EP 470: OpenAI’s Deep Research Updates - What’s new and how to make it work for you

This is the Everyday AI show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business and everyday life.

I think one of the biggest AI updates of the year so far just dropped, and hardly no one's going to notice because it's not a new model. I'm not talking Claude Sonnet 3.7. I'm not talking Grok 3. It's technically not even a new feature, but this is OpenAI's deep research updates that just dropped. It's exponentially more powerful. But the big news here is actually the access. Deep Research is now available to all paid users. So, yes, if you are on the $20 chat GPT plus plan, you now have access to what I think is probably the most useful AI tool that I've ever used, and I've used thousands of them. All right, so I'm excited to go over today, not just going over this access, right? And now there's hundreds of millions of people that now have access to this tool that maybe couldn't, you know, shell out the $200 a month for the pro plan, but also to look at some of the new features inside of Deep Research. All right, I'm excited for this one. I hope you are, too. If you're new here, welcome. Thank you for tuning in. My name is Jordan Wilson and you're listening to Everyday AI. This thing's for you. It's your daily live stream, podcast and free daily newsletter helping everyday people like you and me not just learn generative AI, but how we can leverage it to grow our companies, to grow our careers. And I want you to be the smartest person in your company at AI. That journey, well, it starts here, but then it continues, right? You can learn everything here but how you leverage it. It's on our website. That's your Everyday AI dot com. Sign up for our free daily newsletter. We're going to be recapping today's shows as well as keeping you up to date with everything else that you need to know. All right, before we jump into what's new at Deep Research, and there's a lot, let's first go over, as we do almost every single day, what's new in the world of AI news. So first, Meta is considering a $200 billion AI data center project in the US so according to the the information, Meta is planning a $200 billion data center campus for AI projects. Potential locations include Louisiana, Wyoming and Texas, with executives visiting sites this month. This would be one of the largest AI infrastructure investments yet, as Meta aims to expand its AI capabilities. So Meta CEO Mark Zuckerberg recently announced 65 billion in spending for AI infrastructure in 2024. This is obviously a lot more Competitors like Microsoft and Amazon are also ramping up AI data center investments with budgets of 80 billion and 75 billion respectively. Although it was reported that Microsoft might be scaling things back a little bit. And this comes just hours, literally after the Apple news that Apple was investing 500 billion in AI infrastructure. All right, let's go from big money to no money. So Microsoft Copilot is expanding its free access to advanced AI powered by OpenAI's O1 model. So Copilot is rowing rolling out free unlimited access to its Think Deeper feature, which is powered by OpenAI's O1 model, and it offers users advanced reasoning tools to handle complex decisions and tasks. So think deeper if you haven't used it. And it enables in depth analysis for decisions like comparing electric cars, planning career moves or evaluating home renovations, helping users make smarter data driven choices. So this new update also includes unlimited use of the Advanced Voice feature, the Copilot version of it, allowing hands free extended conversations for tasks like practicing languages, mock job interviews, or real time recipe guidance. I've actually used it for that many times. So Co Pilot Pro users maintain early access to experimental features and better performance during peak usage alongside integration with Microsoft 3 Microsoft 365 apps. So yeah, pretty crazy. Co Pilot just said, hey, we're going to give everyone unlimited access to advanced to their Voice mode and to their thinking deeper which uses 01. You heard that right, unlimited 01 usage for free. Because yeah, you don't even get unlimited 01 usage on the $20 a month chat GPT plus plan. So pretty crazy. All right, speaking of new things for free, one last piece of news. OpenAI has also introduced their free version of Advanced Voice mode. So the Advanced Voice mode in Chat GPT is much better than the Voice mode in Copilot, even though it's based off similar technology. ChatGPT is just much better. So OpenAI did say though it is going to the free version is going to be based off of GPT4O mini, a lighter cost effective model than the paid version of Advanced Voice Mode which is based off of GPT4O. So again the paid plan gets the Advanced Voice Mode based off GPT 4.0 and the free version gets the Mini mode. So free users will see a preview of Advanced Voice Mode displayed across the platforms. Though OpenAI has not specified exact usage caps. The strategy is seemed designed to attract new paid users while also maintaining server demand. Don't worry, if you're that $20 a month plus subscriber, you still retain full access to the GPT4O powered advanced voice mode. So they're not making it, you know, mini for everyone. They're just creating a essentially a light version of advanced voice mode for free users. All right, so we're going to have a lot more on those stories and a ton more in our daily newsletter. So if you haven't already signed up, make sure you go to your everyday AI.com all right? I'm excited for this one live stream audience. I hope you are too. You know what, I'm curious at least for our live stream audience. Are you using the free version of chat GPT? Are you using the 20amonth plus version? Or are you using the 200amonth pro version? Or maybe you're on an enterprise plan? You know, I'm always curious, just so everyone knows, I use them all right, I have free accounts, I have paid accounts, I have teams accounts, I have enterprise accounts because companies, you know, Fortune 500 companies hire us to teach their hundreds or thousands of employees how to use Chat GBT because this is all we do. And I also have Pro account, so I have every single tier. I use ChatGPT every day. And I wanted to put that out there to talk about Deep Research because I've been using it since it came out, since the day it was released February 2nd. So like I said, up until today, Deep Research has only been available if you've had that $200 a month pro plan. So very few people have been able to actually take advantage of Deep research. So here is how open AI describes Deep Research. They say it is an agent that uses reasoning to synthesize large amounts of online information and completes multi step research tasks for you. Are you still running in circles trying to figure out how to actually grow your business with AI? Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on Genai. Hey, this is Jordan Wilson, host of this very podcast. Companies like Adobe, Microsoft and Nvidia have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead. And some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for ChatGPT training for thousands or just need Help building your front end AI strategy. You can partner with us too. Just like some of the biggest companies in the world do. Go to your everydayai.com partner to get in contact with our team or you can, you can just click on the partner section of our website. We'll help you stop running in those AI circles and help get your team ahead and build a straight path to ROI on GEni. It's good. Let me just say that, right? Actually first let me kind of set the stage and then I'm going to jump straight forward into Deep Research. All right, so like I said, right now the limits have also changed. So previously if you were on the Chat GPT Pro account, you had 100 deep research queries a month. Now that's been upped to 120. And now starting today, if you are on any other paid plan, so ChatGPT plus Teams, Enterprise or EDU, you get 10 deep research searches a month. So it's not a lot. But follow along with what I do and it's going to make it worth it and soon. CEO Sam Altman OpenAI CEO Sam Altman has said that free users will get two a month. That has not rolled out yet. But plus Teams Enterprise and edu you now have access to by far. It's technically even the most powerful model in the world because as far As I know, O3, what this is based off of has not been benchmarked. So we've seen O3 mini and O3 mini high benchmarked from open AI and this is their advanced reasoning model that came after 01. Right. So we went 01, 01 Pro, skipped O2 because of some copyright reasons and now we're at O3 mini and O3. The actual O3 full model is not available except in Deep Research. So it is a fine tuned version of O3. It is mind bogglingly good. So we're going to do this one a little different. Live stream audience, I need your help. Please let me know if you can see my screen. Also, I'm going to be bouncing between a couple, a couple windows here, so hopefully this works out okay. Nothing more I love than doing live demos. That's a bit of humor. I'm actually not the biggest fan, but you all reach out to me all the time and say the live demos are super helpful. So I'm gonna do it. All right, so here's what we're gonna do. I'm gonna get this started. I'm not going to read the full prompt to you yet because it's probably going to take a while. All right, but here's essentially the prompt that I'm putting in in podcast audience. I've opened my Chat GPT account, I'm on my Pro account, but it doesn't matter. I could be on my, you know, any of my other paid accounts. All right, I've pasted in this prompt. So one thing that might be a little confusing, and I actually wish that OpenAI would clear this up. So as an example, you have your different modes, right? So you know, you have your GPT4O. GPT4O with TAS0103 mini, oh, three mini high 01 pro, etc, right? You can click Deep Research. It's actually a button that is right under where you put your text in, okay? Which is a little confusing because when you click the Deep Research button, it still shows up as whatever model you're using. So as an example, I can be an 01 Pro and I can click that Deep Research. That doesn't mean it's using 01 Pro. So it's a little, it's a little confusing. But essentially, once you click that Deep Research button, it should be down there by the search button. So look for the little telescope, right? And if you don't have it yet, you know, try logging out of your account, logging back in, and it should pop up there. But essentially, once you click that Deep Research icon, it's going to override whatever, you know, model or mode that you chose in the model selector. So it's the. This is the only instance that does this inside of Chat gbt. It's a little confusing, if I'm being honest. All right, so here's what I'm going to do in this long prompt and I'm going to show you some of the things that I'm doing. Okay, so I've exported a ton of data here out of Google Search Console. Let's see how many. Okay, so I have about 6,000 cells here of data in spreadsheet one, and then I have a spreadsheet two here. Let's see how many rows of data. It's probably roughly the same. Let's see 8, 000 rows. So I have just under 15, 000 rows of data. So all this data is. Well, it's. It's a lot. So this is our, for our website, your everydayai.com. this is our Google Search Council data. So I've been doing SEO on and off. I mean, technically, I've been doing it for like 25 years. I've been getting paid to do SEO in some way, shape or form for more than 15 years. So I'm a dork. I spent a lot of time in Google Search Council and this is one of the reasons why I want to do this in deep research. So what I want you to think of as you're staring at all of these numbers on the screen, if you're joining us live, think about tasks that you do all day that require a lot of research. For me, I do a lot of research on what type of content that I'm going to cover, right? So all of our podcasts, right, we put the transcripts on our website, we put the YouTube video, we actually put the podcast itself. So you can go to our website, literally it's for free. You can go listen to 500 episodes, read our write ups, look at the transcripts, watch the video, it's all on the website. So essentially, you know, we hope to attract a lot of search traffic by putting all of this information for free on the website. But I do spend a lot of time researching new content types that we should be covering based on a lot of things. So I still do a lot of old school manual research. Well I did before deep research. So I say all that because I'm trying to invite you into my world. I'm always using some advanced Boolean searches to find different trends and all of these things of things that we should be cover things we should be covering, more things you all care about. You know, we take requests in our newsletter, we do polls say like, hey, what do you want me, like what do you want us to cover? But I still do a lot of research. So this is an actual use case that I am refining to hopefully use a lot. So it's, it does a couple of things. It's trying to improve areas where we were getting a lot of traffic before, but maybe we don't anymore, right? So maybe we used to get, you know, I don't know, a hundred people coming in from mid journey and we don't need more, right? It would take me a lot of time to go through this data and to start piecing all these things together because we have thousands of web pages and we technically rank inside of Google search for thousands of words. So let me just give you one, one example, right? So something that brings in a decent amount of traffic to our website is the term Chat GPT free, right? So usually about three times a year I do a podcast that is just chat GPT free versus chat GPT paid. I do it for a couple of reasons One, it brings a lot of users to our website and they find our helpful content. But number two, Chat GPT changes what's free all the time, right? It was just one of our news stories. So this content often becomes stale. So think of that. If you have thousands of words that bring in organic traffic to your website, it's, it's changing all the time. Especially in something as fast paced as AI. This changes all the time. All right, so essentially I have these two different spreadsheets. I'm going to be exporting them and then I import them into deep research and I'll read this long, this long prompt that I'm putting in here, but I want to get it started so I have this prompt. I'm going to upload these two files, all right, And I'm gonna explain this. But the first thing that Chat GPT is going to do is it's going to ask me questions, right? Which I love. So without getting into all of it right now, I'm essentially having Chat GBT spot trends in my data and then go do a bunch of research, right? And come back and report to