ChatGPT 5.2 vs Gemini 3 Pro 🤯 (HONEST Comparison)

Google released Gemini 3 Pro and it crushed the benchmarks and everybody was going crazy about it. But now OpenAI has clapped back and we have ChatGpt 5. 2. And according to OpenAI, their new model beats Gemini 3 Pro in most of the major categories when it comes to the benchmarks. So in this video, I'm going to do a head-to-head comparison of both of these models to see which one is actually the best. And for this test, we're going to be comparing the Thinking Mode. So, this is thinking with 3 Pro and we're going to be comparing it against GPT 5. 2 with thinking mode and extended thinking so that it's a head-to-head comparison. Since Gemini 3 Pro is the king of the heel on the internet right now, we're going to use Gemini with each example first and then we're going to follow up with 5. 2. For

the first example, I'm uploading my entire YouTube channel analytics and I'm going to run several different tests using this data. The prompt is you have my uploaded YouTube analytics exports task. Find 10 insights that are not obvious at a glance and each insight must include the exact metrics used, the time range you use, the specific evidence numbers pulled from the files, the action I should take next week. Constraints. Do not guess. If a metric isn't in the files, say not in the data. Prioritize insights that increase click-through rate, retention, returning viewers, and RPM. Output format, a 10 row table, insight, evidence, why it matters, action, then three big bets for 2026 with a one to two sentence rationale each. We're going to send that prompt over and then we're going to go over to chat GPT 5. 2. We're going to upload the same files, press enter, we're going to paste the same prompt, and then we're going to send it over. What I'm looking for in this test is to see if it can crunch a lot of data without hallucinating. So, while Chat GPT is finishing up, we'll go ahead and look at what Google Gemini 3 Pro gave us and then we'll look at ChatGpt 5. 2. And because this is my channel data, I'm probably going to instantly know which one did a better job. So, 10 insights table. This is the inside evidence why it matters action. They got that right. I don't think Gemini did a great job of giving me 10 insights that weren't easily discoverable. The 500,000 impression leak. This one seems to have a little bit more promise. How to get 50,000 real Facebook followers. Got 501,000 impressions, but only 4. 81% click-through rate. This single video is being pushed heavily by YouTube, but rejected by viewers. A 2% bump in click-through rate equals 10,000 extra views. AB test thumbnail. Replace the current thumbnail. Use a high contrast before and after follower account image. Three big bets for 2026. Pivot to business strategy over platform tutorials. That's good. Own the annual update formula. Search first content packaging. You have videos with massive search impressions but low click-through rate. The algorithm wants to serve your content to searchers, but your thumbnails aren't winning the click. These are good insights. They're just not things that wouldn't be obvious. If you remember the original prompt that we ran and chat GBT is still thinking about it. The task was to find 10 insights that are not obvious at a glance and each insight must include and now not is in caps. But overall, Gemini did give me a useful response. I can use this, but many of these insights are obvious at a glance. These are things that I'm definitely aware of, especially with my thumbnails. But now, let's look at chat GPT's analysis and 11 minutes and 18 seconds. It should have done well. It had enough time to do it. So, time range notes. Your chart data and totals export span uh 2001 to 2025 based on date column. Your table data export doesn't include dates. So, I'm treating it as the same range. So, it definitely addressed the date aspect. And this format reminds me a lot of the 03 model when it first came out with the horizontal tables. Let's see if it came up with some insights that are not obvious. Once again, your biggest click-through rate win is hiding in already high impressions videos. And it pulls up the same video that Gemini did with a 4. 14 click-through rate. Now, it did add an extra note that there is a 13 minutes and 30 second average view duration. Oh, this is a different video. How to increase your reach and engagement on Facebook in 2024. How to get 50,000 real Facebook followers fast with 500,000 impressions. So, it's the second video that Gemini pulled up. This first one is something ChatGpt brought up. Another video uh that it brought up right here. So, it pulled up more videos than Gemini. Two, some of your best packaging is attached to weak early retention. Uh, click-through rate, average viewed, average view duration range among videos with a click-through rate less than 6. 83% top quartile and average view less than 31. 98% bottom quartile. The top examples include how to fix Facebook payout issues, Facebook monetization, Gemini 3 Pro can build websites. Your data suggests a two format strategy. Here's the evidence. Duration bucket, median click-through rate, median RPM, median subs, 1,00 views. Then in this column, why it matters. 3 to 8 minute looks best for clickability. 15 to 30 money plus deeper conversion. Shorts are weak for RPM. Next week, publish one long form pillar, 15 to 30 minutes for RPM conversion, and one tighter companion, 3 to 8 minutes that feeds into it via endcreen and pinned comment. Five longer videos convert subscribers dramatically better. Proxy for future return. How to increase your reach and engagement. 723 subs, 23 views, 34. 