HN Debrief

Kagi added a setting for removing paywalled links from search results

  • Search
  • AI
  • Privacy
  • Media
  • Developer Tools

Kagi’s update adds a simple switch: remove paywalled links from search results entirely. Kagi already marked paywalled results with a dollar icon, so the new piece is not disclosure but exclusion. People liked that because the practical annoyance is not moral opposition to paying for content. It is clicking a promising result, hitting a wall, and bouncing. For many, especially when searching for quick answers or technical information, a paywalled result is dead inventory. That fits the broader case users were already making for Kagi. They pay for it because it gives them control Google does not, through source downranking, custom redirects, lenses, and slop filtering. The mood was strongly pro-Kagi and strongly anti-Google, with a lot of users saying paid search feels like one of the few consumer software products still aligned with the customer instead of advertisers.

If you build discovery products, users increasingly want results filtered by access friction, not just relevance. For media and subscription businesses, this is another sign that paywalls inside open search are becoming a ranking disadvantage unless you also have a bundle, a strong direct audience, or a smoother one-off purchase path.

Discussion mood

Very positive toward Kagi and the feature, driven by frustration with Google, paywall friction, and a sense that paid search aligns incentives with users. The notable sour notes were concern that hiding paywalled sources can bury better journalism, irritation at Kagi boosterism overwhelming the actual feature discussion, and a recurring political objection to Kagi’s Yandex ties.

Key insights

  1. 01

    Paywalls distort ranking, not just access

    The sharper objection to paywalled results was not simply "I do not want to pay." The point was that search engines rank those pages on the full article text while the user gets only a teaser or a wall. That means inaccessible pages can displace readable ones even when the engine knows most searchers will bounce. This framing turns the feature from convenience into search quality control.

    If your product ranks or recommends content, treat access constraints as part of relevance. A result the system can analyze but the user cannot consume should usually be downgraded or clearly segmented.

      Attribution:
    • Nextgrid #1 #2
  2. 02

    Micropayments fail on business incentives

    The most convincing pushback on pay-per-article was that the machinery is not the hard part. Wallet top-ups, batching, and periodic settlement are straightforward. The failure point is that publishers prefer predictable subscription revenue, and readers prefer all-you-can-eat pricing over repeated tiny purchase decisions. That is why the market keeps recreating bundles instead of true micropayments.

    Do not assume a smoother checkout fixes fragmented paid content. If you need users to pay for many small units, test bundles first and treat "mental transaction cost" as a core product constraint.

      Attribution:
    • kelnos #1 #2
    • notatoad #1
    • boplicity #1
  3. 03

    Bundles beat one-off article sales

    Apple News and public library access stood out as the only models people described as workable at scale. They reduce account sprawl and make payment feel settled in advance. The complaints were about platform lock-in, ads, and mediocre apps, not about the bundle concept itself. That says the demand is less for perfect micropayments than for one subscription that unlocks many publishers cleanly.

    Media products should look harder at interoperable bundles and entitlement tokens across participating sites. Discovery surfaces like search will increasingly punish stand-alone paywalls that force a fresh buying decision on every click.

      Attribution:
    • tensor #1 #2
    • reaperducer #1
  4. 04

    Search-first AI is useful but still brittle

    Kagi’s assistant won praise because it behaves more like guided research than pure chatbot synthesis. The important caveat was that it still amplifies source errors with confident prose, and several users said they use it as a complement to manual search rather than a replacement. The win is speed and query reformulation, not reliable truth extraction.

    When deploying AI over search, optimize for source visibility, brevity, and easy verification. Do not design around the fantasy that a synthesized answer removes the need to inspect underlying pages.

      Attribution:
    • kelnos #1
    • QwenGlazer9000 #1
    • freehorse #1
  5. 05

    User control features are Kagi’s real moat

    A surprising amount of enthusiasm was anchored not to this paywall switch alone but to the wider toolbox around it. People cited redirects that swap Reddit links to Redlib or SafeReddit, site priorities, lenses, and slop filters. The pattern is that users are willing to pay when search becomes programmable and they can encode their own preferences instead of accepting one opaque ranking stack.

