My prompt injections on Google’s Gemini model provoked a new leak of its internal system information.

Today, I will reveal some of the data contained in the upcast_info block used by Google to send instructions to Gemini. And the potential implications for Search.

This is the most beautiful AI research I’ve read this year! :D

I'm Elie Berreby, the author and I'm sharing first on HN. Using standard UTM parameters (utm_source, etc) on internal links is a dangerous anti-pattern.

Architecturally, it creates two massive problems: it forces web crawlers into infinite, redundant loops that destroy your crawl budget, and it permanently fractures your analytics by overwriting the user's original session acquisition source in tools like GA4.

The modern fix is to completely decouple your routing and presentation layers from your analytics.

Pass structured payloads using HTML5 data-* attributes and pick them up via event listeners. It keeps URLs clean and makes your telemetry immune to CSS refactors.

Apologies for the image loading issues with Chromium based web browsers. I had strong (and incorrect) settings in place because I forgot 1 character somewhere in the config

That was to protect from eavesdropping and avoid modifications by man-in-the-middle attacks. I corrected + purged all cache.

Issue now resolved, just force refresh :)

You can email me at: rxnx8obtw@mozmail.com

Elie Berreby, head of SEO & AI search at Adorama, explains how e-commerce brands can improve their visibility in AI Search answers by using agentic commerce.

Google wrote that "structured data won't make your site rank better"

Let me be clear: the official Google Search statement you read is wrong.

A good structured data strategy will positively impact organic rankings.

In this 3,500+ words article, I tried to explain ideas and concepts in detail:

1) The role of structured data in organic search

2) How structured data adds value for humans

3) How structured data helps search engines match content with user intent

4) Why I believe Google downplays the role of structured data in SEO

5) The impact of Click-Through Rates on organic rankings

6) How visual layout influences human behavior

7) How structured data reduces our cognitive load

8) Why using good structured data will increase CTR

9) How a CTR increase can create a positive divergence with the Click-Through Rate Google expects

10) How positive divergences push Google to re-evaluate rankings

I did my best to clarify things but I'm exhausted. If you spot typos or plain mistakes, PLEASE let me know. I always WELCOME constructive criticism!

I co-authored this short article for Malwarebytes.

Ask me anything :)

The life of many creators today?

AI answers, no clicks, no revenue.

My short post about the limitations of cosine similarity used for semantic matching in organic search.

My simplified guide with 30+ slides to help web publishers and creators understand the basics of crawling, rendering, indexing and conversion rate optimization (CRO)...