37 subs per 1,000. I believe I saw this in Gemini, even if I didn't mention it directly. Let me copy this like so. Yes. How to increase your reach 2024 drives 34. 4 subs per 1,000 views. And ChatGpt says it drives 34. 37 subs. So, Google Gemini rounded up to the nearest 10th. ChatGPT probably gave me the exact math. That's a nice little nuance right there. And overall, ChatGpt gave me far more evidence. If you look at each of these in this second column for the evidence, there are at least three videos supporting each of the ideas or insights that they're laying out with a time range even though it's a standard time range. And I guess the videos don't have a publish date. So, it did this, but it followed the instructions better. Add a single specific subscribe reason tied to a series. I'm releasing it next and place it after the first proof moment, not at the beginning. Some of your most lucrative videos are undercllicked. High RPM, low click-through rate examples. Chat GBT agent mode review. How to give chat GBT unlimited memory. My system. How to increase reach and engagement. These are money videos that aren't getting their fair share of clicks. Raising click-through rate here. Boost RPM and total revenue. Pick one. Click rate video and redo the thumbnail and title with a clearer outcome, stronger specificity. Your internal linking can be a real growth flywheel. It already works when you use it. Increen click-through rate. Increen clicks shown. Card click-through rate. Uh, the good news about the bonus ending 17. 56. Facebook pages actually pay better. Higher session depth increases the eyes. One end screen template. You're currently relying on two hit videos for a huge chunk of recent engagement. The best way to make money on Facebook and how I got 100,000 Facebook followers in 48 hours without ads. This is an opportunity and a risk. It's telling you exactly what the algorithm is rewarding right now, but also showing dependence. Next week, ship two sequels part two mistakes new update and interlink them as a miniseries. End screens pin comments. Number nine, you have real momentum and it's not evenly distributed across days. When you publish, you can change the size of the first wave which impacts early and the evidence is year-over-year comparison, day of week averages, top daily distribution. Schedule next week's upload for Tuesday or Wednesday, and plan a community post the next day to push a second wave. 10. Returning viewers isn't measured from these exports. You say returning viewers is the priority, but these files can't confirm what's working. re-export YouTube analytics with returning viewers plus new viewers plus unique viewers for the last 90 days plus the same breakdown by video. Then we can repeat this analysis specifically for return behavior. And quickly the three big bets for 2026. Build a high RPM long- form pillar channel strategy. Your data shows median RPM jumps from 11 cents on less than one minute videos to 9 to $10 on 15 to 30 minute videos. Long form is where you can realistically move RPM. That's obvious. Systematized packaging optimization, quarterly thumbnails, title testing on your top impression library. A single click-through rate lift on a high impression video can translate into hundreds to thousands of watch hours. Your own numbers show 3,63 hours of modeled upside across just three videos. Three. near a returning loop via series plus internal linking and that's just kind of like an insight from earlier. So when we compare Gemini and their bet is to pivot to business strategy over platform tutorials because of the RPM but for different reasons. They're saying to do it because of the topic. Your data proves that marketing keywords earn 36 RPM while Facebook keywords earn $5 RPM. ChatGpt is saying to do the same thing, but they're saying to do it because of the depth of the video. But if we remember these examples with the marketing keywords, I believe that these examples were from the chat GPT video. So, I'm just going to give it a very quick follow-up prompt here. Give me two example of videos that have the marketing keywords that earn $36 RPM. And for chat GPT, I'm going to ask, give me two examples of videos, the 15 to 30 minute range that have this nine to$10 RPM. And you notice that chat GPT doesn't even mention the $36 RPM. So these are definitely two separate insights. So the first one is what is marketing 3657. That is one of the first videos I posted on this channel. And this second one is how to use chat GPT and email marketing to grow your online business. I did this video about a year maybe a year and a half ago but it didn't get a lot of views. So while Gemini is looking at the data and saying yeah these are good videos because they earn more money you need to double down on those. I think this is just my hypothesis. I think that chat GBT is taking an approach that's more in line with the way a human would think. What is this person actually doing right now that's working overall? And so here are the videos. Facebook monetization 2023. We could easily change that to 2026. Easy to get and new ways to earn money. Why your followers never see your content and how to fix it. So that confirms my suspicions. ChatGpt is actually looking at my channel and looking at what content should I realistically create. So even though it's looking at the full range, I think in the context of the totality of my channel. So I'm definitely going to have to go with chat GPT for this one. So for the next test

