Jesse Cunningham Examines Changes Affecting Claude Website Builders

0:00Things are about to get weird as this rolls out. So, Claude is watermarking its AI. Let's talk about it. People are saying you can avoid the watermarking. How does Claude even do it? What does it mean for people like me and you who, I mean me, but I don't know about you, I almost exclusively build with Claude right now, my websites. In fact, on that computer right there, I'm building one.

0:19And it's going to take like 30 minutes. It's awesome. I really love it. It's a great way to build websites. But, is it going to kill our websites, right? So, someone commented, "Does it matter that all Claude code is now invisibly watermarked, so everyone knows you used it?" And someone else said, "I got the same question. I would love to see Jesse make a video." So, let's do just that. I think it's going to be bad for those who don't know what to expect. And I think there's ways around this, too. Now, we need to understand this one right here.

0:45Nothing is added to the text. Now, this is the official Anthropic press release thing. Nothing is added to the text, and there's no hidden characters. So, at first, I'm like, "Man, there's probably just hidden characters. We'll push it to notepad, whatever text you do, push it to notepad or Google Docs, strip it all out. We're good." That's not the case. That's a big problem. It's a big problem. They're a bit more sophisticated than that. And here's why it's a problem. So, if Google can tell it's AI, does my site derank, does it die? So, let's say AI writes the site,

1:14Google spots the watermark, that's a problem, and your rankings drop. No calls, no rent. So, what I do is rank and rent often, right? And if Google's not going to rank my websites, we got a problem. We got a big problem. And we build these websites almost with 100% AI. In fact, that's what I'm doing right now, 100% AI. But, here's what Claude actually announced. So, in this month, right, Anthropic said Claude will start adding the invisible watermark to the text.

1:41It's a hidden pattern woven into the words. Pretty intense. You cannot see it, and it does not change how the writing reads. The reason, right, was not to snitch to Google, but it's to comply with the new European transparency law, the EU AI Act. And this is like a legal tick box, like a box to tick applied worldwide. So, if Anthropic is doing this and they're the first ones to do this, do you think OpenAI and GBT, right, their product, they're going to follow suit? I'd be willing to bet everyone's going to follow suit. So, we have to get in front

2:11of this thing. Now, we're going to get into the weeds, but realize this whole thing with Anthropic is a potential extinction event. That Uh what does that even mean? Well, an extinction event is where you have a strategy and it no longer works. Like 3 or 4 years ago, the HCU, the helpful content update on Google, wiped out niche blocks, a big problem. Anthropic's thing could do that, too, if it takes hold. If you want to stay ahead of the curve, there are tactics to do so and you have to stay up to date. Join a community. The Rank Expander Academy,

2:37I'll have a link in the description, is where I spend most of my day. All right, let's get specific here. So, how does this actually work? What is Anthropic doing? We have three buckets here. So, what it's doing is messing with text, images, and code. So, Anthropic, you know, Claude, it can do text really well. It can do code even better. And images, meh. And we're not really concerned about images, but we'll talk about it, too. So, text is marked. So, the words Claude writes carry the hidden pattern and it's not like injected into the words. We'll talk about it. Images

3:08have a label. You can strip this out with a tool like XnConvert. I'm not concerned. And then code barely has any water marking because code has to be exact. So, what we're going to focus on is text because, well, if a website is all text, for the most part, we have to worry about it. So, it nudges word choices. This is how it works, nothing else. So, as Claude writes, it is constantly picking the next word for a few good options. The watermark leans those tiny coin flips into a secret pattern using a method from Google DeepMind called SynthID. So, to the

3:40reader, it feels normal, natural, no weird characters, no visible marks, nothing. To the pattern, the word choices line up to a hidden signal that only a special key can read. So, how AI language models work is they have n-grams and n-grams is a prediction model. So if you have the cow, so two words the cow, what does it do next? Well, the prediction model can figure out like a cabillion different next words, but their secret key, Anthropic has now a secret key, right? A secret way of doing this can show other people that yes, this was by Anthropic because of the word choices

4:15that it picked. And if you have no key, you have no way to see it. So without the key, it looks like regular text. But with Anthropic's key, right? You can see, all right, this was definitely 100% made by Claude. With their key, and they say they're going to have a public, maybe not public, but they are going to have a detection tool coming soon. And if that becomes public, anyone's going to be able to go to any website and if there's API, scrape websites and see, yeah, that's all AI. That's a big problem. I can tell you why it's a problem. It's

