Hi 👋🏻
I came across a couple of posts on social media recently that highlighted the need for a discussion on the topic of results personalisation - on both traditional and AI search systems.
One is this funny example of (what we can assume based on Natalie's caption from her LinkedIn post is) a highly personalised selection of PAA questions, related to a query she submitted to Google.
Most of the other posts I’ve seen on LinkedIn talk about result personalisation in AI search, but they lack any real contrast with traditional search. They rarely acknowledge how personalised systems like Google Search already are, instead fixating on user personas in AI search or the challenges of tracking rankings from AI-driven traffic.
Yes - personas are important, and yes - AI search systems' degree of personalisation makes it harder to anticipate visibility. I absolutely do not disagree with that, just simply want to highlight that these challenges exist for traditional search systems, too.
Let's explore.
how are search results personalised on google
Looking at the example of the personalised PAA box, this patent explains how questions for PAA boxes are selected.
PAA questions are generated from query logs + click data + topic sets per result + question/answer quality signals, so they’re deeply shaped by the specific query, the results shown, how users historically interacted with those results, and the dominant ways people semantically describe that topic.
But that's not all 👀 Another Google patent Generating Query Variants Using A Trained Generative Model explains how Google generates query variants for search features like People Also Search For and People Also Ask.
When you see PAA questions that feel uncannily relevant to you, it's not just because they come from aggregate user behavior—Google is also actively factoring in:
- Your search history - Previous queries you've submitted
- Your location - Where you are or frequently search from
- Your inferred intent - Based on your calendar, emails, and browsing
- Temporal context - Time of day, day of week, time of year
This makes PAA not just crowdsourced, but personalized ✨ crowdsourcing.
Google's results are a lot more tailored to you than you might realise — your history, your device, your intent, and even your scroll behaviour.
1. User & Session Context
Google doesn’t treat each search as a blank slate. It looks at:
- Your past searches & clicks - Boosts topics you’ve shown interest in and deprioritises things you’ve already seen and ignored
- Your location - City searches (e.g. “London”) adapt to where you are and local tickets, tours, and offers change based on geography
- Your device - Mobile vs desktop can see different rankings, and results are optimised for how people typically behave on each device
- Your session behaviour - adapts results to things you clicked earlier in the session (or search sequence) and considers similar queries you’ve run recently
All of this is used to prioritise “more of what looks relevant to you right now”.
2. User Behaviour & Implicit Feedback
Your behaviour on the results and pages becomes a ranking signal.
- Clicks & CTR – How often a result gets chosen
- Dwell time – How long you stay before bouncing back
- SERP interactions – Scrolling, pausing, “reading” segments
- Mouse hover – Hovering before clicking helps predict interest
- Query reformulation – Changing your query signals that previous results missed the mark
All of this helps Google reorder and refine what you see next.
3. Entity Understanding & Knowledge Graphs
Google doesn’t just see keywords — it sees entities (people, places, things).
- Disambiguation helps tell "Apple" from "apple"
- Richer, personalised answers - Knowledge panels, carousels, and voice results shaped by your history, device, and prior interactions
4. Search Intent & User Personas
Google tries to answer the question “What are you really trying to do?” by:
- Intent classification - Informational, transactional, navigational, local, etc., but also refined using location, device, and past user behaviour
- Content alignment
- User personas (inferred) - Queries with words like “luxury” vs “cheap” can signal different segments, shifting results and content types accordingly to the inferred user profile
5. Information Gain & Query Augmentation
Google wants to avoid showing you the same info over and over.
- Information Gain (IG) - Measures how much new value a page adds vs what you’ve already seen and helps filter out repetitive results in a session
- Query augmentation - Expands your query using recognised entities and attributes. These refinements are tuned by behaviour signals like long clicks and CTR
Google, the librarian 📚
Google's search personalization system acts like a highly attentive, experienced librarian. When you walk in and ask for a book (your query), the librarian doesn't just match your exact words to titles (traditional search). Instead, they subtly tailor the results based on everything they know about you—where you live, what genres you read last week, which authors you usually linger over, and what others who share your interests found valuable (location, session context, implicit feedback, and similar user data). They also quickly clarify ambiguous requests (entity disambiguation) and constantly suggest related reading material (query augmentations) that offer genuinely new information (information gain), ensuring you leave satisfied with a search experience tailored uniquely to you.
Are search results more personalised on AI search than traditional search?
Modern search, regardless of whether it results in traditional links or AI-generated summaries, is already highly personalized and relies heavily on AI/machine learning (ML) models and user data. 🌟
The key difference is that despite the hundreds of different SERP features, traditional search is a lot more rigid in presenting results to the user. AI Search challenges this and systematically reviews and selects results to present to the user. New challenges appear:
- Being used as a source doesn't equate to being cited
- Queries are longer on AI search (more explicit context to work with)
- Tracking is not scalable
- Retrieval mechanism is fundamentally different
- Query fan out introduces a new layer of IR and can skew search intent
This simply means, AI search has:
- More signals (conversational history, prompts, feedback)
- More flexibility (it can change the content of the answer, not just rank links)
- More room for adaptation (tone, depth, format, examples, local context)
Watch this video for more information about SERP Snippets and ML-enabled SERP analysis 💜
Side note, we've started publishing lesson snippets on our YouTube channel, covering concepts like: information gain, EAV model, implicit user feedback and user behaviour, query augmentation, query paths, and more.
A lot more is on the way - subscribe if you'd like to see more of this content in your feed. 💜
LEarn more about how google personalises traditional search results in our Featured course 🌟
Let us teach you fundamental concepts about semantic query understanding and how traditional search engines have systematically personalised search results for years, combining multiple explicit and implicit signals. We cover dozens of patents to show you exactly what you need at the keyword research phase to ensure you create meaningful content.
SEMANTIC AI-powered keyword research
Two new modules are in development, to be released in November. ✨
|
|
|
This course features 🔥 10+ hours of content already, with a ton of practical exercises, tools, checklists, scripts, and more.
Heads-up 💸 the price of this course will increase when the new modules drop in November —lock in now.
✨ Use code COMMUNITY30 for 30% off at checkout - that's our little thank you for being part of the MLforSEO community
80+ forward-thinking marketers are already taking our courses 💜
recent discussions from our slack community 💬
Here's some of the recent discussions there:
Or check out a small sample of the resources shared:
Join 670+ AI/ML-interested marketers on our Slack community to stay up to date with discussions on AI/ML automation in SEO and marketing.
Happy learning! ✨
Lazarina