we're going to use the same data. And all we're going to ask Gemini to do is use only the uploaded channel data. You are my chief growth strategist. I want a decision memo that answers what three video topics should I publish next and why are they the biggest expected value requirements? Choose exactly three topics. Each topic must include target viewer, core promise, title angle, thumbnail text angle, and a 60-second hook concept. Show your reasoning with a simple expected value model using numbers from my data. For example, baseline views, click-through rate, retention proxy, returning viewers, etc. Include two counterarguments per topic, why it might fail, and how to derisk it. End with a 14-day execution plan daybyday that a solo creator can follow. Hard constraint. If the data doesn't include a required metric, explicitly say what's missing and substitute a proxy you can support from the export. So Gemini's just responded. It took a little moment to think about it, but Gemini is typically responding much faster than chat GPT, but speed isn't everything. Accuracy and quality is probably more important to most people. two chief content officer from growth strategy team December 12th 2025 subject high leverage content plan three topics to publish next volume leader Facebook growth revenue leader business wealth growth leader AI automation strategy we will not choose just one we will publish a barbell strategy portfolio one for volume one for revenue one for hybrid growth topic one the profit multiplier concept how to build digital products with AI in 1 hour. Target viewer, the creator who has some followers but no income. Core promise. I'll show you exactly how to turn your content into a paid product using tools that do 90% of the work. Evidence title angle Gemini 3 can build your first product in 14 minutes. Thumbnail text 0 to 1k sideby-side comparison 60-second hook. Most people use AI to write emails. I use it to build a $50 digital product in 14 minutes. Here's the exact prompt chain I use. That is not a 60-cond hook, but these are some good video ideas. The 2026 Facebook monetization road map, the trust builder, why your reach is dead and how to revive it. These are all some good videos and spins on old ideas, but I can roll with that. So, for chat GPT, it didn't even write it as a memo. It just says decision memo. next three videos. Topics with highest expected value, uh, using only your uploaded content, what's missing, simple expected value model. At least Gemini did try to format it like a memo with the to from subject, even though it wasn't formatted correctly. So, I'm going to have to say, you know what, Gemini did better at, you know, contextually understanding what was going on and how to respond. And once again, we have that horizontal table that reminds me of the 03 model. I kind of like these tables. They really give you a good way to look at the data because it's all right there. I do wish it would kind of shift over here to the left a little bit so that it's more centered, but channel medians all videos topic chat GPT course creation Facebook monetization 2026 angle AI builds your website page product. Same thing as Gemini. Geminis's was how to build digital products with AI in 1 hour, Facebook monetization 2026 angle and why your reach is dead and how to revive it. So I think that chat GPT doubled down on the AI. So chat GPT course creation and then AI builds your website or product page. So, when it comes to this particular example, I'm kind of torn in between the two, but I do have to lean towards Gemini 3 Pro a little bit simply because it did try to format it as a memo because both of them essentially gave me the same three topics. Chat GPT just doubled down on the AI course creation, product creation angle. So, I do feel like chat GPT understands the context of my channel better because I would personally rather do those three videos even though there's major overlap between the two. But for the next prompt, we're going to test something different. So, one thing they've been

talking about is the ability to create slides. So, we're going to turn the analysis we just did into a 10 slide PowerPoint. Each slide has these specific requirements. Executive summary, what's working, what's failing, audience signal, packaging, retention patterns, three fixes, three topic bets, 14-day execution plan, stop doing list, five items, and a scoreboard, what we'll track weekly, and then these are the designs. Clean, modern, use charts only if they're directly derived from the data. Include speaker notes with the key natter for each slide. If you cannot create a PowerPoint directly, output slide by slide content, format it so I can paste it to slides quickly. All right, because I cannot generate a PowerPoint file directly in this environment. So this is what I'm going to tell it to do because I do want to see its design. So I think it can do it if I use canvas. So what I'm going to do is just start a new chat. I'm going to choose canvas and then I'm just going to tell it to or let me get the prompt that I used and we're going to run this using canvas and see if Gemini can create the slide deck using canvas creating presentation slides. Once again, chat GPT is infamously thinking long. We're at 2 minutes and 8 sec, but we do have extended thinking turned on and I'm going to go take that coffee break. So, ChatGpt just finished after 10 minutes and 13 seconds. And it says that it thought for 10 minutes and 21 seconds. If you look right here where it's buffering, we can see that the YouTube Analytics Insights artifact in progress 98%. I think that this is a new notification feature. I've never seen this with Chat GPT before. It's at 99%. And so while chat GPT is finishing up, I was able to use canvas to create a slide deck inside of Gemini 3 Pro. And there's a light dark theme button here, but only for the background. And so if I look through this channel growth strategy presentation, we look at the executive summary, which is going to give us an overview of