4:44already happening in some industries. Um, people are getting penalized. Google and Claude are different companies. Thank goodness. So for this to be a problem on SERPs, on Google SERPs, they have to cooperate somewhat, but I think there's a bigger problem for Google, right? Checking the whole web, the whole internet is expensive. So even if Google had the key tomorrow, it crawls hundreds of billions of pages and running a detection check on every single one for a signal it is never asked for is a massive extra cost. Do you think they're going to spend money, a for-profit company? Maybe. Maybe. Google rewards

5:19helpful content, however, it is made. And here's an interesting thing. If you're going to start penalizing AI content, realize that there is, quote unquote, real companies across the whole web, right? Some people claim when they watch my videos, AI slop, and they freak out and they and they write mean comments because AI slop. Which what is AI slop? Like is it any AI that you're writing? What is the distinction? Is it Is it value-based? Is it your perception? Your Your Your subjectivity?

5:47That's the problem because there's, quote unquote, companies that you know that are using AI right now. And if you just do a blanket statement of AI bad, and they have a key, and Google starts to smack them down, I don't think that's going to work. Now, like most things in SEO and online, when there is money to be had, it's a cat and mouse game. It's been like this forever.

6:07You have people like me who are SEOs who want to get on top of the algorithm and beat the competition because ultimately Google is a big competition. Whoever shows up number one is going to make the most money. It's a competition, and it's a cat and mouse between the arbiters of the SERPs, the people who dictate who's on top, and the people who want to be on top, right? So, the cat are the detectors, every new tool that claims to spot AI text, such as this one, and the mouse are the creators. And oftentimes, especially with text, the mouse is much

6:36faster than the cat. But the cat is starting to catch up, right? The cat is starting to catch up. Big problem. All right. So, the mark lives in the words, so the words break it. So, how can we beat this thing? Well, listen. If it's it's it's not embedded in the words, if it's literally the words, then you can spin the words.

6:56There's been spinners forever. How hard is that? And then all of a sudden, it's not written by Claude, right? The watermark will be broken immediately if you spin it. And there's other ways to do this we can talk about, but there's still problem here. There's still problem. So, you can edit and rewrite, so the pattern sits in the word choices, so a real editing pass disturbs it on its own. You can add 10% human, right?

7:20The one real detail per page you already add from the blueprint mixes the words up further. Um code, don't worry about that. So, how are people doing this right now? They're literally just spinning articles, but it's still, if you spin an article, it's still detected as typically 100% AI. Other people are adding ad-lib. They're talking to their computer. They're using dictation. That can work, too. There's other ways to do it that I can't talk about on YouTube because it's a cat-and-mouse game, right? And sometimes the the cat is watching channels like mine. So, we're not going to talk about it here.

7:57But, if you can break the pattern that is dictated by Anthropic, then you break the key. You break the code. Now, for me, this is what I think, right? I'll give you my take. Right now, nothing changes. I think I don't care. I mean, quite quite literally, I don't care. I think it's a nothingburger. I think it's going to be the policy of all AI to do this, and I think companies like Google cannot just have a big sweep of all AI is bad. I mean, they have used AI in their overviews, their AI mode.

8:28They're using AI throughout. You can't just say AI is bad and and banhammer websites for using it. So, nothing changes for me. Um I'm not losing a minute's sleep over it. Two things that would have to happen together, right? Anthropic hands over detection on a large scale, and Google flips its policy. Right? Google just wants helpful content. And your mode was never hiding the AI, at least in my opinion. We have genuinely useful pages. We have real services behind what we do, and we have track calls, not rankings. We have all the things that Google is looking for.

9:04So, just appease the algorithm. Make sure the the consumer, when they come onto your website, they're not bouncing off immediately. And that's the true north. Make sure your consumer is happy. How long do they stay on your website? Listen, people in this community right here, you your mind would be blown if you're not in here already. And by the way, it's a 7-day free trial. You have nothing to lose. Come check it out. What the heck? But, you have people in here who have 100% AI content websites that are ranking position number 1, 2, 3, 4, 5, 6 with six other different websites

9:34for a query they want to rank for. All roads lead to Rome. It's insane. And that just happened recently. Very exciting stuff. You can learn how to make money with AI. You can build a microsite empire. You can do all sorts of things. And typically, it's with quality, right? But this whole Anthropic thing right now, to me, is a nothingburger, but it could be a glimpse into the future.