me. So one thing I like about open AI or you know, chat GPT's deep research before it gets started because sometimes it can take anywhere from five to 30 minutes. It asks you questions, clarifying questions. So I need to answer those. So I'm going to go ahead and answer them live. It's going to be shorter answers than I would normally do because I don't like typing live. So number one, it says SEO goals. Besides improving rankings and traffics, do you have specific conversion goals? So I'm going to say more traffic and newsletter signups. Okay, then number two, it says competitive landscape. Are there direct any direct competitors you want me to consider when analyzing ranking potential? No. 3, content focus. Are you prioritizing any particular AI related subtopics? For example, business AI applications, generative AI, AI tools, ethical AI. I'm going to say pretty much anything related to AI gen, AI, AI tools, LLMs, etc. All right then number four, content types. Are you focused on long form blog posts, pillar page guides, videos or other? I'm going to say all focused on all of those things. All right, and then number five, it says keyword difficulty tolerance. Are you open to targeting high difficulty keywords with long term potential or do you prefer lower competition, quick wins? So I'm going to say a mixture of all. Okay, so what I want to do now before I get back into kind of the rest of the show, I want to make sure that Deep Research is starting. So then we can check back in on it in about 10, about 10 or 15 minutes. Hopefully it should be done. So you'll see here it's giving me a positive message that says starting Research. I always like to wait until I actually see this bar start and then you can click on this bar and this is great. And you can actually follow Deep Research step by step activity to see how it researches. And then you can also go and click on the Sources tab. So right now it's just compiling sources. I don't see much activity yet, but it does look like it's about to kick off because we have a heading that says SEO Analysis and Future Content Strategy for Everyday AI. All right, so I think we are good. All right, so let's jump back in and go over a little bit more about Deep Research in general. And let me answer the question that I don't know if any of you are thinking, but I'm just going to go ahead and answer it. Open AI's version of deep research is the 90s bulls unfreaking, stoppable, right? Yeah. There's some other teams, there's some other versions of Deep Research that have their pros. Right? But open AI's version of deep Research is the pro, period. Right. So I've covered them all. And like I said, I have been using OpenAI's deep research since the day it came out. I've used it almost every single day since it's came out, even on the weekends. It is very hard for me to do much of anything, if I'm being honest. You can imagine, as someone that covers AI every single day, the amount of research that I do, the amount of time that Deep Research has given me has given me back, I cannot even begin to compute it. I would need literal GPU compute to compute how much time this has saved me. But I've covered all of them. I've given each and every one of these except some of the newer ones. But I've given the main Deep Research tools their own dedicated show. So even if you want to listen to, you can go listen to episode 425. Google's deep research was the first Deep Research tool came out in December. You can go listen to 454 when I did a deep dive on OpenAI's deep research. All right, 464, I covered perplexity deep Research. And then the next day I did a Deep research throwdown. Perplexity versus Google versus OpenAI. Now I'm sure A lot of you are going to be like, yo, what about Grok? Grok is amazing. Grok has a new deep search grock. No, don't use it. No, I mean if you use it, use it with like a, like a grain of salt that's bigger than this pot can that I'm holding on screen. Here's why. Grok Deep Search is largely based off Twitter data. Yes, it browses the Internet, but it also browses Twitter, which is bad, right? Studies have shown that Twitter is the number one in terms of misinformation and disinformation amongst all social media networks. So I'm personally, and I would advise you again, huge grain of salt before you jump on board and you're like, oh, grab Grok Deep search, right? I wouldn't. It's almost like searching Google in reverse order, right? Page one is usually the most highly authoritative pages, right? If you're searching for something manually on the Internet and then when you're like page 30, you're like, where am I? That's what it's like. I think if you're using X or Twitter posts to search something for business, I don't know, unless you're trying to search something pop culture or a recent news event, I would not use Grox Deep search for anything, if I'm being honest. But what I'm trying to say here is Open AIs is in a league of its own. All right? There's certain things, Google, Gemini, Deep Research does great. There's certain things, perplexities. Deep Research does great, but it makes up so much. OpenAI's deep research, I think is the AI tool of the year. It's early 2025 so far, but the amount of time that it can save the average everyday person, mind boggling. It really is, right? I think so many of us either, number one, don't realize how much time we do spend researching something, or number two, we really only scrape the surface because it is so time consuming to research things at the depth that we may want to. So instead of doing a level research on a project that you need or on a, on, you know, on a, on an assignment that you're completing and you're spending a, you know, couple hours on, you just might not have the time to weed through 20, 30, 40, 50 different websites. Make smart decisions between those searches, right? And, and you know, think of it like I think of deep research, OpenAI's version, like a fork down a road, right? You go up and you meet a fork in the road. But instead of it forking into two, it's forking into four. Right. And that is when OpenAI's model O3 model makes a decision for you. Right. So it doesn't just take this wide, you know, cast a wide net, kind of like the Google Deep Research does. It makes very smart, informed decisions. It visits one page, it makes a smart decision and goes to another page. And sometimes somewhere along that tree, it will go in a different direction that maybe you or it thought it would originally go because it acts like a smart human researcher. It's fantastic. All right, let's keep this going and talk about what else is new. So, number one, the biggest new update here that you need to keep in mind, I already said it's out to all paid users, so there's going to be, I believe, hundreds of millions or at least tens of millions of paid users that are going to get access to Deep Research. Right now, that's number one. Number two is now it can embed images with citations in the input. All right, that may or may not be helpful for you. If you're a visual person or if there's visuals that are highly relevant to what you're trying to research on, I'm sure that could be helpful. But the thing that I'm interested in and that I showed in this demo is it is now better at understanding and referencing uploaded files. That's huge. That is a, a huge update that I think is going under the radar and that's why I did this little demo that I did. Also open AI did release a Deep Research system card before we