the entire thing. There's an efficiency gap, traffic leak, and a growth engine, which basically talks about the annual and AI toppers are currently driving three times subscriber growth, if that's what I'm looking for. Traffic leak, we have 500,000 impression opportunities in search, failing due to low click-through rate, less than 4%. We are earning 10 cent on shorts, but saying don't create shorts, focus on business content. This is because of that. This is marketing uh skewing the data. It's an old video of about 2 minutes and it has a couple hundred views and Gemini 3 Pro is looking at that and it has skewed everything that it's shared with me from that one single point. And so it's inability to capture the accurate context of my channel data overall is glaring to me because I have a very close relationship with this data. So I can instantly spot that. But other than that, it does put together a visually appealing presentation. Now, albeit when it comes to images like this thumbnail here, you're going to see when we get to the end that all of these images were borrowed. And so, essentially what it did was it borrowed these images from different sources. And I don't even know if that was supposed to be slide 12 or not, but the data was good, the design was good, it was a decent presentation. kind of generic, but it it'll pass. And chat GPT is still at 99% developing this PowerPoint document. And if we look over here, we can see that it's thinking again at 2 minutes and 50s something seconds because it made some mistakes applying a patch command. Uh, and I guess this is was in the process of trying to actually create the images and the presentation. And so right now we're currently at 13 minutes overall waiting for ChatGpt 5. 2 to actually create this presentation. And so one of the biggest things with 5. 2 that's going to be a make or break for it is how people view the quality of the outputs. If I'm waiting 13 minutes for an output that's only slightly better than Gemini 3 Pro, then even if it is slightly better, I still think that's an L because I'm waiting 15 minutes just for an incremental upgrade. Okay. And so after 10 minutes and 21 seconds and an additional 7 minutes for a total of 19 minutes and 38 seconds, we have our YouTube growth memo 10 slides PowerPoint presentation. Uh we can play them right here and let's just take a quick look at this. So Chad GBT and Gemini are telling me the same thing. Make longer videos, work on your thumbnails, and [snorts] double down on AI content, product creation, creating things with AI. So, here's what's working. How to increase your reach and engagement on Facebook in 2024. High RPM content, course cluster median, expected 431. 1 video model, internal linking, 17. 56% endcreen. The good news about the bonuses ending, no one is talking about. I think that's kind of an anomaly, but Gemini and Chat GBT both latched on to that particular idea. Uh, but one thing I will say when it comes to the design, Gemini just did a better job of packaging the data of designing it. And Gemini, to be fair, Gemini did use Canvas. And I could probably take this slide information and ask chat GPT to design it in canvas and maybe it will give me something better. And I'll do that now and we'll come back to it at the end of the video. Maybe redesign these slides so that they're beautiful. And of course with chat GPT we can use canvas or we can use other apps like Canva, but we're just going to use Canvas. For the next prompt, we're going

to tell Gemini to create a downloadable spreadsheet that I can use weekly to run my channel like a business. Include weekly KPI tracker, video backlog with status, packaging, test log, content calendar, and then the rules include formulas where appropriate, make it clean and actually usable. Prefill it with the best baseline values you can pull from my uploaded exports. If you can't generate an XLX XLSX, output a structured CSV table for each tab or use Google Sheets. Copy that and let's run it with chat GPT and let's wait 30 minutes. See you guys tomorrow. So, chat GPT failed to generate the spreadsheet. It retried it still failed and it actually finished it. It said it finished thinking and it was actually outputting the sheet and it stopped. I'm going to hit this retry button again, but we're going to keep moving forward because that was like 10 minutes I waited and didn't get anything. So, I don't know, maybe this stuff is a little too complex for GPT 5. 2, but let's see how Google Gemini handled it. So, Gemini actually gave me a download link for an Excel SSX file. I didn't think Gemini was going to do that. I think it was going to default to Google Sheets, which is why I added that little comment at the end. But look at that. Okay, smarty pants. Both of these models failed to actually give me what it said it would give me. The link doesn't give me an actual file to download. Perhaps you should use Google Sheets by chance because I don't think that Gemini can give you download links. Everything that Google has, they try to do this closed garden. So, it's always use canvas and then export to sheets. I don't want to turn on canvas because if I do sometimes it changes the conversation and kind of forces you to always use it or something. But I'm going to go ahead and turn it on right here and just let it have edit. But in the meantime and in between time here is the slide deck that we asked chat GPT to redesign using canvas. I believe you can see how it added these bars at the top, you know, the circles here, there, but the text, it's not really handling that well at all. It's just so garbled up at the bottom. This was not a good design. So 5. 