Search Personalisation

Search personalisation is when a search engine or AI assistant adjusts what it shows based on who is asking and in what situation. Location and language are the most common signals, but recent searches, clicks and, in AI assistants, saved memory can also shape the results. MLforSEO treats it as part of Session Context: context does not reset between searches.

How it works

Personalisation works at two levels, which MLforSEO calls micro and macro:

• Individual (micro): the system uses what this user has searched, viewed and engaged with in the session to refine later results, for example promoting fresh results when earlier ones went unclicked.
• Cohort (macro): behaviour across groups of similar users can lift pages they favour and quietly demote ones they keep skipping. See Implicit User Feedback.

Google says its results use location, language and recent search activity, and that signed-in users can turn off personal results. AI search goes further upstream: ChatGPT, Gemini and Google’s AI Mode can use conversation history and memory to decide which sub-queries to generate in Query Fan-Out, so two people asking the same question may trigger different searches.

How it’s used in SEO and marketing

• Treating rank tracking as an approximation: tracked positions come from neutral settings, while real users see variations.
• Mapping query to landing page to next page paths by combining Search Console and GA4, to see real journeys.
• Planning content for the follow-up questions people ask within a session, as well as the first query.
• Covering the different personas and contexts that generate different fan-out queries.
• Local SEO, where location is the strongest personalisation signal.

Example

Two people search “running shoes”. One has just searched “plantar fasciitis” and the other “marathon training plan”. The first query is identical, but session context makes different results more relevant to each, and an AI assistant with memory might fan out into “shoes for heel pain” for one and “carbon plate racing shoes” for the other.

Limitations and things to watch

• In 2018 Google said personalisation in core web rankings is lightly applied, mostly location and immediate context from a prior search, and that many differences come from location, language and platform instead. Avoid over-attributing ranking changes to it.
• Privacy settings, signed-out users and regulation all limit how much can be personalised.
• You cannot see another user’s personalised AI answer, so measure trends across many prompts and contexts.

Learn it on MLforSEO

• What is Query Context and Session Context in SEO?
• What is Search Query Sequence and Query Path?
• What Is Implicit User Feedback in SEO? How Behavioural Signals Shape Search Rankings
• Semantic ML-enabled Keyword Research