ChatGPT is the world’s most widely used GenAI application, and has of course made an impact in education – for better or for worse. In this free mini-PD video, I demonstrate how to use ChatGPT and run through all of the features of the free and paid versions of OpenAI’s product.
This video is part of a series running throughout August. To access the upcoming videos, make sure to join the list here.
How to Use ChatGPT
More Information on ChatGPT’s Features
In this video, I discuss several recent additions to ChatGPT. You can read more about those features here:
First Impressions of ChatGPT’s Study Mode
Everything I’ve Learned so far About OpenAI’s Agents
Transcript
Transcribed with otter.ai, reformatted with Claude
Hi, I’m Leon Furze, bestselling author of Practical AI Strategies, consultant, and a PhD student studying generative artificial intelligence in education. I’m also an educator myself. I’ve been in the classroom for almost 20 years in K-12, higher education, and vocational education. For the past few years, I’ve worked with thousands of educators across all of those sectors, in Australia and overseas, and I’ve heard about a lot of the problems, challenges, and opportunities with generative artificial intelligence.
I know that one of the biggest challenges educators face is keeping up with the rapid pace of these technologies. The technology companies seem to drop a new feature about every six hours or so it seems. In just the past few weeks, we’ve seen releases like ChatGPT Study Mode, new releases into education from Google in their Gemini chatbots, and various other platforms.
In this series of mini PD, I’m going to be running through some of the major platforms and talking about how they work, what the different features are, and what’s the difference between the free and paid versions and different models. Now, all of this information is correct as of August 2025, but we all know that it can change very rapidly. So I’d encourage you to jump onto my blog, leonfurze.com/blog, and make sure you stay up to date with the articles as they come out.
In this first video PD, I’m going to help you to understand ChatGPT, which is still the most commonly used platform, and a platform which has so many features that it can be really overwhelming to understand which model to use and whether you should pay the 20 US dollars a month for the paid version. I’m going to walk through some examples in an educational context, both from the point of view of a teacher and a student, to explore all of these different modes. Stay tuned and sign up to the mailing list for future video PDs, which will be coming out once a week for the whole of August.
Free Version Features
Let’s take a look at what we’ve got in the free version first. We’re looking at the free but logged-in version of ChatGPT. You can see in the dropdown, we’ve only got two options: ChatGPT and ChatGPT Plus, which we have to pay for. We can add photos and files. We can connect things like Google Drive, OneNote, SharePoint, and we’ve got all of these various options here, which I will go through one at a time, including pointing out which ones are slightly different in the paid version. We can dictate and we can use voice mode. It’s not as advanced as the voice mode that you get in the paid version. We’ve got this panel on the left where we’ve got new chats, search chats, library. We can look at Sora, but we don’t get much use of that in the free version. We can use GPTs, the custom chatbots, but we can’t create them.
Settings Configuration
I want to pop down into the settings and show you a few important things. Straight off the bat, I’m just going to put dark mode on, just for the sake of it here, and I can change some of my personalisation options. I’m going to disable custom instructions. Those are little instructions that we can give to change the way that every single ChatGPT interaction works. I’m also going to turn off memories, and you can delete any existing memories that you find in there so that it’s not storing that information. We can connect or disconnect those apps that I mentioned before. I’m going to turn off “improve the model for everyone,” which means it’s not going to use my chat data to train future models. I’m going to, in my security settings, probably enable multi-factor authentication, just to make it a little bit more secure. I can upgrade my account from here, but I’m just focusing on the free one for now.
Basic Functionality
Let’s start off with a really simple overview, a simple prompt to show the kind of typical response produced by ChatGPT. “What’s the capital of France?” Here it is. There’s just a bit of an arbitrary response. I can show you some of this stuff here: dislike, like, read aloud, opening Canvas, or switch models. In the free version, you get limited access to different models, but I will talk through the difference of these models as we go along.
Image Recognition
I can drag and drop an image in, and I can use image recognition, something like, “Can you tell me anything about this image?” And that will work on any images. It will work on photos of handwriting with a reasonable degree of accuracy as well. Here you can see it does a pretty good job of articulating what that image looks like. So even in the free version, we can upload images.
Document Upload
We can also upload PDFs, like this one. This is just an open access article that I’ve got a copy of here. You can get a summary of that. It will read the document and pretty swiftly be able to do things like summarise, pull out quotes, pull out key information. So we can upload lots of file types in the free version: Word documents, spreadsheets, PDFs, all kinds of things can go into there.
Study Mode
I’m going to open up a new chat and turn on Study Mode. That’s a brand new feature as of a few days ago when I’m making this video. Study Mode has a little instruction in the background which tells it to avoid giving answers immediately, check backwards and forwards for understanding. You can see here, I’m using a literature class example trying to get a definition of synecdoche in figurative language. It doesn’t give me the examples straight away. It prompts backwards and forwards. I’m not super impressed by Study Mode, and I wrote an article about that on my blog, which I think you should check out. I’ll pop a link to that below this video if you’re looking at it on my website, or I’ll include it in the email. It’s not particularly impressive, but it is available now in every single version.