2 still doesn't have that clean design that I'm looking for. It just doesn't have it. But I know who does know how to design something. These PowerPoint slides don't look good at all. I need something that looks great, that feels like my channel, that something I can present to the public and be proud of. And you might say, well, it's the prompt you're giving it, but I just think it's the skills and the tools that the model actually has. And so if we just look at the YouTube growth memo that Claude was able to create, we see that this is much more in line with something that we're actually looking for. So Claude is still better at designing and creating these type of assets, even better than Gemini if you ask me. I cannot directly create a Google sheet for you, but I have generated the four separate CSV files below. Download each file, open Google Sheets, go to file, import, upload. This is the KPI tracker. This is the video backlog, packaging test log, content calendar. And if we go back to the prompt, the prompt was create a spreadsheet that I can use weekly to run my channel like a business. tabs required. Four tabs, weekly KPI tracker, video backlog with status priority scoring, packaging test log, content calendar. And so after 13 minutes and 18 seconds, Chat GPT 5. 2 was able to create the spreadsheet with the weekly KPI tracker that we can view right there, the video backlog, the packaging test log, and the content calendar. And it is interactive. Here's the video backlog. Build course with chat GPT Facebook monetization 2026 decision tree. These are the videos that I should be working on. The three ideas topic cluster status idea priority score expected impressions expected click-through rate target viewer core promise title angle thumbnail text angle 60-second hook concept. Chat GBT gets a plus. Not the most beautiful thing, but it was able to do it. Took it a while. Wasn't able to oneshot it, but did a much better job than Gemini. So now what we want to do

using this same YouTube data, we wanted to find a short list of brands that will be in alignment with my YouTube channel. And then we wanted to create two specific deliverables. Number one, I wanted to create a media kit. And number two, I wanted to write an uh an email pitch that I can pitch them on for sponsorships on my video. So Gemini couldn't create the media kit because it's an image and it needs Banana Nano to do that. And so I just dropped it in there and asked it to create it to get an overview. I knew it was going to do all the slides at once and I would have separated it slide by slide if I had the time to do that, but I just wanted to see what it would create for me. And this is pretty good. And I actually think a better way to create this with Google's products would be to use Notebook LM. But when it did a research

of the brands that fit my channel, I was surprised to see that Metricool was number one because I'm subscribed to them. My email, they always do these in-depth studies about social media that I love reading. I've actually done a video or two about it and I actually have one that I planned on doing sooner about a study they put out about Facebook, Instagram, YouTube and all the platforms right now. The second brand is Standtore, the third is Nvidia AI. The fourth is school and the fifth is app sumo. So this is pretty good especially considering that I've worked with two out of these five in the past before and the best brand pick is metrical. And they even provided me with an email to contact. So I put the slide deck together and reach out and see which one responds. And then when we look at chat GPT, they say five sponsorship fit brands. Pick from your channel data plus real contact info minihat. Why it fits from your exports your channel's biggest winners are Facebook monetization plus growth. Uh you already have a minihat related video. Uh that's old and it's something I just put up to help people. So I don't know. Kajjabi. No, they have an affiliate program at Kajjabi and so if you don't use Kajjabi, they're probably not going to pay for a sponsorship. Metrical number three. So Metrical was number one for Gemini, but it came in number three right here behind Minihat and Kajjabi. I'm not sure who has the best email because Google Gemini actually gave me the name of a person to get in contact with whereas this is the PR team and this might have something to do with chat GPTs content policies and personal information stuff like that. But my first mind is Gemini 3's because they said they got theirs from LinkedIn I believe and this is press work with us contact on metrical newsroom. So GPT 5. 2 to big field right there. Gemini 3 Pro definitely got it right. Uh Web Flow. I've used Web Flow in the past and so I might reach out to them. I still know how to use it. It's a complicated platform, steep learning curve, but once you learn it, it's really good platform. Uh Kit, I'm actually an affiliate for Kit. I like them. So, this is a good one as well. All right, step two. Okay. Uh number one, best fit brand, Minihat. I don't know about that. that, but I'm going to go with it. it because you recommended it, chat, GPT, and we'll see where it goes. This is the sponsor media kit. Sit slides. Let's open this up or let's download it then. Let's open it up. All right. And so, here we are inside of the sponsor media kit for Minihat partnership proposal. We have a channel positioning uh slide, audience, and value proposition. What we can and cannot claim, none of the data. uh proof points from your channel, sponsorship packages, mini chat integration concepts. So, it put together a media kit, but I can't say that this is the best even by the standards of the content like proof points from your channel. It just didn't provide the data and it didn't list