talk about some of the things that are in that system card and some of this new information that we got. Right. So that's more or less a system card is a technical report essentially that research labs put out on different models. All right, so we do have that. But first I like to explain the difference between some of these new agents that OpenAI has released and agent esque behavior. All right, so OpenAI has released tasks. So that is when you can schedule open schedule, chat, GBT, specifically GPT4O to go do anything semi autonomously. Right? You schedule it, but then it does it. So you could have it do, do it every day, have it go do something at 9am you can't schedule Deep Research, you can't schedule Operator, but in tasks you can go have GPT4O do anything that you want. Operator is an agent. Okay? Operator will use an actual virtual computer. So that is a little different. Deep Research is not using a virtual computer per se, and you can't have it, you know, go log on to your, you know, your email or your LinkedIn account or something like that. Like you can with Operator, you can have Operator, you can take over the screen, you can log into certain things with your credentials, have it perform tasks for you. Deep research isn't like that. Deep research is literally something that's just going to do a bunch of research for you. All right, so a little bit more on how deep research functions. So and a lot of this is from the model card. Some of this was already known. So it is based on O3. This is an early version of OpenAI's full O3 model, but it has been fine tuned for web browsing capabilities. Multi source analysis is the big thing. So it searches across the Internet to gather, interpret and analyze text, images and PDFs establishing connections between different information sources. Also it has Python integration so the system can write and execute Python code to perform calculations, analyze data sets and generate data visualizations. That's huge because that's what I'm going to be showing you. Also there's an advanced reasoning framework. So I talked about how in between each essentially web page it visits, it reasons, right, this is a reasoning model. This is not a transformer model. So it literally goes to a web page, it analyzes it and it makes it a decision like hey, where am I going to branch my research off next? So it essentially continues to accumulate knowledge and depending on how good you are at prompting it, this is when prompting actually really matters because it does have that advanced reasoning framework. Here's something that's pretty interesting that we just found out in the model scorecard. So it actually has a dual model architecture. So deep research utilizes a secondary O3 mini model to summarize complex chains of thought, making extensive reasoning processes more accessible to users. So yeah, it has a fine tuned version of O3 full and then it has a secondary O3 mini model. So there's technically two different O3 models working on this Deep research. It's wild. And then it is specialized training. So it is trained on purpose built browsing data sets with reinforcement learning where responses were graded by chain of thought models against ground truth answers. That's wild. It was trained against a data set of browsing and using reinforcement learning, but synthetic reinforcement learning, right? This is in the system. Like I'm like, I'm reading this in the system card and I'm like this is so meta. All right, I know it's open, open AI, but this is so meta. All right, so let's that's all the details. That's the technical details. So now let's jump back in live. We'll see how far along we are. All right, let me first read this prompt, all right, because it's. It's kind of big and I want you to understand what's happening here. So remember the other big thing that's new here, and this is why I wanted to highlight this is it has Deep Research, has an improved ability to interpret, understand, and work with advanced data sets or files that you upload, y'. All. I uploaded, what was it, 15,000 cells of data. Some pretty complex data, too. So this is kind of my prompt to Deep Research. So I'm giving it a little context, I'm giving it some instructions, and I'm asking for a very specific output. All right, ready? I'm going to try to make this one go fast. So. All right, here's what I said. This is my Google Search Council data for queries and pages for my website. This is a year of data. The last six months columns are the most current six months, and the previous six months column is the six months before that. Also make sure to look at the difference column, as that is where certain keywords are trending. This is from my website, everyday AI found@your everydayai.com in the queries that are bringing us organic traffic via Google, this query data shows comparative trends for clicks, impressions, CTR and average position. Also, I've added external keyword data that shows average monthly search, volume, CPC and competition. We obviously want to focus on SEO optimization and improved content strategy for terms with very high search volumes that we have a shot at ranking for. I've also uploaded similar data for pages, but that data does not have monthly search, volume, CPC or competition. In short, a positive click difference shows that we're getting more clicks for that page or query than we were in the previous six months. First, analyze the data to see, number one, where I'm gaining the most visibility on queries. Number two, where I'm losing the most visibility on queries. Number three, where I'm gaining the most visibility. Oh, that should say pages. I made a mistake. Let's see if operator will even know. I should have said pages there. I'm a human, all right? So I'm essentially asking for wins and losses on queries. And I was trying to say wins and losses on pages, but I got that wrong and I just repeated it. So we'll see if for number three and four, if it found. If it figured out what I was trying to do. Or if it just repeated it then number five I said the 15 biggest SEO opportunities based on the above data. Be extremely specific and give me detailed logic behind your choice without wasting words. This should be either extremely accurate SEO optimization trends or content strategy gaps and opportunities based on high value opportunities. 6 Future Content Plan below based on deep research findings for number six Future Content Plan, please use deep research to find trending topics AI related topics in 2025 that may be aligned with my current content or content gaps based on your analysis of my files. Focus on huge areas of opportunities that I may be missing in our SEO content coverage. Make sure these are timely and areas that I'm likely able to rank for if I create compelling content based on trends and opportunities. You should tell me if I should create new content that doesn't currently exist or update current content that is already ranking but may be losing traction in the content plan. Please Give me either 10 new pages, posts or updated pillar pages that can bring in the most amount of organic traffic in the near future. Each one should be pinpoint specific and based on specific trends and research. As an example, AI for business automation would be too general or vague and a bad example instead. A better example would be AI powered deep research tools. How they'll Change Business automation. Yeah, now we're getting meta. Also, each of these 10 pages or posts should