RT @abondance_com: ChatGPT bouscule Google : un nombre croissant d'utilisateurs l’adoptent comme moteur de recherche, avec des questions pl….

RT @metehan777: 🔍 I analyzed ChatGPT's actual configuration files (you can check yourself: any chat → View Source → search "rerank"). Found….

What do you want to know?.

RT @natzir9: The beginning of the end for prompt injection?. If you were hiding instructions in HTML. the party's over. ChatGPT’s agent….

Trying to get a free backlink from . Not indexed by GG but still not forbidden in robots.txt ==> do you think it works ?. Save time : Grok has nofollow, Claude and Perplexity are in disallow. .

Google has patched the flaw that made it possible to access the list of current experiments on their interfaces. 🥲. So we won't be getting any more updates to the list, which I'm preserving here for the record:

And plenty of other things to dig into… Here’s the URL to download the file for those interested. 85,000 lines of code, no less!.

A list of third-party sites with which Google shares data, including Facebook, Telegram, Discord, Instagram, Microsoft, Pinterest, Tencent.

Another list of domains receiving special treatment, this time for a video game experiment. Notice that search engines like to write domain names backwards, in the form extension.domain.subdomain, it takes up less space to store that way…

Some sites are manually excluded from certain experiments, including Amazon, Pornhub, Wikipedia, Reddit or PayPal.

Numerous events are tracked, such as abandoned navigations from Google Search, timing metrics, or predictors of clicks on links and user interactions on visited websites. Your navigation after a Google search is tracked, but we already knew that ^^

Yet another filtering system, this time related to Google Lens, along with an educational layer for .edu domains. You can find both positive and negative filters in it

More descriptive lines to identify product pages that are candidates for a couponing system being tested, for a limited list of domain names

Here’s a list of words that Google uses as a blocklist on product pages, in order to exclude them from certain experiments. Funny to see that not everything is handled by algorithms, engineers sometimes rely on manual descriptive files.

I just came across something that definitely shouldn’t be online, related to Chrome. It’s a file containing numerous browser experiments and configurations… .Here’s a thread with some funny things you can find in it. For once, I’ll share the file URL at the end ;). ⬇️ 🧵.

RT @dejanseo: This tool allows you to see if a person, brand, product or service is a known entity in Google's knowledge graph. https://t.c….

RT @abondance_com: Top 10 Google ≠ Top des citations IA 🤯 .12% seulement des liens cités par ChatGPT & co figurent dans les SERPs classiqu….

Ever wondered which sites Google thinks are the best sources? 🤔 . I scraped almost 15k domains from Google's new Source Preferences feature. Searchable list now live - find out if your favorite sites made the cut!.

AI Mode only giving me a list of 10 blue links 💕

By the way, we’ve recovered hundreds of sensitive conversations. Some are really disturbing :

🔍 Analyzed 87k public ChatGPT chats:. First msg: 8 words (vs Google's 3.5).60% have 5+ exchanges.Only 48% seek facts.10% contain code. It's not search 2.0, it's something new entirely.Full study :