Joshua Squires shared one of the most interesting AI Overview leaks and for some reason it was mostly ignored by the SEO industry. I’d like to draw your attention to it today because it provides two key details framing AI Overviews as an implementation of Google’s Dialogflow agentic framework which is backed up with an […]| DEJAN
While traditional keyword research tools provide valuable data, they often fall short in discovering truly novel or long-tail search query variations that a business might not yet rank for, or even be aware of. This is where our query fan-out model comes in. Using advanced language models to generate a vast array of related search […]| DEJAN
Some people use AI to speed up the process of getting their ideas and message out. Others use it to polish up their language which I think is really cool use of AI, especially if they’re not a native speaker. But there’s also the hordes of mindless AI slop generators masquerading as meaningful human engagement, […]| DEJAN
Introduction “Help Me Write” is Google Chrome’s AI-powered writing assistant designed to help users create short-form content directly within their web browser. Launched with Chrome version 121, this feature leverages artificial intelligence to generate text suggestions based on user prompts and the context of the webpage you’re viewing. How Does Help Me Write Work? The […]| DEJAN
Stop Guessing, Start Optimizing. Introducing Tree Walker for the New Era of AI Search The digital marketing landscape is in the midst of a seismic shift. With the rise of AI-powered search engines and generative experiences, the old rules of SEO are being rewritten. Marketers and content strategists are asking the same urgent question: “How […]| DEJAN
This test is designed to show whether Open AI’s browsing tool does a better job at supplying their model GPT-5 with grounding context from a page with schema. We took the exact HTML from the original experiment here, stripped off the “experiment” from the title and header and uploaded here and here and then ran […]| DEJAN
Chrome is about to give all websites a voice through a built-in version of Gemini. Your visitors will have completely private chats with it. No external API calls to Google’s servers and once loaded you can even switch off the internet – it will still work! What will they talk about? The Silent Web is […]| DEJAN
I decoded Chrome’s internal semantic search, found the exact chunking mechanism, embedding logic and am now able to browse, search and cluster my own search history through decoded vector embeddings. This is an in-depth technical analysis of Chrome’s history embeddings system based on Chromium source code and official Google documentation. Google Chrome has implemented a […]| DEJAN
Modern search engines are still fundamentally based on information retrieval, but they’re now powered by two distinct layers of AI augmentation: a strategic Agentic Layer and a user-facing Interpretative Layer. The Agentic Layer The Agentic Layer acts as the engine’s strategic decision-maker. This layer, which involves multiple systems and models, determines how to best fulfill […]| DEJAN
The two pillars of AI optimization are model understanding and control with well-established analogues in the machine learning industry called mechanistic interpretability and model steering. SEO Machine Learning Understanding Mechanistic Interpretability Control Model Steering Mechanistic Interpretability A subfield of AI interpretability that aims to understand neural networks at the level of individual components (neurons, attention […]| dejan.ai
The Temperature parameter is a crucial setting used in generative AI models, such as large language models (LLMs), to influence the randomness and perceived creativity of the generated output. It directly affects the probability distribution of potential next words. Understanding the Basics What the Temperature Value Does In Practical Terms Using the sentence “The cat sat on […]| dejan.ai
GEO stands for Generative Engine Optimisation, an acronym easily confused with, the well-established “geo-” prefix commonly associated with Geosciences. What is a ‘Generative Engine’? Generative engine is recently made up term by the marketing community in an attempt to rename Chatbots, more recently known as AI Assistants including ChatGPT, Claude, Grok, Gemini and Perplexity. Basically […]| dejan.ai
John Botman For nearly two centuries, journalism operated under the assumption that truth mattered, stories should be original, and humans should write things for other humans to read. Quaint, right? We trusted journalists—those quirky creatures who collected facts, verified sources, and occasionally spelled words correctly—to give us nuanced, insightful accounts of the world. Oh, how […]| dejan.ai
In our previous post, Training a Query Fan-Out Model, we demonstrated how to generate millions of high-quality query reformulations without human labelling, by navigating the embedding space between a seed query and its target document and then decoding each intermediate vector back into text using a trained query decoder. That decoder’s success critically depends on […]| dejan.ai
Prompt Engineer’s Guide to Gemini Schemas| dejan.ai
DEJAN is a marketing agency specialising in medium-to-large brands and eCommerce websites. We use machine learning to transform data analysis, strategy design and campaign execution. Book a conference call with our senior strategy team to discuss your project in detail. The consultation is free and highly constructive. Data. Discovery. Testing. We design and deploy sharp, […]| dejan.ai