Image Generation
In the free version, you can also generate images. You get a limited number of images per day—I think it’s five images per day. There, it does take quite a while, but there’s my photo of a cat. If I click on that in the sidebar, you can see the chat window has changed, and it gives me an option to change the styles a little bit like Adobe Firefly, if you’re familiar with that platform, although simplified. I’ll try to generate a second image. Again, I’m just speeding up the video here, but it actually hits a problem, which is quite common I found in the free version of ChatGPT: “A network error has occurred.” I’ve rarely seen that in the paid version, and it does concern me that this error in particular seems to come up in the free version and not so much the paid.
Reasoning Mode
I’ve turned on the button that says “think longer,” and it’s actually got a little light bulb there in the word “reason.” This is going to switch it into using a model called O3, which should give a more robust answer. You can see it’s done this little thinking mode here. It produces a table, but I’ve noticed right away that the GPT-O3 column is actually almost totally hallucinated information. So I’m really concerned there again—on the free version, it does seem to hallucinate significantly more than in the paid $20 a month version. When you use O3 in the paid version, as I’ll demonstrate in a short while, you do generally get pretty accurate responses. So it does concern me that that was hallucinated.
I’m going to throw in another reasoning question, maybe something a bit more straightforward: “What’s the best approach for chopping wood when the wood is well-seasoned but quite knotty?” This is a problem that I’m dealing with myself at the moment out here in southwest Victoria, where it’s freezing cold. It does its little thinking thing and gives quite a detailed response. So reasoning or thinking mode, using O3, just improves the overall quality generally, but not always, as we saw in that earlier example.
Web Search
You can force search to turn on, and this forces ChatGPT to use Microsoft Bing search and do an internet search. So if I want the accurate weather, I’ll get that. But at this point in the video, I’ve actually hit the free plan limits. You’ve seen everything I’ve done so far, so you can see how quickly I’ve hit that limit.
Paid Version Features
I’m going to switch over to my paid version. Currently, because I’ve been testing out the agents feature, I’m actually on the highest tier, which is the $200 US a month tier, which is ludicrous, frankly. But all of the features that I’m going to demonstrate other than agents are available in the $20 US a month tier.
Deep Research
I’m going to fire up Deep Research first. That turns on the little telescope and the research button, and this will use one of those O3 reasoning models, but also do a lot of internet searching—20-30, up to about 50 sources of information. So if I say, “What are some of the challenges faced by secondary teachers in Victoria, Australia?” it will ask me some follow-up questions, and I can just say, “Look, all of the above.” I want all of those areas there. The Deep Research model takes seven minutes to produce this report—20 sources, 80 different internet searches. We can scroll down the side, we can look at all of those sources of information, and it gives us this lengthy but much more accurate report.
You can use Deep Research on the free version—you get five per month—but it does seem to be a slightly lightweight version. On the paid version you get about 30 a month.
Canvas Mode
I’ve enabled another mode. You might have seen Canvas in the dropdown menu of tools. Canvas is like an editing tool, like a word processor or a code editor, depending on how you use it. I’ve asked it to write a little newsletter about Fitzroy. I also gave it the task of going online and finding some recent events. You can see it handled that pretty well, and it’s produced this little newsletter intro.
I can use the toolbar on the right there. I can add emojis if I really want to, shorten the text, lengthen it, or I can highlight, I can type over the top, and I can actually prompt ChatGPT directly to focus in on a piece of information here. So I’ve highlighted the second half of the first paragraph, and I’ve said, “Provide me more information with this.” You can see over on the left there that it’s expanding on that section. It’s going to fall back on its internet search from earlier, looking at a few websites around Fitzroy Market and other places, and after a little bit of thought, it’s going to come up with some updates just to the second half of that paragraph. So I could edit directly over the top of this, much like using a word processor. You can see here, it’s just rewriting that little short passage. I can also highlight different parts of the text and use a few simplified formatting tools—bold, italics, changing to Heading One, Heading Two, and doing some of the paragraph style and all of that stuff. It’s a lightweight text editor, and it’s something you might use if you’re doing a lot of content creation within ChatGPT.
Code Generation
It also has a mode which is specifically designed for writing code. If I come out of the text editing mode, but I leave Canvas turned on, and then I’m going to put in a second prompt here: “Using this information, create a community web page.” I’m going to give some subsequent instructions. I could talk about the style of the website page, I could talk about the colours and so on. I’m just going to say here: “Create a community web page with a simplified calendar app built in, relevant, interesting stock images, clean, smooth scrolling, aesthetic,” and I want it self-contained so that it can actually run within the ChatGPT window. I’m going to demonstrate a few different variations on this in a moment, because there’s quite a few things that we can do within this code execution window here.
