ChatGPT vs Claude vs Gemini: I Gave All 3 the Same Real Job

So AI really is quite interesting now because the big companies are fighting for the consumers. Everybody wants to have the agentic consumer product. It started with Claude's Cowork and this week Gemini entered the ring with their Spark product. Last week we talked about ChatGPT Work. So now all of them have a product. So what I wanted to do today for, well, myself and therefore also for you is is compare them on day-to-day use cases, on things that I regularly do with ChatGPT. We prepared some mock data, we used some connectors in some of them, and we ran the same tests inside all three of these apps. That's going to be the theme and topic of this week's News You Can Use video, the show where we look at all the latest releases and then I give you my take on what I think is worth using. This week we'll be comparing the three big agentic consumer products plus a few other stories like OpenAI releasing their first hardware product. But those are really just side notes. Let's get to the meat of this video, starting with the methodology. So look, I guess you could do this in a really complex manner. What we did is we found a bunch of use cases for ChatGPT work as we were researching it throughout the past week. But then in our discussions, we kind of arrived at, hey, like a lot of this stuff is one-off uses. What about the stuff that we actually do regularly? Like summarizing meetings or looking through business documents or getting summaries of your meeting inbox. And the fourth one is letting AI help us with customer support and responses. We deal with a ton of customers. All of these are real things we do day to day. A bunch of the other use cases, they're interesting, but this is really at the core of how we use it as a company. And I thought these four use cases were just relatable for a lot of people. Day-to-day work, the typical stuff that consumers do. Which one of these tools is actually the best. And as mentioned, the free tools are going to be ChatGPT Work that released a week ago, Gemini Spark that, well, has been releasing over the past 2 weeks, but now it's finally available to all Ultra subscribers. That's the $100 plan. And then Claude Cowork, which is available on the $20 plan. So the cheapest one is definitely ChatGPT Work, but you can't even do all the things that we tried today on that plan. So essentially, you need the $20 plan for Claude or ChatGPT. And the Google one is only available behind the $100 plan. For all 3 of these models, I set it to the state-of-the-art model at the default setting. So that's Fable high, ChatGPT Sol high. And then in Gemini, you don't even get a choice on prompting. Okay, with that being said, what we're testing here is their version of Claude Cowork. And I'm saying that because they were first. It's basically a consumer-focused agent that plans and connects to a bunch of tools and then gets you output. It's supposed to work for you, not with you. And I think that's enough context. Let's actually look at the results here, starting with the first use case. Meeting transcript to action plan running in all three of these. Okay, let's start by Gemini because it's the new kid on the block. So meeting transcript analysis and planning, it's pretty simple. We just run this prompt that extracts five different things like this key decisions, action items, blockers, follow-up emails, and a Slack summary. And we attached a meeting transcript that in this case is a fictional one, but it's based off real conversations that we had with real team members. You can see it's a pretty short one. This is what we got from Gemini Spark. This is what we got from Claude Cowork. And this is what we got from ChatGPT work. So after looking over this for a bit, I would conclude the following: ChatGPT is the most complete and detailed. One thing that I particularly liked, and also our AI analysis, which I'm also going to show you, noticed, is that it flagged if certain ownership is implied, and I need to confirm it. I thought that was very thorough. And it just gave me this feeling of me being able to hand off important work to this. The others just made assumptions for everything that needed to be assigned. Overall, you can pause the video and look at these results if you want to look at them. Claude was the most concise and created a separate markdown file. Gemini just gave me the answer like so without many bells and whistles. And ChatGPT, again, very thorough with a lot of functionality here, like being ready to send this email, if I add the recipient right here, stable formatting that I always like, and clean sourcing, which you know, the others didn't have. Overall, I would prefer ChatGPT. And as mentioned, I run all of this through AI analysis. And the result for this one was ChatGPT with a 9.1 and Gemini being last. One thing that should be flagged is that Claude, as per usual, has the best writing. It was just more human sounding, but it invented and assumed a bit too much. I would totally agree with that. So overall, this one goes to ChatGPT. All right, next up, and we'll keep this very brief. It's just attaching a bunch of business context. We have this file called business DNA with all details on our products and what the outcomes are. And then I just ask rather generic questions. What