the titles, names, nothing. So, yeah, that's not going to be it. So, here's the email. Hi, Minihat partnerships team. I'm reaching out to explore paid YouTube integration with Minihat. My channel publishes practical playbooks on Facebook monetization and creator growth. A couple of proof points for my YouTube analytics exports the best way. So, here are the videos and the information that should have been pasted or placed inside of the media kit that it just didn't include. Many chat is a na natural next step for this audience. After I teach them how to get reach and qualify viewers, we can show how to convert that attention into leads and sales through IG Facebook DM automation. Let's look at Gemini's pick. They picked Metrical. Your channel is effectively a Facebook help desk. Metrical is the tool that solves the number one problem your viewers have after they watch your video. How do I manage this page easily? Volume alignment, retention fit, uh deliverables. This is the media kit. Facebook growth authority audience and research performance proof sponsorship packages 450,100 300 and those prices are too low but uh the integration concepts for metrical 30-day scheduling challenge contact. Here's the email. Hi Aniston. I run a YouTube channel dedicated to Facebook professional mode and page growth. My channel is the help desk for creators struggling with Meta's native tools. I noticed metrical is aggressive on helping creators streamline their workflow and I think my audience is your exact ideal user. Why this fits high intent, deep intention, problem solution. I think this is actually the better email. I got to give this to Gemini 3 Pro. I feel like this is going to be a good pitch for them uh and if they accept a good partnership. And so

for this next example, we want to test the ability of 5. 2 2 to C because they claim that the vision got better. And so what I've done is provided a screenshot of my meta ads manager account and I'm asking Gemini and chat GPT to look at this meta ads manager and goal is to set up a new campaign for YouTube lead magnet funnel. Uh, I gave it the objectives, the budget, audience, placement, creative, tracking, give me exactly what to click next, step by step, the exact settings to choose, the three most common mistakes at this step. If you cannot see something clearly in the screenshot, ask me for the one missing screenshot you need, only one. I personally don't expect Gemini or Chat GPT to ask me for any screenshots because they've been trained on this data and they can look at the website and so forth and understand how to get this job done. All right, so Gemini is done already. All right, leads versus sales. Choose leads. What to click next step by step assuming you're on the campaigns tab. Click the create button. Leads. Continue. Manual leads. Do not choose advantage plus shopping or tailor leads if offer. We want control over the audience stacks. That's true. Campaign level name it top of funnel uh lead magnet YouTube. Scroll down to advantage shopping campaign budget formally CBO. Toggle it on. Set daily budget to $50. Why? With $50 a day and multiple audiences. This lets Meta auto shift money to the winning audience. True. The exact settings at the ad uh set level. Uh you will need to duplicate this asset three times to create your stacks. But here are the settings for the first one. Conversion location website assuming you have a landing page. Performance goal maximize number of conversions. Select your pixel lead. Start event tomorrow at 4:00 a. m. United States. United Kingdom. Tier one English. And I can tell you right now that Gemini 3 Pro knows how to advertise. At least it knows how to set up the ads. Three most common mistakes. Optimize it for link clicks instead of leads. People should know this by now. This has been an old mistake for years with Facebook ads. Uh the buttonclick pixel era. Do not fire the lead pixel event when someone clicks the submit button on your landing page. No, you have to fire when they get to the next page URL. Thank you. That way you know you got a conversion. Let's go over and check out chat GPT. Objective choice. All right. Exactly what to click next from your screenshot. Campaigns. Create button. Leads. Continue. One campaign. Three adsets. Two ads per adset. All right. Uh asset one. Next image. Oh, creating the ad image. Image number two. Short video. Publish or go back and duplicate. All right. The exact settings to choose. Campaign settings. Objective leads. Auction. It's always a auction. YouTube uh campaign name. YouTube leading leads. Braw three stats. $50 a day. That's a long name. Uh, CBO on daily budget $50 AB test offer now special ad category none. That's a good catch right there by 5. 2. Just want to give it credit that it actually, you know, made a mention of that. Gemini probably left it out because it's not there based on the context. I'm not going to be advertising about housing, credit, or employment. Asset settings conversion location website performance goal maximize number of leads conversion event lead budget and schedule audience broad location set to where you actually sell or serve. Oh, this might be off right here. This might be off. I'm going to wait. I'm just going to look over this a little bit. Interest stack. So, these are the interest stacks. Okay. So, placements. This is good that it caught that creative image plus one short video. This is technical information. It's not really giving me the creatives, but we didn't ask it to either just to be fair. But tracking must be installed configured. Uh link clicks over landing page views. Letting placements auto run without creative built for them. That is definitely a huge mistake. If you have an placement