not only be pinpoint specific but they should be extremely relevant to the research and trends from 2025. Also, you should outline in bullet point 5 to 10 subtopics or statements of facts to cover for each of these. Oh, I made another mistake. I should say each of these 10 pages or posts. It should say 10. I put 20 already two mistakes that I did. None of these should be generic or vague in any way. They should be factually relevant, timely and provide Context to the 10 page post ideas. Be exhaustive in your research. Please act as a seasoned content strategist, making sure to research in in both obvious needed and yet unexpected places. Huh. That was a lot. All right, so had to take a quick drink after that one, y'. All. You guys following what what I'm asking it to do here? This is I would have to hire an SEO agency to do this and they would not do this good of a job. I don't Even know what OpenAI deep research is going to spit out here, but I'm already guessing it would do a better job than most humans would do. But I could be wrong. So let's go ahead and explore a little bit and it looks like we're almost done. This kind of bar here, it says 15 sources. So let's kind of see what we can do. So it's compiling sources. So right here it has my files that I uploaded. It looks like it went to Forbes. All right, And I can click that later. It should give me the exact page that it went to right now, which is giving me the domain name. Then it went to Sloan Review, Dot, mit, it went to News Microsoft, it went to Exploding Topics. That's great. That's a great website to look at. It just looks at different trends in different industries. So I probably should have, in all honesty, prompted it to go to Exploding Topics. But that just knows that Deep Research is super smart. It knows to go to Exploding Topics. It also went to Sapphire Ventures, it went to Gartner, and I'm sure it's going to go to a lot of other places. So I'm not going to read off this entire kind of summarized chain of thought. But in the same way, if you've used open AI's Reasoner models, right? So you have your Transformer family of models, you know, your GPT4, GPT4 Turbo GPT4O, right? GPT4O mini, then you have your Reasoner models, your 01 Pro, your 01, your O3 mini, O3 mini high, right? Alphabet soup, all these names that are not marketing, right? So in your reasoning models, you can usually see a summarized chain of thought. So the same thing in Deep Research if you go on the right hand side panel. All right, I would also just recommend always, anytime you're using Deep Research, don't have your sidebar open because it just gets super clunky. So just have your kind of middle main area open and then look over into the activity. And if you don't see it right, all that means is look at the bottom and where that progress bar is, that's where you're going to want to click. And when it's done, you can still click there. Don't worry. Okay, so when you click there, then it gives you back that activity and sources bar. So let's just quickly look and hopefully we can see a little bit. So before it even gets started, it says, I'm examining everyday AI data to uncover key insights, analyzing shifts in queries and Pages, uncovering top SEO changes, chances, and pinpointing AI trends for 2025. All right, so it says, I suspect time might be causing browser search issue with user files. So it looks like there was some issues with the file and it looks like it tried to use a Different process to look at those files. So we'll see if it actually did. I did run this search once already. I went through, changed some things, improve the search for this second one. So hopefully it works here. If it doesn't, I will fall back to the one that I just ran like an hour or two ago. Although I did want to do this whole thing live, but it did look like I might be running into an error. That's fine. Full disclosure, I've probably run at least 80 to 90 deep researches so far on OpenAI. I've only had one error out so far. So just by math that's like a one point something error rate. Go figure if my second error ever would be on one of these instances that I'm doing live. But such as Lightfoot. All right, so it says I'm weighing the pros and cons of opening the everyday AI page trends and query Trends CSVs files directly with Python or through a browser, considering the potential for better content display in the browser. So interesting. It looks like for whatever reason it is having some issues opening up the files. Like I said, I did do a very similar version of this query previously and it did fine. All right, so it looks like hopefully it did read it. So again I'm reading through kind of the activity summary here. So it looks like, it looks like it did read it. All right, so it read one file, it read the second one. Then it says I'm analyzing keyword trends and their impact on rankings, impressions and click through rates. Then it says, okay, so I know it did read the files because it says I'm digging into a certain term, right? So the term chat GPT free has 2.7 million monthly search volumes, right? That's huge. And it's also showing an impression difference from a six month period to another six month period. So remember, I essentially uploaded one year of data to deep research here. It's the last six months versus the previous six months, right? And there's thousands of terms and there's also this data. So you can see it's comparing, it's comparing the data. So you know, one example, let's, let's just see if I can get an actual, actual good example here for y', all just so you can kind of, kind of see what we're working with. Okay, the, I think that the, this one's, this one is a good example. Just the Chat GPT free, right? So just as an example, in the last six months there's been a thousand 62 clicks for that exact term chat GPT free. Right. And there's like 50 different terms that sound just like that, Right? So chat space, GPT free. Right. People type in Chad GPT and chat GBT. Right? So it's just the term chat GBT, one word, free. I had 1062 clicks, and then the previous six months there was 258 clicks. So, you know, it's looking at the difference. So, you know, as a. As an example, if a human were going to. To look at this, they would say, oh, okay, well, we probably did better, or you were receiving more traffic for chat GPT free. Chat GPT free. It could be because more people are searching for it in the last six months versus the previous six months. And then you would. To confirm that, you would probably look at the position and the impression. So, you know, as an example, the average position was 15.6. So in the middle of page 2 for the previous six months, and for the most recent six months, the position was 11. So essentially for that term, our website moved up to the. The cusp of, you know, page one and page two, which is a pretty good place to be for a term that has, you know, 3 million people searching for it every single month. So you can see how something like that would be extremely valuable to someone like me. Right. If you can show up on page one consistently and people land on the page and they'll listen to the podcast or, you know, sign up for the newsletter, maybe that's exactly how you found us. I don't know. All right, so let's see. Hopefully we're. Hopefully we're about done here. All right, we might. We might have stalled out, folks. We might have stalled out. Let's see. You can always click refresh. Oh, good. We didn't stall out. I just had to literally click refresh. It was done. Yeah. Sometimes the progress bar will fully update. Sometimes the progress bar just stalls. Okay. But what you can do is go down here. Okay, not bad. So this. You might miss it after your query is done. In deep research, this can be easy to miss. You essentially need to go find this thing that says research completed. I really wish there was a little bit better user interface to know, like, oh, I should be clicking this. All right, but I can click that. It says it took 18 minutes and it went to 15 sources. All right, so I can click on it and then I can see here the activity. I was starting to read through it. So essentially, after it went through those two CSVs, it started to identify some trends. Then from there, it started searching for AI trends. 