The Google Trends Glitch Nobody’s Talking About

**Chris** (0:02)
Okay, guys, there's something that we need to talk about. It's controversial.

**Dave** (0:10)
Let me pull up your X account real quick, because I think I know where you're going with this. Very controversial.

**Chris** (0:16)
Why does every single person on X, if you say anything remotely negative about a company or stock, behave as if they're the founder of the company? These are just investment thesis. That's all it is.

**Jordan** (0:33)
And just because, by the way, just because you're saying something negative doesn't mean that you're negative on the company, but you have to be realistic about all the possibilities.

**Chris** (0:43)
Exactly. And by the way, I did this thread. I worked really hard on it. I spent my entire Sunday two weeks ago.

**Jordan** (0:52)
I think it was a fantastic thread, by the way.

**Chris** (0:54)
Thank you. Thank you. I did work hard on it.
And the thread is something that's been really bugging me because one thing that we've relied on for the past 18 years, almost equivalently to anything else, it's almost as important as the comment analysis that we do on TikTok and other social channels, is search traffic. I've always said that search traffic, G Trends, Google Trends, is the best data source in the world for investors, and it's completely free. But you have to learn how to use it. You can't just like go on there and pull charts because you'll get confused, you'll misconstrue what you think the data is showing you. But our community has become experts at extracting alpha from search traffic. Okay? And over the course of this summer, our community has started to notice wild anomalies in Google Trends data. Things that I've never seen in 18 years. I probably spent more time on Google Trends than anyone in the world. And part of the reason why I got so amped up on Lululemon is because like 14 different product categories at Lululemon were hitting all-time highs in the middle of the summertime on Google Trends. And these are seasonal items that normally peak on Black Friday or the holiday week of Christmas. I've never seen anything like it before, ever, in 18 years. And we started speculating something is off, something is really off, okay?

**Dave** (2:35)
These are some examples from your tweet. This is Lululemon shorts. And you see this, like, all-time high, like, eclipsing even the peaks, the annual peaks.
And it just said something has to be wrong. There's something going on with the data. What could have changed with Google Trends, not just for this search term, but for a bunch of other search terms?

**Chris** (3:02)
And Dave, Jordan, remember earlier this year when we talked about, you know, Google has to be losing one or two or three points of search traffic to Chat GPT based on our estimations.
And then they came out with earnings and didn't. And then they came out with earnings again last quarter when we were like, we think they might be down three or four points of search traffic, which is pretty big. And they actually grew search traffic and grew search revenue. And now we're seeing these anomalies on G Trends. And I start doing research and speaking to people, realizing that a lot of people working with AI, AI scientists that I spoke to, said that they are now tapping into Google search at ChatGPT and other AIs, other LLMs. Now, they used to, ChatGPT used to rely mostly on Bing, but this summer they started expanding well beyond Bing. And Jordan, you might even have some experience with this, because you're working on some apps that do a lot of search. So we looked into the data agreements that Google had, both on AdWords and Google Trends. And what was astonishing is that they don't mention anything about external third party AI being included or excluded or filtered out. So I think on AdWords, I think it was AdWords or maybe Google Trends, they mention that they do not include the Google AI summary data, okay, because they have full control over that. But what they don't have control are the 1 billion people using ChatGPT. And in some situations, when you put a query into ChatGPT, it generates not just one search, but it could generate dozens or theoretically, if you're doing deep research, hundreds of search queries that not only search, but then will do deep search and actually click on links that could be advertiser links, okay? And nobody was talking about it.

**Jordan** (5:16)
Yes, so we don't really know. I'd hate to point at things that we don't 100% know. But yeah, I think that anytime you're working with AI now, especially some of the thinking models or the deep research models, we know that they've got access to the Internet. And they use one method or another to get that information. Generally, the easiest way is a search engine, right? Because you can use natural language that you generate from the LLM to then interact with the search engine. So yeah, it could be one query. It could be tens. It could be, like you said, it could be 100 queries. We don't know. And it probably depends on the context and the things that you're doing with the LLM. But yeah, we're pretty sure now that these LLMs are talking to Google.