Google discovered how to generate millions of high-quality query reformulations without human input by literally traversing the mathematical space between queries and their target documents. Here’s How it Works This generated 863,307 training examples for a query suggestion model (qsT5) that outperforms all existing baselines. Query Decoder + Latent Space Traversal Step 1: Build a […]| dejan.ai
Google’s embedder uses dot product between normalized vectors which is computationally more efficient but mathematically equivalent to cosine similarity. How Googler’s work and think internally typically aligns with their open source code (Gemini -> Gemma) and Chrome is no exception. It’s why I look there for answers and clarity on Google’s machine learning approaches. After […]| dejan.ai
Generalist, Open‑Set Classification for Any Label Taxonomy We’ve developed a search query classifier that takes any list of labels you hand it at inference time and tells you which ones match each search query. No retraining, ever. Just swap in new labels as they appear. Old workflow Pain New workflow Build + label data + retrain […]| dejan.ai
If Marie Haynes, Barry Schwartz or Cindy Krum had written an article declaring SEO dead and proposing we rebrand our industry you’d seriously consider it. Wouldn’t you? What about Zach Cohen and Seema Amble? I don’t know either. Looked them up just now. Two VC people with insignificant footprint or long-term interest in SEO, Machine […]| dejan.ai
Embedding Methods Evaluation: Results, Key Findings, and a Surprising Insight On June 6, 2025, we ran a comprehensive evaluation comparing four different embedding methods—regular, binary, mrl, and mrl_binary—on a dataset of paired sentences. The goal was to measure each method’s speed, storage footprint, similarity quality, and accuracy against a ground-truth of sentence pairs. Below, we […]| dejan.ai
As a technical SEO, you might be diving into machine learning (ML) to understand how tools like Google’s Gemini process text. One foundational concept is subword tokenization—breaking words into smaller pieces called “tokens.” While tokens themselves are context-agnostic (they don’t consider surrounding words), they do carry an inherent bias: each token’s likelihood reflects how prominent […]| dejan.ai
1. ULM128M 2. LLMIT1B 3. GEMINI2_NANOV2 4. GEMINI2_NANOV2_EE2Q 5. GEMINI_XS 6. GEMINI_XS_DRAFTER_6LAYER_CAUSAL_USM_700M_RESIDUAL 7. GEMINI_XS_LUSM_700M_RESIDUAL_BOTTOM15 8. GEMINI2_NANOV2_EE12Q 9. GEMINI2_NANOV2_EE2_LUSM_700M 10. GEMINI2_NANOV2_CAUSAL_700M 11. GEMINI2_NANOV2_EE20_CAUSAL_LUSM_700M 12. GEMINI_XL_DRAFTER_24LAYER 13. GEMINI_XS_FA1 14. GEMMA2_8B 15. GEMMA2_7B 16. GEMMA2_2B 17. GEMMA3_1B 18. GEMMA3_4B 19. GEMMA3_12B 20. GEMMA3_27B 21. STABLELM_4E1T_3B_PHI_2_TF_LITE| dejan.ai
I recently stumbled upon a fascinating aspect of how Google’s AI Mode (powered by a custom Gemini model) interacts with the internet. I ran a simple test, and the results suggest that instead of performing truly live fetches for all URLs, the AI Mode relies on Google’s existing index or a cached version of the […]| dejan.ai
Using the same tech behind AI Rank, we prompted Google’s latest Gemini 2.5 Pro model with search grounding enabled in the API request. A total of 10,000 prompts were collected and analysed to determine the grounding status of the prompt. The resulting data was then used to train a replica of Google’s internal classifier which […]| dejan.ai
Activation Logging and Internal State Monitoring One foundational approach is activation logging, which involves recording the internal activations (neuron outputs, attention patterns, etc.) of a model during its forward pass. By inspecting these activations, researchers can identify which parts of the network are highly active or contributing to a given output. Many open-source transformer models […]| dejan.ai
The “Probability Threshold for Top-p (Nucleus) Sampling” is a parameter used in generative AI models, like large language models (LLMs), to control the randomness and creativity of the output text. Here’s a breakdown of what it does: Understanding the Basics What the Threshold Value Does In Practical Terms Imagine you’re asking the model to complete […]| dejan.ai
Google’s Gemini models are designed to provide users with accurate, timely, and trustworthy responses. A key innovation in this process is grounding, the ability to enhance model responses by anchoring them to up-to-date information from Google Search. However, not every query benefits from grounding, and Google has implemented a smart mechanism to decide when to […]| dejan.ai
It’s an exciting time to be in SEO. Honestly, it feels like 2006 all over again – a period of rapid change, innovation, and frankly, a whole lot of fun. For a while there, things had gotten a little… predictable. Technical SEO, keyword research, competitor analysis, link building, schema… it was all necessary, of course, […]| dejan.ai
UPDATE: Addressing guardrails, hallucinations and context size. 1. People are reporting difficulties in recreating the output due to guardrails and hallucinations. 2. Snippet context sometimes grows to several chunks. Guardrails Google attempts (and in many cases) succeeds at blocking these requests, but it does so in a very clumsy way so that we actually get […]| dejan.ai
