Of all the great and wonderful ways that we can use AI today, I believe that the most beneficial to everyone across the board is AI powered web search. Giving AI the tools and the ability to search the web for us. If you think back to this time last year, we only had one way to connect LLM to the web. Perplexity. It was the first tool that connected LLMs like ChatBT or Claude to the internet. But over the last few months, all the major AI players added web search to their tools. Chat GBT has web search. Google Gemini has web search. Grock has web search. Claude has web search. It's become very standard. And then when Anthropic released MCP servers, we had the ability to add our own custom web search tools to Claude or any other client that supports MCP like Brave Search to Villian and Firecrawl and make them work better for us to the best of their ability. But here's the thing, getting valuable, accurate AI search results is crucial whether you're coding, doing market research, making big decisions, or even just writing. Because when AI has access to real-time current information, it reduces hallucinations and gives you answers you can actually trust. And the truth is, despite us being way more efficient than when we were just searching Google, AI search still has major limitations. Websites just don't want bots looking at them and taking their information. And that's why sometimes you have to solve captures or choose all the bikes in the pictures or click buttons to prove that you're human. These are all designed to stop bots from crawling websites. And unfortunately, these tricks usually trick up the AI or the website itself can tell that it's not a human and will just cut off access. So in this video, I'm going to show you hands down the best tool for doing AI web search. And that tool is the Bright Data MCB server. It could solve captas. It solves geoloccation issues. It's the only one you need. Let me show you what I mean. A few days ago, OpenAI released Chat GBT agent which essentially combines their operator tool and their deep research capabilities. And this is supposed to be very powerful. It's supposed to be able to do all these things for you on the web. And of course, when they announced it, it looked amazing. So, I had to try it out. So, just to show you agent mode really quickly. The first thing you do is do / agent. I made a specific prompt here telling you chatbt to look at this Zillow property and give me information on it. I tried to open up Zillow. It wasn't able to. Then I went to realtor.com and then Redf Fin. After 14 minutes, it said the Zillow page is asking to press and hold a button to confirm that we're human before it will show the listing details. To proceed, it wants me to interact with the button. So this is one of the features of chat GPT agent. It's supposed to be able to interact with the web and do deep research and if not, it can ask the human to take over. So now I'm taking over Chat GPT. This is the website they're on. It's asking me to press and hold this button. So eventually Chad GBT was able to get to the Zillow link. ChatGpt worked on this for about 20 minutes. gave me a decent report. It wasn't able to access Zillow itself. I had to interfere and take control to click through to Zillow. And that's exactly where today's sponsor, Bright Data, comes in. It is the only tool I need to search all these different websites and give me up-to-date information. Now, I'm going to send the same exact prompt to Claude with Bright Data. Okay, so now I'm in Claude. I'm going to give it the same exact prompt, but I'm going to tell it to use Bright Data. So, within a few seconds, Bright Data was able to access Zillow. As you see here, it went to the URL. It got all this information. Then it's using its own built-in search engine. Now it's using its built-in Reddit tool to directly search Reddit. After all that, it's getting specific information about the neighborhood. So Claude did this in a few minutes. And if we look back at Chad GBT's answer, it's not really put in a way that's easy for me to understand. And this could just be the difference between Claude and Chad GBT. I think Claude writes a lot better. But the bottom line here is with the bright data MCP, it was able to access Zillow directly, go to Reddit directly, and not have to find all these workarounds to get the information we needed. So I just want to show you the bright data MCP within cloud. They have 60 tools enabled. They have scrape as markdown, scrape as HTML. They have extract. They can get web data from Amazon, Walmart, eBay, Home Depot, Zara, Etsy, Best Buy, five different LinkedIn tools, Crunchbased, Zoom Info, four different Instagram tools, four different Facebook tools, four different Tik Tok tools. It could get information from Google Maps, Google Shopping, the App Store, and then Play Store, different news sites, GitHub repositories, Yahoo Finance, Twitter, Zillow, Booking.com, YouTube, Reddit, and it also has a scraping browser. And this is really powerful because if it doesn't have a specific tool, it's able to use its own browser and go get the information like we see in some of the other MCB tools like Tilly, like Firecrawl, like Puppeteer, but it's way more powerful and it's way smarter. And that's because Bright Dead has been doing this for years. By the way, last week I was vibe coding an app with Kira.dev. It's a great tool. I'll link to the video above. And anyways, I ran into a problem that wasn't mentioned in the documentation. And I tried using Brave search to search the web for others that had similar problems or find some type of solution