2025 predictions. And then it went to an MIT Sloan website, a Forbes for AI trends website. IT Pro Today Microsoft News. It went into Forbes and then here's, here's where this is pretty cool. So it said, I'm noting a potential shift to Sloan MIT due to Forbes being locked. Right, Like a normal human. It went to a website. It went to a Forbes website to try to do some research based on the data and the trends that it spotted. In the 15000 rows of data that I gave it, it went to a Forbes article. The Forbes article was locked. So it looked for a free, a free way to essentially read that article. And it looks like it found it on Sloane Wire MIT Edu, right? So then it said, we're examining the Sloan's article on business trends, focusing on agentic AI and measuring generative AI return on investment. Simplifying sections and cross checking with a Microsoft news summary might help streamline the process. Right? So they're like, yo, we found some good information here. It was originally on a Forbes article, then it was on this Sloan review. Let's cross check it with something very reputable to make sure that something that it found was actually a trend and not just a one off thing. Right? Because it wants to do literal deep research and you know, if you give it a lazy prompt, it's not going to do this good of a job, right? The thing, even with, I don't care what anyone says, Even with the O1 models, O3 deep research, the better your prompt, the more data you share, right? Even when I was answering those questions, I should have gone into much more detail. I just wanted to do it live and I hate typing live. But you should go into extreme detail and take care of when doing this. Especially if you're a chat GPT plus user. You only have 10 of these a month. You have to make them count. All right? So I'm not going to go through and read every single line. You know, if you want, if you're, if you're listening, just say activity. You know, just send the word activity. If you're listening on LinkedIn or Twitter, I'll just copy and paste this because I don't want to turn this into a three hour podcast. So this is, this is great though. So it's even looking at like, you know, rumors and trends, right? So it sees that, you know, GPT5, it went to explodingtopics.com and it saw that GPT5 was a huge explosion. But then it just started to do some research to see if it was, you know, actual. If it was a trend, if it was rumors, what it was. This is, I mean this is great. Looking through the activity, this is where OpenAI's deep research shines because the others don't really do this good of a job, I think in the decision making process between sources. Yes, Perplexity and Grok and Google go to way more sources. Open AI makes it count. All right, but let's quickly look at the report that it put together because I already know the right answers for many of these. All right, so let's see what it found. So it said chat GBT massive impression surge. Gained about a quarter million impressions over the last six months. Correct click screw, correct AI podcast breakthrough to page one. Yes, that's been a, a huge, a huge boon. Something I've been personally investing in for about 18 months to show up, you know, as a top three result when someone searches AI podcast on Google. Not easy, but we're finally there. So operator did that. So why, I'm like, why am I spending so much time going through something that was a bunch of spreadsheets? Because this is one of the new updates inside Deep Research. Aside from the 20amonth chat GBT plus access, now it has a better ability to work with files, to understand them, to interpret them and to do deep research based on them. And I said, what better way than yo, here's like 15,000 pieces of data, here's what they all mean. Go do a bunch of research, right? So as I go through here, it always has citations. So the citations in this case early on is the actual file that I uploaded. All right, so it's, it's doing, yeah, it's, it's finding all of these essentially notable queries, etc. So that was the biggest. What was that? That was the biggest gains, right? Then it has the biggest loser. So the biggest losers for our website were the terms knowledge cutoff, Taplio, how to make chat GPT sound more human. Free chat GPT with a space. That's smart. It noticed that, hey, actually free chat GPT no space was a big winner. Free chat GPT with a space, big loser. But it says this is smart. It's essentially being like, yo, don't worry because the net traffic gain from free chat GPT one word is still positive. So it's super smart. Super smart. You know, let's, let's scroll down here. Let's see if it actually found my mistake because again, I accidentally told it to do the queries. Wins and losses twice. I meant to say pages so we'll see if it actually picked. Oh, it did. My gosh. Deep research is good. I literally told it to find winners and queries and losers and queries twice. It realized that I made a mistake and it went, did the right thing. Right when everyone says, oh, AIs. AI is never going to be as smart as humans. I'm a human. I made multiple, multiple mistakes. Deep research is like, yeah, you made a mistake. Here, here's, here's what you meant to do. Thank you, Deep research. All right, so top pages it went through, it found my biggest gains, right? The home page free Chat GPT versus Chat GPT plus, what's the difference? Apparently the cling AI video review that we did is now getting, I don't know, 100, 000 impressions, something like that. Category page for AI news, AI's impact on society article and a couple other things. Notebook LM other big gains. All right, let's go down to the good stuff. Here's our top page. Performance biggest losers. This is what I wanted. Fifteen data driven SEO opportunities. So it says, number one, leverage Chat GBT free versus paid dominance. Keep the content fresh. That's good. It's telling me make sure to keep that content fresh. It says boost that page the site. Okay, so it's essentially saying that that page is hovering around position 11 to 17. And it's saying cracking page one here is challenging. But even moving from 11 to 8 could double the CTR opportunity. Create a concise guide or a tool page using Chat GPT for free that targets the broad term specifically. Great advice. All right, optimize for AI newsletter and AI podcast number four. Create a best AI newsletters list posts. And it's not a bad idea. A lot of people do that, you know. Could work. Let's see. Improve the click through rate for chat GBT plus queries. Revive the chat GPT Sounds human. That's a good one. Right? Good. A good piece of advice there. So it says the query how to make chat GPT sound more human