    For search and assistant products, configurability can be more defensible than raw model quality. Let power users redirect, suppress, and re-rank the web, and they will tolerate imperfect recall if the system feels governable.

      Attribution:
    • optionalsquid #1
    • Phemist #1
    • Saris #1
  6. 06

    AI is cleaning up a mess AI made worse

    Several people captured the current search paradox cleanly. Search assistants feel necessary because the web is clogged with SEO spam and low-grade content, and generative AI has accelerated that flood by making garbage cheaper to produce. That leaves AI tools acting as compensating filters for a deterioration partly caused by AI itself.

    Expect more demand for filtering, provenance, and domain-level trust controls. Products that help users escape synthetic junk now address a structural problem, not a temporary annoyance.

      Attribution:
    • sleight42 #1
    • II2II #1
    • Calazon #1
    • QwenGlazer9000 #1
  7. 07

    Yandex remains a live reputational risk

    A separate but persistent objection was that Kagi still relies in part on Russian search data and therefore sends money into the Yandex ecosystem. The founder defended this as a product-quality choice rather than political endorsement, but critics treated it as a clear moral line and a reason to cancel or avoid subscribing. This did not dominate the thread, but it was one of the few issues that cut through the goodwill.

    If your infrastructure depends on politically sensitive vendors, expect that to shape customer trust even when the dependency is technically small. Offer clear sourcing transparency and opt-outs before the issue becomes your brand story.

      Attribution:
    • freediver #1 #2
    • jwr #1
    • neoromantique #1

Against the grain

  1. 01

    Best source can still be worth the hassle

    Some people want the opposite default because the highest-quality information is often behind a paywall, and they would rather know it exists than have the engine quietly hide it. One commenter extended that to privacy, saying they would not pay for a metasearch service with an account requirement anyway and prefer splitting searches across multiple engines.

    If you add friction-based filters, make them transparent and reversible. Advanced users will want visibility into what was suppressed so they can trade convenience for completeness when the query matters.

      Attribution:
    • cormorant #1
    • OroPla #1
  2. 02

    Readable is not the same as trustworthy

    A skeptical line running through the praise was that removing paywalled links can push users toward the wrong pages. Free results are often lower quality, and AI summaries built on top of them can smooth over uncertainty and make weak sourcing feel authoritative. That weakens the case for using accessibility as a proxy for usefulness.

    Do not let convenience filters masquerade as quality filters. Pair them with source-quality controls, labels, or saved exceptions for outlets your users actually trust.

      Attribution:
    • SamBam #1
    • kelnos #1
    • dspillett #1
  3. 03

    The filter is opaque without a source list

    A practical criticism was that Kagi did not explain which sites count as paywalled. That makes it hard to predict what disappears and harder for competitors or users to reproduce the feature. For a setting that materially changes recall, the lack of transparency stood out.

    Any suppression feature that changes what users see needs inspectability. Publish the matching logic, expose affected domains, or at least show a suppressed-results count so people can audit the tradeoff.

      Attribution:
    • dmtryshmtv #1
    • BrunoBernardino #1

In plain english

AI ↩
Artificial intelligence, software systems that perform tasks such as analyzing code or generating text.
Redlib ↩
An alternative front end for Reddit that lets people view Reddit content without using Reddit’s standard interface.
SEO ↩
Search engine optimization, the practice of improving a site's visibility in search engine results.
UX ↩
User experience, how a product feels to use in practice.
Yandex ↩
A major Russian technology and search company whose data and services were discussed as part of Kagi’s search stack.

Reference links

Kagi product features and docs

Paywalls, micropayments, and media economics

Search quality and web decay

Privacy and alternative access tools

Journalism and news references

Kagi background and governance concerns