selected, you need to have a creative for it so it fits right and your creative isn't getting cut off. Words aren't easily seen because you're paying for it. Overtargeting interest, too narrow with low budget. If you go too narrow, it's going to raise your budget up. Let me look at Gemini once again. I didn't find anything wrong with what Gemini laid out. But this one single fact right here. broadest oneapper asset location set to where you can actually sell serve. So Gemini, this isn't a wrong answer. It's the right answer, but Gemini is demonstrating more foresight. Set it to these four countries. It's understanding that it's a YouTube lead magnet. So if magnet, then it's a lead magnet having something to do with possibly growing your YouTube channel or something like that. And so it's targeting the tier one English countries, the US, Canada, uh, Great Britain, and Australia, right? That was perfect as far as targeting goes. Chat GPT 5. 2 says audience keep it broad, which and this is it's not wrong. I want you to understand that. It's just that Google Gemini 3 Pro is better. So I have to give that to Gemini 3 Pro. [snorts]

But for the next response, I have here some HTML uh that I used inside of Gemini to create an AI app and it needs to be debugged. So basically, I'm going to ask Gemini to find the bugs and fix it and I'm going to ask ChatGpt to do the exact same thing. But in addition to asking ChatGpt 5. 2 to do this in the chat, I'm also going to open codeex and I'm going to run this prompt over there. And when you run these type of debugs and codecs, I can get several different versions if I choose to, but I'm only going to go with one because of course if you have several versions, you could probably get it right. But I just want to see if, you know, [snorts] if codeex is better than 5. 2 and if 5. 2 is better than 3 pro. And one of the major setbacks of Chat GPT and just OpenAI as a platform is this right here. After Gemini finishes debugging and writing the code, I'm going to be able to use this application right here inside of Gemini. And there's a little tab that should pop up in the bottom right corner where you can turn on Gemini Power and you can power this app you build inside of your account. And you can do the same thing inside of Claude, but you can't do it inside chat GPT yet. And it's not just that, it's several other smaller things. Chat GPT is a platform that constantly focuses on hitting the benchmarks, being the number one platform when it comes to the benchmarks. But their ecosystem, features, I feel like they're starting to fall behind and they haven't released any new features or at the very least updated their platform so it can do the same things that Claude and Google can do with Canvas. So, Gemini 3 Pro is done and you can see that it's asking me to enter a Gemini API key and you can get a free key at Google AI Studio. I'm pretty sure that works. But what we're going to try to do is add Gemini features right here. And we're going to see if this takes this away. And if we come back to chat GPT, we see that the top five issues preventing reliable operation have been identified. We're going to copy this code because there is no environment that you can run it in inside of chat. Actually, I'm not going to have to wait for it to finish writing the code because it's still writing. But at least it's done. And Codeex is still working on its version. So, we should have those shortly. But back in Gemini, we can see that the app has been updated. I've added two powerful new large language model features to your app to complete the packaging workflow. AI virality check. Step one. Step seven, a launch kit. Here's the updated app. And we just have to wait for Gemini 3 Pro to finish writing the code. So, we ran the same code through codeex. And this is the file that we got when I saved the HTML and I open it up. So, of course, it's going to be some problems, some debugging things that need to be done, but I didn't think that it would look like this when I open it up. 5. 2 is still writing the HTML. So, we'll go over to Gemini 3. This is the code. Let's go to the preview. I'm going to click right here to go over here, get a free API key. All right. And so, now we set this app up. It's asking us for the video topic. Chat GBT 5. 2 versus Gemini 3 Pro. Target audience, content creators, viewer outcome, choose the best eye. We can analyze the virality or generate titles. I'm going to analyze the virality first. And we got an error there. Got a fail API error. That shouldn't have happened. So, let's just see if we can get rid of that. And we can't Oh, I think I finally pasted in the right API key. Generate titles. And that was it. So, it's analyzing viral patterns. So, it's going to do that first anyways. And then we have the AI recommended. Stop using the slow AI for scripts. I tested both. The creator verdict. Choose your AI fast versus accurate script. How I choose five AI fails that wreck my script. I don't know why it's bringing this script writing in there. I think because I said content creator. Uh but I'm just going to pick one. I'm going to go with that one. And so now it's going to come up with the thumbnail text. You're doing it wrong. Let's go with that one. Video hooks. Uh, this is our title. This is our thumbnail. Now it's writing our hooks. If you're still typing prompts like, "Generate a 60-second YouTube script about productivity. " Stop. You just wasted 5 minutes. Uh, you know that feeling? How long does it take to get a high performing script from AI? 3 minutes. 