clearly has user interest. It's essentially saying like, yo, you used to get a ton of clicks and impression for this. It fell off. You should probably update the tone in the article. Some great stuff here. What's interesting is I was, I was gonna see if chat GPT in deep research was smart enough to go visit these actual pages if I prompted it to. I'm sure it may. I specifically did not because I wanted to see if it would recognize if it should or not. It didn't. Or at least it didn't list it in the sources. So, you know, I guess, I don't know, maybe that's when we've reached AGI, is when you know, it does all these tasks that you kind of secretly hope it would without telling it to. All right, let's just go down to the good stuff. So it went through. It gave me all my 15 very pinpoint specific SEO optimizations. Here's the thing that I really wanted to see, how well it did the actual research, right? It went through because people I don't think are using this deep research to its full advantage. They're using it as is like, oh, you know, really good Google. No, no. These new updates are extremely exciting because it can work better with your files. And when it picks up on nuances in kind of trends in files that you upload, it will go off on its own in research. So let's see if it actually did that. So again, everything that I just did so far, more or less, aside from reading the activity of deep research, everything else has just been kind of general. Okay, here's a large language model doing cool stuff. But now let's see how well it took that information, how well it pulled insights, pulled trends, pulled opportunities out of all of this data based on the pretty advanced query that I gave it. And did it actually go out and do deep research? And I asked a ton of it. Let's see. All right, I'm liking this so far. All right, so remember I said give me 10 timely content ideas. So presumably things number one, I have not covered. Number two, it went out and researched these things. So it knows that there's appetite, it knows that there's search volume and it knows it's something either I haven't covered very well or I haven't really covered at all. So this is good. The rise of autonomous AI agents in 2025. Number one, it gave me a why, right? I'm not going to read this for all of them. Let me go ahead and just do the first one. So number so you can see the quality that it gave me. So number one, it says the rise of autonomous AI agents in 2025. It says to write a new long form article why AI agents that can be, that can act autonomously are a hot Trend. Heading into 2025, both experts and big tech are touting agentic AI as a game changer. And then it sources two different articles. It says everyday AI can demystify this for readers. And then it gives me subtopics to cover. Oh, this is good. This is Good, right? I could literally use this as an outline for my next podcast. And part of the research is already done for me. So here's the subtopics I should cover. What are AI agents? I need to divine. Define agentic AI, give examples. Next sub point, hype versus reality. Discuss why agents are trending. Next sub point, real use cases emerging. Talk about real use cases. And it gave me some. And links to talk about those gave me survey stats. A UiPath survey showed that 68 of I T leaders plan to invest in agents within six months. And then I can go click that and read it. This is really good. Agents at work, co pilots and more. Explain how tools like Microsoft 365 copilot act as agents executing tasks, summarizing email, scheduling meetings with minimal oversight. Then it gave me a link to that Microsoft article on six AI trends. This is good. It says possibly include everyday AI podcast insights if you've discussed agents. Then it says next sub point, risk and challenges. A balanced look at why some are skeptical. How to get started with AI agents. The road ahead conclude with an expert outlook. 2025 is likely the year agents move from hype to initial real world adoption. And then it has different articles that it cited from Sapphire Adventures. Then it says, predict that year by year end we might see at least one major success story from AI agents, such as claiming a billion dollars in savings. What does that mean for everyday users? Summarize the potential. Then at the end it says by covering agentic AI in an accessible way. This piece aligns with Everyday AI's mission to explain cutting edge, edge trends for a general audience. It should be published as new content, then updated as agents evolved. This is so good, y'. All. This is so good. I'm probably, I mean I'm probably going to do that exact episode with those exact subtopics right now what I'm going to do is I'm going to take this, I'm going to take this outline here. Then what I'm going to do is I have, I've done about three or four, maybe five or six podcasts so far on agents. So I'm going to get all those transcripts, I'm going to upload those transcripts into deep research. I'm going to copy and paste this. I'm going to do a little bit of back and forth and then I'm going to have it go playing my show, right? Fantastic. Fantastic way for me to, you know, people are always like, jordan, what's, like, how big's your team? Right? You know Very lucky that, that you all listen to this podcast when I ramble on so long and, but people are always like, you know, how big is your team? Right? You must have, you know, 50, you know, 50 people. Right? You know, I'm, I'm very lucky, you know, that we're like a top 10 tech podcast on Spotify and all the other tech podcasts. For the most part. They have huge teams. They have big companies backing them with a ton of money. Not us, we're a very small team, self funded. Right? This is how we do it because we're, we use AI tools to help us work smarter, period. So that's a great first example. Let me just go through the, the other ones here. Quick. Two says AI personal assistance, your AI companion for everyday life. Three, and also let me know, live stream audience, which one of these do you want to see most? I'll maybe you guys vote on which of these topics and we'll do this one next. 3 GPT 5 and beyond. What to expect from the next generation of AI models. New content. Forward looking. All right, there we go. Let's see. Number four, AI regulation is here. What new AI laws mean for you haven't covered that a lot. This did a really good job. Number five, measuring AI's impact. Are we really more productive? It. I would love to do an episode on that. All right, number six, AI and education in 2025. Personalized learning in action. So it's essentially saying I have, I had an AI and education post that was just tanking. So it's giving me new, new idea for new content to put on that page that's a little more relevant. Great idea. Let's see in all of these, y', all, it gave me 10 sub points for each and every one cited and sourced. My goodness. All right, let's keep going. Seven, AI and marketing and sales. 