The editorial argues that search quality is increasingly about what you don't see, and Kagi's set-and-forget toggle spares users from wasting attention on links they can't read. It contrasts this favorably with Google's AI Overviews approach, which paraphrases paywalled content rather than removing it from the results.
The editorial highlights that Kagi chose to host the paywall domain list on GitHub rather than building a proprietary classifier, so adding or removing a domain becomes a pull request instead of a support ticket. This transparency is framed as the most interesting design choice of the release, empowering users to shape their own search experience.
By submitting the changelog note and driving it to 1,082 points, the submission signals that a simple always-on toggle resonates far more than the prior per-query lens. The editorial reads this response as validation that users want a persistent default, not a filter they must remember to invoke each search.
Kagi shipped a settings toggle that removes paywalled results from search entirely. Not a lens, not a lowered ranking, not a badge — a switch that makes the results disappear. The changelog entry landed as item #11296 and hit the Hacker News front page with 1,082 points, which for a single-line changelog note is a rare signal that the feature scratched a real itch.
The mechanism is deliberately simple: a curated list of paywall domains, maintained on GitHub, that Kagi filters against when you flip the toggle in settings. Because the list lives in a public repo, adding or removing a domain is a pull request rather than a support ticket. Users have already been submitting PRs for domains the maintainers missed — the New York Times, Bloomberg, the Wall Street Journal, the Financial Times, Substack posts behind reader-limits, and the growing tail of Medium-hosted publications that hide the article after two paragraphs.
This is Kagi's second swing at the paywall problem. The first was a lens — a saved search filter you could toggle per query — which required you to know a lens existed and remember to use it. The new toggle is set-and-forget: enable once, never see a paywalled result again unless you disable it.
Search quality has quietly become a UX story about *what you don't see*. Google's answer to the SEO-slop problem has been AI Overviews, which paraphrase the paywalled content back to you without a click-through — a workaround that sidesteps the paywall by consuming the article on your behalf. Kagi's answer is the opposite: remove the source entirely so you never spend attention on a link you can't read. Both approaches treat paywalls as a signal, but they draw opposite conclusions about what to do with the traffic.
The interesting design choice here is who owns the blocklist. Kagi could have built a proprietary paywall classifier, run it as a ranking signal, and never published the list. Instead they put it on GitHub and let the community do the domain-hunting. That's cheaper to maintain, harder to game, and — critically — auditable. If a publisher complains that they're being unfairly filtered, the reply is "open a PR against the repo" rather than "email support." It's the same move Pi-hole made with blocklists and uBlock Origin made with filter lists: outsource the taxonomy to the people who care most about accuracy.
The HN comment thread reveals the fault line this feature exposes. One camp argues paywalled journalism is a legitimate business model and hiding it from search is a rounding error away from theft. The other camp — clearly larger, given the vote total — points out that search results are already ranked by relevance to *you*, and a link you cannot open is by definition not relevant. There's a middle position that the toggle nicely accommodates: leave it off by default, let users self-select. Nobody's being deplatformed; the WSJ still ranks for WSJ searches. The change is that a query for "kubernetes ingress timeout" no longer surfaces a Medium article gated behind a subscription wall.
The broader pattern here is that user-agent-side filtering is eating server-side ranking as the mechanism for search quality. The old model — Google decides what's good, you accept it — is being replaced by a stack where the search engine returns candidates and the user configures the filter. Kagi's entire pitch is built on this: personal blocklists, domain ranking sliders, lenses, and now the paywall toggle. Each feature is a small transfer of control from the search engine to the searcher. That's a strong story if you're a paying Kagi user (starting at $10/month for the Professional tier); it's an existential threat if you're a publisher whose SEO strategy depends on being ambiently indexed.
If you've been running uBlacklist with a hand-curated list of Medium, Quora, and Pinterest domains — and if you're the kind of dev who reads changelogs on a Saturday, you probably have — the Kagi toggle collapses that into one setting. You still need uBlacklist for the domains the community list won't cover (personal preferences, competitors, that one blog that keeps outranking your docs), but the paywall category is now handled upstream.
For anyone building a search or content-discovery product, the takeaway is that filter transparency is becoming a feature, not an implementation detail. "We use a proprietary algorithm" plays worse every year. "Here's the GitHub repo with our blocklist, send us a PR" plays better. It's the same trust-through-legibility argument that pushed encryption libraries toward open source: if the mechanism is auditable, the trust doesn't have to be.
There's also a practical implication for anyone publishing behind a paywall or considering it. Kagi's user base is small — a few hundred thousand paying subscribers — but it skews heavily toward the exact audience most publishers claim to want: senior engineers, technical decision-makers, high-intent readers with disposable income. If you paywall your content, you now have a settings-level opt-out from that audience. The math on "partial paywall with a metered soft-wall" versus "hard paywall filtered out of Kagi entirely" just shifted, at least for the segment of readers who search via Kagi.
The interesting question is whether this pattern jumps. Brave Search has flirted with similar ideas, DuckDuckGo has bang commands but no paywall filter, and Google is structurally incapable of shipping this because AI Overviews depend on the paywalled content being indexed. If the community-maintained blocklist grows past a few hundred domains and starts covering the long tail of Substack-hosted publications, expect it to become a de facto standard that other engines either adopt or explain away. Either way, the toggle is a good reminder that in 2026, the most valuable search feature might be the button that hides results.
The one thing I find slightly grating about links to Kagi blogs is the top comments are almost always "I use Kagi and it's great!" rather than about the content of the blog. And I'm a happy Kagi subscriber!I get that this is an option, but what it really shows to me is how broken
I've been enjoying Kagi for the last couple years.Even as LLMs slurp up most of the internet and replace search I think Kagi is still useful.Reddit has recently blocked access to old reddit without an account and the ability to filter out stuff like that is useful.
This makes sense to me as a Kagi user. The chance that I'm going to subscribe to read an article I found via a search is zero.
Killer feature. It would be awesome to have some plugin or userscript to auto-swap the e-begging scammy link for an Archive link instead.
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
I think this is amazing. Love Kagi. I’m happy to pay for a good search.Nobody speaks about their AI Assistant, but it is really good. They somehow harnessed it so that it mainly searches info first and sticks to the verifiable data. I prefer it to Claude or any other, because it actually answers the