Jon von Tetzchner, the co-founder and CEO of Vivaldi, didn’t bother with the usual corporate-speak. In a blunt, almost manifesto-like blog post titled “Keep browsing human,” he argued that stuffing AI assistants into browsers — turning the omnibox into an “assistant prompt” — risks turning curiosity into passive consumption. “We choose humans over hype, and we will not turn the joy of exploring into inactive spectatorship,” von Tetzchner wrote. “Without exploration, the web becomes far less interesting. Our curiosity loses oxygen and the diversity of the web dies.”

It’s not theatre. Vivaldi’s stance is a real product decision from a small browser maker pushing back against an industry trend: a growing number of companies are building AI-native browsers and embedding assistant-style features that summarize pages, automate tasks, and — critics say — deprioritize the original websites those assistants draw on. The result, the argument goes, is fewer clicks to publishers, fewer eyeballs for creators, and a web that fades from a marketplace of ideas into an API served by an intermediary.

Why Vivaldi’s pique matters

Vivaldi is small but unafraid to be principled. Unlike Microsoft, Google, or Apple, which are actively folding generative AI and assistant layers into search and browsing experiences, Vivaldi has doubled down on personalization, privacy and letting people do the digging themselves. Its announcement is both philosophical and tactical: the company says it will avoid turning its browser into an “AI-first” interface unless AI can be used in ways that don’t steal intellectual property, violate privacy, or hollow out exploration. That positioning is pitched at power users, researchers, and anyone who values control over convenience.

The timing is telling. Startups such as Perplexity (Comet) and The Browser Company (Dia) are aggressively pushing the opposite proposition — that tightly integrated assistants and agentic features are the future of how people will interact with the web. Perplexity markets Comet as an assistant-centric browser that “acts as a personal assistant and thinking partner,” automating research and workflows; The Browser Company has likewise built AI experiences into Dia to shortcut the path from question to answer. Those products are explicitly designed to reduce the friction of searching and organizing information — even if that comes at the cost of fewer direct visits to source sites.

The publisher problem — and the data that backs it up

Vivaldi’s warnings aren’t just ideological. Independent research and reporting show measurable harms for the sites that make journalism and creative work. A recent Pew Research-centered analysis found that when search results include an AI-generated summary (an “AI Overview” or similar), users are far less likely to click through to the underlying links — sometimes by roughly half. In other words, when the assistant hands you the answer, you often don’t go to the place that produced it. That has real economic consequences for publishers who rely on referral traffic.

The Financial Times and other outlets have documented an industry scrambling to adapt to what some call “Google Zero” — a world where search engines’ AI summaries satisfy queries without sending people to the original reporting that produced the facts. Publishers have reported drops in referral traffic even as their content is used more widely in AI answers; the FT covered executives talking about diversifying revenue away from search-driven advertising toward subscriptions, events and direct reader relationships. That’s exactly the dynamic von Tetzchner invoked: if people stop clicking through, the web — at least the parts that sustain independent journalism and niche communities — becomes poorer.

Convenience vs. curiosity — the tradeoffs

Agentic browsers promise convenience: automate a research workflow, summarize ten pages into a short brief, or let the assistant “do” the web for you. That’s powerful. But convenience has tradeoffs. Brave and other critics have raised security concerns (prompt injection, hidden scraping) and privacy worries about agentic browsers that retain and analyze richer behavioral data than traditional browsers. Tech commentators warn the new wave could revive — or even intensify — the worst parts of surveillance advertising, because assistant features thrive on understanding what you do and why. For people who value privacy, agency and the joy of following a hyperlink into a rabbit hole, those tradeoffs are existential, not academic.

To Vivaldi, the risk is not only surveillance and broken incentives, but also quality. AI assistants are spectacular at sounding confident, and spectacularly fallible when they hallucinate or misattribute facts. If the browser becomes the default arbiter of truth — producing polished answers without nudging readers to the original sources — the web’s fact-checking mechanisms weaken. That’s part of why Vivaldi insists on resisting the rush: it wants users to remain active participants in discovery, not passive recipients of canned answers.

Is Vivaldi swimming upstream — and can it win?

Vivaldi is not the first to promise resistance to AI’s siren call, but its message is crisp and potentially resonant. There’s a clear market of privacy-conscious, power users and researchers who dislike the idea of a browser that automates curiosity away. For that cohort, Vivaldi’s branding as a tool for “curious minds” could be a meaningful differentiator. At the same time, convenience is a brutal competitor. Perplexity, The Browser Company and other startups — backed by big money and heavy user enthusiasm — are moving fast to make agentic browsing feel indispensable. If users find that assistants save them time and frustration, many will accept the tradeoffs.

There’s also a middle path: browsers could offer optional assistant features that are transparent, opt-in, and designed to credit and link to original sources. Perplexity and others are even experimenting with revenue-sharing models to compensate publishers for content used in AI answers — a tacit acknowledgment that the old referral economy is under strain and that some kind of new settlement may be needed. Whether those arrangements will scale fairly, and whether they will preserve serendipity and deep engagement, remains to be seen.

What this means for the web — and for creators

If the future tilts toward agentic assistants, websites will have to adapt. That could mean:

• making content more programmatically accessible (APIs, structured data, machine-readable licenses),
• building stronger direct relationships with readers (subscriptions, newsletters, events), and
• experimenting with technical and commercial ways to make sure AI-served answers link back, credit, and — ideally — compensate.

If Vivaldi’s thesis is correct, those shifts will be critical to preserving a diverse, creator-funded web. If the thesis is wrong, and people mostly prefer an assistant that does the clicking for them, publishers will have to be even more aggressive about monetization and audience ownership. The FT’s reporting shows many publishers are already moving down that road.

The bottom line

Vivaldi’s pronouncement is less about technology than values: it’s a bet that exploration — the act of following links, reading varied viewpoints, and doing the work of research — is worth preserving. It’s also an argument that convenience should not automatically swallow context, credit, or privacy.

Whether Vivaldi’s human-first stance slows or merely protests the tide is an open question. The market will decide how many people want an assistant running the web for them and how many still prefer to explore it themselves. For now, Vivaldi has staked a clear position: if the web is to remain a rich place of discovery and not just an answer machine, someone needed to say it out loud.