is our core product? What is the outcome most of our customers can expect? I want clear citations, short answers, and the practical next step. The first question, all three gave an identical answer. But here on question number 2, the expected customer outcome, Gemini shows that the expected outcome is getting 2 to 5 hours back every single week by being a part of the club, which is our core product, which is very obvious from the business DNA. But the practical next step differs. So Gemini recommended that we audit all upcoming content streams to ensure every piece of content clearly serves this metric, the time that people get back by participating. Claude flags the same thing, but the practical next step is different. It says we should make the 1 hour per week outcome the headline of every piece of member-facing copy and the onboarding touchpoint. I actually like that more. It thinks about the onboarding, which is the beginning of the journey, rather than just all the content. And then it adds this filter that, hey, if the content inside doesn't fit the filter of saving time, then take it out. I think that is smarter than trying to force like a number onto each piece of content. And then finally, ChatGPT is the most concise, which I actually like because we asked for that. The practical next step is that we should establish a baseline for one recurring task, implement a taught AI workflow, and measure the time saved each week. So it's the most granular and practical from all three of these. I think personally, I enjoyed the Claude answer the most, but it's very close. And the AI analysis actually reflects that. ChatGPT with 9.0 and Claude with 8.9, saying Claude gave the richer answer, but ChatGPT followed the brief best. Short-sighted, practical, and no extra baggage. I would agree with that. This is a very close one though. Maybe I'd give this one to Claude. I will point out that Claude did not cite the source, which both Gemini and ChatGPT did. I like that. Okay, two more examples, and then we'll look at the theme between all of these. So this one is going to be customer feedback analysis. We have so many different customers, and we love looking at their responses through both a human lens and the AI lens. It's just useful. So what we did is give it data on customer feedback, just a table again created for this test, but based on real feedback. And the prompt here is very simple. Analyze it, top 5 recurring themes, give me the percentages, sentiment, 3 most important product priorities, a short leadership summary. Be specific, do not just summarize, turn the feedback into decisions. We're going to be looking for that. I always like saying something like this in the end of prompts that are rather vague, like give me the next actions, give me the next decisions that we need to take based on this analysis, not just, you know, a long document. Gemini Spark. Claude Cowork. Ooh, way more concise. And ChatGPT Work. Okay, it's a good table up top and super concise. Okay, leadership summary. Let's start with ChatGPT as we're here. Members generally trust the content and believe in the time-saving promise, but the product repeatedly leaves them at the boundary between understanding and execution. The immediate opportunity is not more content. It is better execution infrastructure. Clear next action, task-based technical runbooks, automatic guidance on using DNA files. Okay, what did Claude think about this document? Nobody's questioning the value. The feedback is about packaging, not product. People consume the content and then stall because they don't know the next action. 3 fixes: a mandatory one-action closer on all content, a de-risked Agent Hub onboarding for mastery, and a DNA cheat sheet. Address roughly 80% of the negative feedback. See how actionable this is? Again, this is just a theme, just like with the previous one. And this is my experience with Claude overall. It just tells you really what to do. Sometimes it might make assumptions, just like in the first example of the meeting analysis. This is really good. This just reads really well. Let's look at Gemini to round this out. Leadership summary. Our offering is structured like an extensive menu, which intimidates trial members. Key takeaways. Setup friction. 28% is the largest blocker, closely followed by onboarding direction and DNA management. Shift in delivery. We must shift from a content delivery long video model to an action-first pre-packaged prompt delivery system. Okay, I'm not sure I agree with that. I think to me clearly Claude won this one. Let's see what the AI analysis says though. Oh, it says again, ChatGPT won. Claude understood the member psychology best, but ChatGPT turned it into the clearest action plan. I think that's fair. He's probably referring to this part, the decision ready readout. And then this comment is interesting. Gemini was polished but over-centered technical setup. Yeah, even from that reading, I agreed. It was really going into the exact tactic, but it wasn't even a good one. And I don't think it had enough consideration for the member experience. And hey, if you're interested in these tools and wonder what else they can do, because they're so, so, so powerful, I created a document where I shared some of my favorite prompts and use cases for, well, all of these agentic