2025 playbook for professionals. So yeah, it looks like I had another page that was losing traffic that I didn't even know about. And then it's like, oh, here you go. Here's, you know, a good start on a podcast outline. Number eight, the new AI search experience. How to navigate AI powered search engines. Let's see. Nine, AI for creatives. How AI is changing art, design and music in 2025. I haven't covered a lot of art, design or music in 2025. My gosh, deep research did a great job looking at that huge spreadsheet with all that data and then went out and did a bunch of deep research. And then number 10, AI for everyone. How AI is becoming more accessible. That's great. I haven't covered accessibility a ton. I think I've only done like two. Two episodes out of like nearly 500 on accessibility. All right. My gosh, I'm exhausted. I thought this was going to be a quick episode. Granted, the deep research took a little longer than I would have liked, but my gosh, y', all, how impressive is that? Now I want you to think about what you do right now and what you wish you could do. All right? Yeah. You can go try. Don't get me wrong, there's some good things about Google Deep Research. There's some good things about Perplexity Deep Research. There's. I don't know if. If you love Twitter, maybe you can find some. Some utility in Grox Deep Search. I personally can't. This changes work. It does. I gave you an example of how this is changing my work. I invited you into my position, into my shoes, right? Kind of, you know, giving you some secrets on how we can plan great episodes. Right. I've been doing these exact things, just in other ways, using other AI tools, using, you know, different AI search, right? Stringing you a lot of tools together. This just makes it so much easier. But I want you to think about what you do, the time you waste searching, right? I've. I've talked about it way too many times about how using the Internet is this terrible black hole. It's too hard to use the Internet. The Internet is made to distract you. Think of how much time you waste on the Internet, right? All of a sudden, whoops, you're on social media. Whoops, you're on, I don't know, espn, right? I'm calling myself out there, right? Using deep research, obviously there's huge gains right off the bat by just being able to research way more, way better. But now take some of these new features that they just announced. So being able to have images in your responses, super helpful. But I think the biggest thing is just deep researches. Increased ability to better use and understand files, right? So think of my example. You have to have a mountain of data that you don't want to go through. Find trends and go research those trends. I'm guessing at least half of the audience, you probably have access to some data in that data. There's probably trends in those trends. There's probably research that should be done, period. Go do what I just did, right? If you are a chat GPT plus user, you have 10 queries a month. Go use them wisely. Just use my example. Apply it to exactly what you do. Or if you want to start off with something simpler. Something simpler. Research industry trends, but go into great depth, you know, first, type out everything you know, everything you want to know, questions you have, right? Use it as an actual thought partner. One of the biggest mistakes that I think people make is taking demos at face value, not trying to be mean here. I don't think most tech companies. No, let me just say it how I want to say it. Tech companies stink at doing demos, right? As an example, when Open AI did a demo of Operator, right? Their, their agentic AI, they had it like, buy tickets, which I thought was a terrible idea for a demo, right? I think the last thing people want to do is click and, you know, click approve 20 times for an agent to buy a ticket. That's a terrible idea for a demo for Operator, right? The same thing. I think what most people have talked about in, in what OpenAI demoed for deep Research was more just like doing like a super advanced, super personalized Google search, right? Like, oh, I need to buy some skis. I'm going on a trip. Here's the type of skis I want. Okay, well, a, a human can probably do okay on that search. Especially a human that knows what they're doing and knows what they're looking for. That's probably a couple good Google searches, you know, maybe 10 to 15 minutes and you're there. Push yourself. Combine the best elements of a large language model. Combine what you already know. Challenge OpenAI to find things that you don't know, but share with it what you know. Exactly what I did. I said, here's all the data. Here's what I think. Here's what I'm trying to do. Go, go research everything. Yo, go find everything I can't find. I don't have time to go dive into these thousands of pieces of data and to drown in this and then go research and research and research and research and research more. You go do it. Be smart with how you use these tools. Don't just blindly follow what other people online or the big tech companies are telling you to do with it. This is not a glorified search engine. This is the future of work. Use it accordingly. I hope this is helpful. That is what's new in open AI's deep research. Use it. My gosh, use it. But first, you might want to share this episode. All right, if you're still listening, share this on LinkedIn. All right? And I'm going to send you. If you didn't share our Last Deep Research episode, it's the same guide, but I put together 10 business use cases for deep research that I think are going to be extremely helpful. So if you want access to those now, like I said, tens of millions of people now have access, but you only have on that 20amonth plan. You only have 10 searches. You got to use them wisely, right? Go look at my use cases. Also, another benefit to sharing this is I'm going to choose one winner for a 90 minute consult. Maybe you need some help, right? I think one of the biggest gains you can get is rerunning a deep research query multiple times and looking at the results. Hey, I'll spend 90 minutes with you walking through, all right? But you have to Repost this on LinkedIn. Also, I'm going to announce the winner in the newsletter. All right? I'm not going to reach out and email you. I'm going to announce it in the newsletter like we did last week. We for our winner last week. And I'm gonna say, hey, you gotta reply, so you gotta repost this and you got to keep an eye on the newsletter. We're gonna announce the winner. All right? I like to make this fun. I get tired. I like to do little contests, all right? It just gives me energy, makes me smile. All right? So if you haven't already, go to your Everyday AI dot com. Sign up for the free daily newsletter, y'. All, I am busting my butt. So you can be the smartest person in AI at your company. Share the love, right? Sometimes people are like, oh, Jordan, this is so helpful. You know, you help me get a new job, you help me get a promotion. Everyone thinks I'm a. I'm an AI savant. All right, well, share this, please. You know, if you're speaking at a conference, throw. Throw the podcast up on a slide. I'd love it. Email your colleagues. I'd love it. If you're listening on the podcast, click that little share button on the podcast. Share it to a friend. Subscribe to the channel. Leave us a rating. There's so many ways that you can help us. I want to continue to do helpful deep dives. I want to bring on even more and better guess. But this thing only works if you tell someone about it. All right? Don't keep Everyday AI a secret. I'm not keeping any secrets from you. I'm giving you everything to get you ahead to grow your company and your career with Generative AI. Thank you for tuning in. Hope to see you back tomorrow and every day for more Everyday AI. Thanks, y'. All.

And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going for a little more AI magic. Visit your everydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.