Google has quietly done something that sounds simple but is actually pretty hard: it’s started making Gemini remember who you are. The new Gemini personalisation features launching in the UK mark a real shift in how the assistant operates — moving from a stateless question-answering machine toward something that actually builds a model of you over time. That’s the pitch, anyway. But the details matter, and there’s more going on here than the announcement might suggest.
Why Google Is Doing This Now
The pressure on Google’s AI assistant business has been building for a while. OpenAI’s ChatGPT has Memory — it’s been out for over a year, first for Plus subscribers, then more broadly. Claude has Projects. Even smaller players have been shipping context-persistence features that let users avoid re-explaining themselves every single session.
Gemini, despite running on some of the most powerful underlying infrastructure on the planet, has felt weirdly amnesiac by comparison. You could have a brilliant conversation one day and come back the next to find Gemini had no idea who you were. That gap was becoming a genuine competitive liability.
The UK rollout is also interesting from a regulatory standpoint. The UK’s data protection regime, post-Brexit, has diverged somewhat from the EU’s GDPR framework — it’s still strict, but there’s been more openness to innovation from the ICO in recent years. That may be part of why Google is piloting these features here before a broader European push. It’s a reasonable testing ground.
There’s also a broader strategic context. Google has been integrating Gemini deeper into its product suite — Gmail, Docs, Search — and personalisation is the connective tissue that makes those integrations feel coherent rather than bolted-on. Without it, every product feels like a separate Gemini instance that doesn’t know what the others know. That’s a problem worth solving.
What the New Features Actually Do
Let’s get specific, because the announcement covers a few distinct capabilities that work differently and matter for different reasons.
Memory and Preferences
Gemini’s memory feature lets the assistant retain information across conversations. You tell it you’re a vegetarian, that you prefer concise responses, that you’re learning Spanish — and it holds onto that. Next time you open Gemini, it already knows. You don’t have to re-brief it like a new contractor every time you open a tab.
This sounds obvious. It is obvious. That it took this long is honestly a bit embarrassing for the industry as a whole, but here we are. The implementation matters though — Google says users have full control over what Gemini remembers and can delete specific memories or wipe everything. That’s the right call, both ethically and practically, and it’s something to watch closely as the feature matures.
Personalised Responses Based on Context
Beyond explicit memory, Gemini is now supposed to adjust its tone, depth, and framing based on patterns it picks up from how you interact with it. If you consistently push back on oversimplified answers, it should start giving you more detail. If you always ask follow-ups about practical implementation, it should start leading with that.
This is harder to verify than memory — it’s more of a background calibration than a visible feature. I’d treat this one as aspirational for now and see how it plays out in practice over weeks of real use.
Interests and Topics
Users can also explicitly set interests — telling Gemini that you care about finance, football, or machine learning — and it will use those to colour responses and proactively surface relevant information. Think of it as a lightweight profile that shapes what the assistant volunteers versus what it waits to be asked.
Here’s the thing: this is exactly what Google has been doing with Search personalisation for years. Bringing that logic into a conversational interface is a natural extension, but it also inherits the same risks — filter bubbles, confirmation bias, and the question of whether the AI is giving you what you need or just what it thinks you want.
What’s Included at a Glance
- Cross-session memory: Gemini retains facts you share across conversations
- Tone and style adaptation: Responses adjust to your communication preferences over time
- Interest profiles: Explicitly set topics that shape what Gemini surfaces
- User controls: View, edit, or delete what Gemini remembers about you
- UK availability: Rolling out to UK users on Gemini Advanced (Google One AI Premium subscribers) from late April 2026
How This Compares to the Competition
ChatGPT’s Memory feature has been live long enough that we have a real sense of how it works in practice. It’s genuinely useful — the kind of thing you miss when you use another tool. But it also has quirks: it sometimes remembers the wrong things, or holds onto outdated preferences you’ve since changed. OpenAI has iterated on the controls, but it’s still not perfect.
Claude’s Projects feature takes a different approach — rather than ambient memory, it’s more like scoped context windows. You create a project, attach files and notes, and Claude operates within that context. It’s more deliberate and arguably more predictable, though it requires more upfront effort from users.
Gemini’s approach seems to blend both — ambient learning plus explicit profile-setting. Whether that’s better or just more complex is an open question. The risk is that users end up confused about what Gemini knows and from where. Transparency about memory sources will be crucial.
For a deeper look at how Google’s Gemini has been evolving this year, our breakdown of the Gemini April drop covers the broader feature trajectory. And if you want to understand the competitive dynamics at play between the major AI platforms right now, the OpenAI-Microsoft renegotiation piece gives useful context on how the whole sector is repositioning.
What This Means for Different Users
For Everyday UK Users
If you use Gemini regularly for things like drafting emails, planning, or research, this is a genuine quality-of-life improvement. The friction of re-contextualising every conversation is real, and removing it makes the assistant feel more like a tool and less like a demo. The key is actually engaging with it — setting preferences, letting it learn — rather than just using it passively and expecting magic.
For Businesses and Professionals
The implications here are more nuanced. Personalisation works best when one person is using one account consistently. In shared or team contexts, it gets complicated fast. Google hasn’t said much about how this interacts with Workspace accounts or enterprise deployments. That’s a gap that will need addressing before this becomes a serious enterprise selling point.
For Privacy-Conscious Users
Google has committed to user controls, but the details of what data is stored, where, for how long, and how it interacts with broader Google advertising infrastructure aren’t fully spelled out yet. The recent conversation around AI privacy and PII detection is relevant background here — the industry is under real scrutiny on these questions, and Google won’t get a free pass just because the controls exist.
The ICO will almost certainly be watching how these features roll out. If something goes wrong — a memory leak, unexpected data use, a high-profile complaint — the regulatory response in the UK could be swift.
Frequently Asked Questions
Is Gemini personalisation available to all UK users?
Currently, the features are rolling out to Gemini Advanced subscribers in the UK, which requires a Google One AI Premium subscription (£19.99/month). Free tier users may get limited versions of these features later, but the full memory and personalisation suite appears to be a paid offering for now.
Can I control what Gemini remembers about me?
Yes — Google has built in controls that let you view what Gemini has stored, edit individual memories, or delete everything. You can also turn memory off entirely if you prefer. This mirrors the approach OpenAI took with ChatGPT Memory, which has become something of an industry standard for this type of feature.
How does this compare to ChatGPT’s Memory feature?
Both store facts across sessions and give users control over deletion. Gemini adds explicit interest-setting on top of ambient learning, which ChatGPT doesn’t do in the same way. ChatGPT has a longer track record with memory, so it’s arguably more battle-tested — but Gemini’s integration with Google’s broader product suite could give it a meaningful edge once the feature matures.
When will this come to other countries?
Google hasn’t confirmed a specific timeline for broader rollout. The UK launch likely serves as a regulatory and technical pilot. An EU rollout would require additional GDPR compliance work, and US availability may depend on ongoing AI regulatory discussions. Realistically, a wider rollout in H2 2026 seems plausible, but nothing official has been announced.
The honest truth is that personalisation is table stakes now — every serious AI assistant needs it. What separates the good implementations from the mediocre ones is the quality of the controls, the accuracy of what gets remembered, and whether users actually trust it enough to engage. Google has the infrastructure and the data advantage. Whether Gemini can translate that into an assistant people genuinely rely on is still being written. I wouldn’t count them out.