consumer tools. I created it for ChatGPT work, but really it's applicable everywhere. And you can get it for free by going to the first link in this video's description. And yeah, we did test all of them and select from a dozen other use cases. So really, these are the ones worth trying if you want to go beyond this video. Let's get back to the results here. Okay, last one, really simple. And this is a use case more and more people are doing. You can simply connect Gmail to one of these accounts and it will just go through all of your emails and then give you a summary once a day. This didn't work so well in ChatGPT because it sort of pulled in the data and then messed it up sometimes. But with these products, these agents, hey, it actually creates a plan and goes line by line and calls itself, I don't know, 20 times if you have many emails. And this works really well. How well? Check it out. Intuitively, I would expect Gemini to shine at least in this one, no? I mean, it is Gmail that we connected, which stands for Google Mail. And this is the Gemini AI. So let's have a look. It's this simple email scan today's Gmail inbox, look at the messages. And then we want to surface a few specific things, mainly stuff that is immediately actionable. And then again, we want a summary and next steps. This was linked to one of my emails that I don't even use regularly. I suppose I signed up to the dentist there. So it highlights that there's some YouTube sponsorship proposal. There's some coworking event tomorrow, StubHub ticket survey, AI masterclass reminder. Let's have a look at Claude. So this is what the result looks like. Way more concise. And again, it highlights the dental appointment that has been canceled, the YouTube collab, and the event tomorrow. Very similar results. But it only gave me 3 instead of 5. Whereas I would agree the StubHub ticket survey reminder should not be in something that's actionable for me. I mean, that's a big miss by Gemini, no? Let's look at ChatGPT. Again, same setup, connector plus this prompt. Google security alert. Okay, so it surfaced that as the only one. I don't know if I like that because it just shows that it's a bit more security focused, ChatGPT model. Then the receipt for the Google AI Ultra account here. I don't know why it highlights that. The dental appointment that got canceled, YouTube collaboration, the coworking event and the masterclass. Okay, so this is similar to Google. I liked Claude the most. I think it did the filtering for what matters to me the best. And to round this out, let's look at the AI analysis here, which says ChatGPT is the best, with Gemini last. So I don't know, I would probably put Claude here. So the AI analysis is 4 for 4 ChatGPT. For me, it would be 2 of them ChatGPT, 2 Claude. And then I'll say one more thing. This entire analysis was created by OpenAI model. So it would often agree with its own results, right? So what I did is I ran the whole thing through Fable, the Claude model. And here, We're just gonna see what Fable got us. Oh, Fable agrees. ChatGPT is the winner on all categories. It likes that the inbox test comes down to security and billing. And then I guess here's the overall verdict from Fable 5. ChatGPT wins on a criteria that matters most for real teams. Groundedness, exact instruction following, and safety judgment. Claude is the clear second for voice and narrative synthesis. Gemini is competent but repeatedly needs cleanup. And then here finally you have recommendations. What you should use for what. Claude for member-facing communications, executive storytelling, emails in Igor's voice, Slack updates and leadership summaries. And ChatGPT for consequential work, action plans, doc-rounded answers, feedback, quantification, inbox security, billing triage, and anything where facts must stay separate from inference, aka the AI's knowledge. There's a lot here. Overall, I agree, ChatGPT for everyday work seems like a good model. Sol is unbelievable. Fable did well. It has the classic Claude strengths of writing well and surfacing actionable items. A lot of people call it like the creative genius, which I think is accurate. But for the day-to-day work, it's probably ChatGPT. And then Gemini just lags behind on everything. They haven't released a model that competes at the level of Soll or Fable yet. So, you know, that could have been expected. But I did want to put it to the test. And now you definitely know that the $100 is not going to be worth it if you're looking for a day-to-day agent to help you in your work. It's going to be between Cowork or ChatGPT Work. And there you go. That's really the core message of this video in yet another hotel room. And look, there was actually one more story that I wanted to briefly talk about. So let's do that. OpenAI is doing hardware now. They released the $230 Codex Micro, which is essentially like a Stream Deck if you're familiar with those. They have a bunch of buttons for agentic coding. I think the biggest feature for most people is a dedicated button to use dictation so you don't have to type your prompts. That's the same thing that I have programmed onto my Stream Deck. All right, that's the video. I'm going to keep using a combination of ChatGPT Work and Claude Cowork. My name is Igor Pogany. And with that being said, I hope you have a wonderful day.