I fully understand that Leo in the sidebar and search.brave.com might be handled by different teams ā but from the userās perspective, itās one experience.
And right now, thereās an important but invisible shift happening:
With web search enabled, Leo pulls live results ā via Braveās search index, yes, but often routed through third-party models (like OpenAI, Google, etc., depending on backend calls). Brave pays for these queries.
In private mode (no sources), Leo responds purely from model knowledge ā no external API cost, no live fetch. The information, however, is not (current), but based on the modelās training cutoff.
So thereās not just a tonal difference ā thereās a real trade-off: Freshness & cost ā Privacy & efficiency
Thatās why a simple hint under the input would make a big difference ā for example:
āUsing live results (current)ā āResponding from model knowledge (not current)ā
This wouldnāt just help users understand why the tone or depth changes.
It would let them choose consciously:
Do I need the latest data ā or a fast, private, cost-efficient answer?
And it would turn a confusing inconsistency into a meaningful, experiential difference ā something users can feel, understand, and use.
This small touch could reduce backend costs, build trust, and make the whole system feel not broken ā but intentionally layered.
Maybe itās just a tiny UX detail.
But for users like me who notice the shift, it would feel like Brave finally said:
āWe see you. And this? This is by design.ā
Just a thought ā but one that could save Brave money and earn real user trust.
Thanks for letting us know, Iāve passed it on to the team to see if itās something weāve run into before. BTW in future it may be better to open a dedicated thread about issues, I may not notice if itās buried in an unrelated thread but itās much easier to spot if you make a separate one
This is true, and I agree it can be really confusing at times. The tricky part is that Leo is married to the browser, whereas Search is designed to be useable in any browser. So itās difficult to build a bridge cleanly between the two
Thatās why a simple hint under the input ā like:
āUsing live resultsā / āResponding from model knowledgeā
ā would be so powerful
Hmm we do try to add a little note whenever Leo is using Search, though there isnāt anything for the opposite āResponding from model knowledgeā. I quite like the idea of adding a little note when no tools are used/if the search API isnāt used, which could help with user clarity and also make users more aware of possible hallucinations
Weāre currently overhauling the search/tool calling/model selection setup which may mean some models wonāt use search anymore, could be a good opportunity for something like this.
This is helpful though, itās unlikely that weāll solve the .search.brave.com ā Leo confusion soon as thatās a big project with a lot of moving parts. But we can try and make things a bit clearer on the Leo end. Thanks for the feedback!
@stephen
Thanks for your reply ā I appreciate the clarification.
Just to be clear: Iām not asking for a feature.
Iām describing a user experience thatās already happening ā one that others seem to notice too.
Whether itās āby designā or not, the tone, depth, and behavior of Leo do shift between modes. And when users feel that shift but donāt understand why, it doesnāt feel like choice ā it feels like confusion.
My suggestion wasnāt about changing the tech ā it was about naming the difference.
A simple label like ā(current)ā or ā(not current)ā wouldnāt force a redesign.
It would just help users make sense of what theyāre already experiencing.
And honestly?
Thatās not extra work.
Itās clarity.
Itās trust.
Itās design.
But either way ā thanks for listening.
Iāll keep testing.