Still a pretty simple prompt. If you’ve read my blog for a while now, you know that I don’t really believe in prompt engineering. I don’t think that’s really a thing. We just have to provide lots of context, which we did earlier by telling it to go and search the internet. When I run it from the little button in the top right-hand corner, you can see it’s had a crack making a little website based on that newsletter. And there’s our upcoming events calendar. We’ve got a missing link for one of the images there, but it’s put a couple of images in. Could be better, and we could go directly into the code now and start to improve it, or we can click and actually tell it to fix its own bugs. So Canvas is an interesting little mode.
Creating Files
I’m going to do another little bit of image generation here, and it is worth pointing out that in the paid version, you obviously get more images and it’s faster, so it really is pay-to-play at this point. “Create an image of a coffee on a table in a rustic cafe.” We’re going to generate this image. It’s going to do its little thing, its thinking process, because I’ve still got O3 mode turned on in the top there. It does a pretty decent job. ChatGPT image generator is much better than it was, let’s say, even six months ago.
I’m going to say, “Take that second image and create a downloadable PowerPoint. First slide, I want the title just ‘Coffee Shop Cafe’ or something similar. And on the second slide, I want the image.” What you’re going to see it do in a moment, a little bit like when we had Canvas mode running, is it’s actually going to write some code for itself, and it’s going to give me a downloadable PowerPoint file. You don’t really need to know what it’s doing if you’re not interested in code, but if you are, it’s using the Python programming language and a module called pptx. It can use that to write itself a little script, and when I download that, it gives me a PowerPoint. It’s pretty basic, but you can prompt around this. You can get a little bit more out of this. Just something that I thought I’d point out. It can also be used to create PowerPoints, PDFs, Excel spreadsheets, Word documents, and all sorts.
Interactive Applications
That coding feature, combined with Canvas that we saw earlier, can actually make some interesting little applications, simulations, little games, little self-contained websites. So if I say, “Write an app which has a simple game that I can play right here in ChatGPT that helps me to learn compound and complex sentences,” then it’s going to go into its thinking mode. It’s going to have a think about the best approach to take with that, and then it will write some code to solve that problem. Again, you can do some of this stuff in the free version, so it’s worth trying out and playing around with. You just get more uses of it in the paid version.
When it’s finished, you have a pretty simple game here where it wants me to choose the sentence type: simple, compound, complex. If I click the wrong button, I get “incorrect” with a correct answer. If I click the correct button, then it’ll add a point onto my score. Again, you know, if you prompt these things backwards and forwards, you can get different results.
Agent Mode
Last of all, I am going to demonstrate Agent Mode. It is in the Pro version, but it’s also coming into the Plus $20 a month version with limited use. This is its web browsing agent. I wasn’t able, while I was making this video, to get it to do anything particularly impressive. I’ve sent it off, as you can see in the prompt here, to go to my website, leonfurze.com, find my contact details, pop them onto a Word document, give me some summaries of my own blog posts. I will say that I’ve written a couple of pretty extensive blog posts about this, and again, I’ll send out the links to those, because in my experiments, I really haven’t been able to make agents do anything particularly successful.
But these companies—OpenAI, Microsoft, Google, even Amazon and Meta—they’re all working on their own versions of this technology. Essentially what it does is it takes a ChatGPT or a large language model AI, gives it a virtual computer, virtual screen, a virtual mouse, a virtual keyboard. It takes screenshots regularly so that it can see what’s going on on the screen, and it uses those screenshots and the coordinates to move the mouse around and click and navigate the internet like a human user.
You can see here, for example, that it’s managed to open up its little browser window. It’s found my website. It’s gone to the contact page, but it’s getting a little confused, because what it’s actually doing here is searching for my email address. It hasn’t really clocked that you can contact me on my website just directly through that form there. So then it starts to search around in other places. It’s quite interesting to watch how Agent navigates the internet. It’s not really like a human user. It’s gone to my About page. It’s trying to find all of that information there. Now it’s just doing a Google search to try to find my contact email. It’s sort of given up on that, and it’s going to the second part of the task. Now it’s reading through my articles. You can see here that it’s got a reader mode where it will just scan the text of websites.
So there’s a lot of features there. Eventually it finds my Practical AI Strategies website as well. It’s still desperately searching for my contact details and not managing to find that. But in the next six months or so, as this technology evolves, we’re definitely going to see more versions of OpenAI’s agents.
Summary
Just to wrap that up:
In the free version, you’re mostly using GPT-4o, which is the basic model. You can use “think longer” for a reasoning model which uses O3 but with limited use. You get about five uses per month of Deep Research. You get Canvas, image generation, Study Mode, and web search, all with slightly limited or restricted availability.
In the paid $20 a month version, you’ve got access to models 4o and O3. If you really want to go higher on that $200, you can use O3 Pro and 4o Mini, though there’s not really much of a difference, honestly. Deep Research increases to 30 or 40 a month, or up to about 400 per month on the highest tier. You get faster and also lengthier responses. They do seem to be more accurate, which I find concerning, as I mentioned earlier, and it is more capable at things like coding, more capable at some of those more advanced features.
Leave a Reply