but didn't find anything. But then I had Kira use bright data search Reddit specifically for this problem and I found other developers with the same exact issues and found a solution in minutes. And the difference was that Bright Data was able to actually access Reddit and search through it properly. Look at subreddits, look at comments, whereas the other tools like Brave Search was just skimming the results. Okay, so I'm just going to show you one other prompt that I believe is simple but demonstrates just how powerful and accurate Bray data is compared to all the other major players and the other MCPs like Brave Search and Tilly. And this is a prompt that I run often and I just ask for the latest products, features, models and pricing of all the other frontier AI models, Perplexity, Gemini, Grock, etc. This should be a straightforward and easy to do task. It just requires going to websites and getting real-time information. And I've asked this to chatbt, Perplexity, Perplexity Labs, Gemini, Claude, Brave Search, Tavilli, Firecol, and none of them get it right. And by the way, this was the first thing that I asked Chad GBT agent the other day. So it took Chad GBT agent 40 minutes to give me this answer. This is right when Chad GBT agent came out, July 18th. And right away, we could see that it fails. It only mentions J GBT40. But if we even go down to anthropic, we see that it mentions the Claw 3 family, not Claude 4. So, straight off the bat, we see this is a fail. It wasn't able to get the latest models. And I also ran the same prompt to Gemini 2.5 Flash. It answered within a few seconds. And looking at it right away, I could already see it's mentioning chat GBT 3.5 mentioning 01 pro mode. This is already out of date information. Let's look at anthropic cloud 3.5 sonnet. There's already been two updates since then, so Gemini failed. Let's also just try AI Studio. I made sure to have thinking turned on and also grounding with Google search turned on. It does mention 03 and 03 promo. So that is up to date. When it gets to anthropic, it mentions cloud 3.5 sonnet cloud 3 opus. Okay, so again another fail. These are outofdate models. So Google their AI models with Google search turned on was not able to get the latest information. So we're just going to give the same prompt to Claude. I'm going to say use bright data. You can pause or read the prompt if you want. It looks like it was able to go to every website and get the pricing and model information. So, let's just check it out. Made us a table here. It's mentioning the correct OpenAI models. Sora codeext agent. So, it's getting up to date information for Google has latest models pricing 250 for AI ultra. So far, everything is looking good to me. Anthropic, the correct pricing plans, XAI, Super Grock, $300 per month, Grock 4. Okay, so this succeeded. It went and got all the information, which for some reason is so hard for Chad GBT, Jadubt Agent, Perplexity, Gemini, all these frontier AI companies to get the latest models, features, and pricing. But when I use bright data, it got it all right. And obviously, you could apply this to so many different things. And I've personally used it for market research, for competitor analysis, for finding relevant people on LinkedIn, and also make investment decisions. Okay, so now let me just show you how incredibly easy the setup has become. You literally just need an API key. First, you go to brightdata.com. I already have an account, but you could just press get started for free or start free trial. So once you create a free account, they'll send you an email with your API key, but if you want to just do it through the website, you just go to your account settings, and here's where you'll see your API key. So here's the bright data GitHub page. has a lot of great information. And the good news is for Cloud Desktop, it's really easy to set up. You just need to click the Bright Data MCP extension button and it downloads a file to your computer. You double click into it and it opens up Cloud Desktop where you just add in the API key. So I'll just show that to you really quickly. So you click here to download the extension for Cloud Desktop. Then if you click the file, it opens up this page which is the newest way to add MCP service to Cloud Desktop. Just press install and then you just have to add in your Bright Data API key. when I first used the bright data MCP, I had to set up the web unlocker in the browser zone, but now it's all set up for you via the API. So again, you just put in your API key here and you're good to go. If you want to set up the bright data MCP server in any other supported client, you just do the standard JSON setup. And again, you just need the API key. Bottom line, AI web search is great, but not all AI web searches are created equal. Even the state-of-the-art frontier models are not able to get past most of the captures and all the little tricks that websites put up to stop bots from crawling. But Bright Data and its MCP server is able to break through. And that's why I believe it is the only web search MCP server you need. And I'm just waiting for all the other Frontier models to add real MCP support. I'm talking locally host MCPS via standard IO. That'll make ChatGBT agent so much better and open the doors for you to use the Bright Data MCP and all these other tools. So I'll put links to Bright Data's website and GitHub in the description below. I hope you found this video helpful or insightful. If you've used Bright Data or you have any feedback or comments, drop them in the comments below. Thank you guys for watching and have a great