5 minutes. Let's just go with this one. Confirm and upload assets. Now I have to upload an image of myself and a style reference. And it'll create the thumbnail. And so now I'm going to upload a face. And then one of my latest thumbnails. Save to title bank. Save style to title bank. I'm going to say white, red, bright. And click generate four concepts. All right. And so now we are inside of the code that chat GPT wrote. Enter your Gemini API key stored locally in this browser. I think I still have it. I did not. Still have it. All right. This page says, "Enter your Gemini API key stored locally in this browser. " So, when I pasted in my API key, it actually showed it instead of hiding it. So, got to cut that part of the video out. Here are 20 title options based on your brief. Click one to select it. I wonder if I could have pasted in my chat GPT API, most likely. Choose your AI. Which AI writes better scripts? And so this is powered by Gemini, but let's just roll with it, right? I'm just going to come down here and choose creator challenge GPT prompt. Let's go back to Gemini. Generation failed. Generation fail. All right. Something's going on with the images. And the problem with this, I know it works because I've created thumbnails with it, but you have to debug the code and it requires a lot of time. Time that I don't have right now. I'll back up and try it again. Oh, you know what I should have done? I should have done is left this blank. I'm just going to leave that blank this time so that it's not trying to save to style bank. All right, let's choose that one. Going to wait for the hooks. I'm just going to choose a hook and then upload the assets. Going to upload a file for my face. Going to upload a reference thumbnail. Same one. And then I'm going to click generate concepts. Not going to choose anything as far as saving it in the bank. Here are six steps in this one. API error. API request failed. So, I wasn't anticipating fixing an API error failed during the video recording. Click on concepts. I'll go back to assets. And then I'm going to click generate for concepts. And this should work. This time, I had to set up a billing account. It's set up, so it should work. And even if it doesn't, you can do this without setting that up because you have Nano Banana right here inside of Gemini. But I'm going to try the same thing over here just to see if it was a matter of setting up the billing with the API. But either way, you see that chat GPT doesn't have the environment to run this natively, but it did actually code this accurately. And so this is really Gemini versus Gemini besides the actual design and layout because we're not using the OpenAI API for this because it really can't generate images like Gemini can. So, we see right here that it actually generated four thumbnails for me. All right. So, why is it so slow? This is definitely a good thumbnail for a chat GPT 5. 2 video because man, it takes a long time. Why is it stuck? Infinite loading glitch. Angry face pointing down. Time flies. Stop the AI. Let's just uh go with this one. And when I click on it, it downloads the thumbnail. I can upload this to my YouTube channel. And now I'm going to click the generate launch kit. And so now I have my own app built inside of Gemini 3 Pro that you know I can start with the brief work through titles, thumbnail text, hooks, assets, concepts, etc. I try to do some other things with uh Google gyms with this. It's a little bit difficult. not the best place to build an app, but it's definitely somewhere where I can come and quickly prototype a video just to get an idea. Like, I could use that thumbnail it created for me, but more than likely what I would probably do is end up just using it as a prototype and then creating the thumbnail myself. So, basically, is CH GPT 5. 2 better than Gemini 3 Pro? I mean, the benchmark say it is, but when it comes to some of the tasks that I requested it to do, it just failed. It couldn't do it. But then there were other things that Gemini 3 Pro couldn't do and Chat GT 5. 2 was able to do. One of the biggest takeaways that I just have to be honest about is the incredible amount of time it took me to record this video because of the response time for ChatGpt 5. 2. Look at the thinking for these messages. Thought for 11 minutes and 18 seconds. Thought for 9 minutes and 24 seconds. Thought for 10 minutes and 21 seconds. Thought for 19 minutes and 38 seconds. Thought for 20 minutes and 6 seconds. Thought for 14 minutes and 39 seconds. Thought for 1 minute and 50 seconds. And then when it rewrote the code for the broken app that I was trying to fix, thought for 2 minutes and 34 seconds. That puts us at a total of around 90 minutes for me to actually get these different outputs from chat GPT just to do this video. That is going to be one of the biggest bottlenecks and things that you should hear about with 5. 2 thinking, especially extended thinking, and that's to be expected to some degree. But 90 minutes, I mean, do what you will with that information, but this was the only video I was able to record today because

it took so long. Listen, I don't know. Which one do you think is the best? 5. 2 is obviously better than Gemini 3 Pro in some regards, but is it worth it when it comes to the amount of time that you have to wait to actually get that output? Is that incremental improvement worth the time you lose? I don't know. I'm going to keep playing with it, testing it on some other things that I use AI for that are more textbased to see how I like it. But let me know what you guys think in the comments. If you got value out of this video, make sure you hit the like button, subscribe to the channel, and as always, take care. Have a good day and I'll see you in this next video right here where I'm showing you how to use Notebook LM Claude Opus 4. 5 which is my personal favorite AI to create YouTube videos in 10 minutes.