 is to use agent mode for complex tasks. Agent mode is where ChatGPT becomes truly autonomous. Think of it as putting ChatGPT on autopilot for complex, multi-step tasks. It can browse the web independently, research information for you, and complete entire workflows on your behalf. What makes it powerful is that it brings together ChatGPT's existing capabilities like image generation, deep research, and connectors. So, it can pull information from connected apps like Gmail, create visual content, and even handle tasks like booking appointments all autonomously while you watch it work in real time through the browser window.
To activate it, you'll need a paid plan, and you'll select "Agent" from the tools drop-down menu. Let me show you with a quick example. Let's say you're a dog trainer in Boston and you're trying to get more leads, so you want to create a lead magnet, but you're not sure where to start. With agent mode, I can give it one comprehensive prompt and watch it handle the entire process.
So, I'm just going to say, "You are an expert marketing strategist creating a lead magnet for a local business. I tell it what business it is, what the lead magnet that I want to create is: a new puppy survival guide. Your mission is to create a complete lead magnet package. So, first, it's going to analyze the pain points of new puppy owners in Boston, in the city environment, and in the coming winter. And then it's going to create a resource, and then it's going to create a cover page as well with the image generator model. And then it's going to compile everything into one PDF document."
So, once I give it that prompt, you can see agent mode working autonomously. It's researching different challenges, creating strategic content, and trying to come up with professional imagery. And after running for a few minutes, here's what we get: an entire lead magnet ready for lead generation. We've got the cover page there, all the content that I asked for, and it looks very professional. This saves a ton of time. The key is that agent mode handles workflows with multiple steps that would normally require multiple tools and skills. This is basically like having a virtual team member who can research, write, design, and execute all while you're focusing on more higher-level strategy.
Now, tip number 11 are the advanced navigation and control features. Now, there's a few features that you may not be aware of that can really help to streamline your workflow and give you way more control over how ChatGPT behaves.
The first one is the edit message feature. Here's a time-saver that most people miss. Instead of sending new requests when your previous prompt didn't work, you can actually click on the three dots on any message and select "Edit" to refine your request without losing the context or having to start all over again. And it actually uses up less context within the context window because you have less messages that you're sending.
The second tool is connectors. For paid members, connectors let you integrate with other external tools and services. You can connect things like Google Drive, Calendars, Gmail, or even project management tools to pull in real data without switching between apps.
The third feature is personalities. This is something that they launched when GPT5 first came out. And if you don't like the default personality of the ChatGPT5, you can actually select different ones: "Robot personality" for more technical work, more precise, direct, and no fluff. "The Listener," whose personality is a little bit more warm, detailed, and supportive. "Cynic" for honest feedback and critical assessment. "The Nerd" for enthusiastic, detailed explanations.
And the fourth feature I wanted to talk about is the legacy models. If you still prefer to access the older models like the 4.0 model, you can do that as long as you have a paid plan. Now, to do this, you're going to click on "Settings," and then you're going to go to "General," and then you're going to see this toggle option for "Show additional models." Make sure that that is toggled on. And then this brings back the previous models when you need them for specific tasks when you prefer their behavior. And then when you actually go to select the model from the drop-down menu on the main screen, you'll see all the old models. These are relatively small things, but they can really change how you work with ChatGPT on a daily basis.
And tip number 12 is to manage your AI memory. Here's something that some people might not realize: ChatGPT has a memory system, and outdated memory can hurt your results. ChatGPT remembers your preferences, your background, and your work situations across conversations. And if that information becomes outdated, then your responses get tailored to the wrong context. So, to double-check this, you're going to go to "Settings," and then click on "Personalization," and then click on "Manage" right beside "Memory" to see what ChatGPT knows about you. So, make sure you review and delete outdated information.
These 12 techniques transform GPT5 from disappointing to genuinely powerful. The prompting fundamentals that we covered first, like the magic phrases, the self-critique, and clear communication, will improve any request you make. But combining them with systematic frameworks like the HubSpot guide creates truly professional results. Try a couple of these techniques this week and see how they transform your GPT5 results.
And if you want to learn more, I've just started my own community called the Innovator Circle on School, where I'll be doing weekly videos, live lessons, and you'll get access to a community of like-minded people who are learning about how to leverage AI and automation in their business, work, and life. You can find the link to join in the description.
Now, if you want to know more about how to use AI to level up your work and your life, then click this next.