Kagi has introduced a setting that lets users remove or downrank links to paywalled pages in its search results. It is an account-level preference in a subscription search engine, not a change to how the wider web is indexed.
Key takeaways
- Kagi is a paid, subscription-based search engine that has long allowed users to adjust which domains appear in their own results.
- The newly discussed setting extends that idea specifically to pages behind paywalls, letting a user filter them out of the results they see.
- The change affects only what an individual account sees; it does not remove any page from the web or from other people’s searches.
- Detecting a paywall reliably is difficult, because access rules vary by article, region, device and how many pieces a reader has already opened.
- The feature has revived a long-running argument about whether filtering paid journalism helps readers or undermines the business model that funds reporting.
What exactly has been added?
The setting in question allows a Kagi user to tell the search engine how to treat results that sit behind a paywall. In broad terms, options of this kind either hide such results entirely, push them lower down the page, or mark them visually so a reader knows before clicking that access may be restricted. The precise wording, default state and granularity of Kagi’s implementation are details best checked in the product’s own settings and documentation rather than assumed, because search preferences are frequently adjusted after release.
What is clear from the way the feature has been described is its scope. It is a personalisation control tied to an account. When it is switched on, the results page that the account holder sees is filtered accordingly. Nothing is deleted from the index, no publisher is delisted globally, and other users are unaffected. This is the same architectural pattern Kagi already uses for its per-domain ranking controls, where a user can raise, lower, pin or block particular sites in their own results.
Why is this being discussed now?
The feature surfaced on Hacker News and attracted a large volume of comments, which is usually a signal that a change has touched an existing sore point rather than that it is technically remarkable. The sore point here is familiar to anyone who searches for information on a specific topic: a query returns a promising headline, the click leads to two paragraphs and a subscription prompt, and the reader goes back to the results and tries again. Repeated often enough, this becomes the dominant friction in a search session.
A second reason for the attention is that Kagi occupies an unusual position. Because its users pay directly rather than being shown advertisements, the company has more room to build controls that serve the searcher’s stated preferences even when those preferences are inconvenient for content publishers. Features that a large ad-funded engine would find commercially awkward are easier to ship in that context, which makes each one a talking point.
The background a newcomer needs
Paywalls came into wide use as online advertising revenue proved insufficient to fund professional reporting. They take several forms. A hard paywall blocks all content to non-subscribers. A metered paywall allows a set number of free articles in a period before blocking further access. A registration wall asks for an email address rather than payment. Some publishers operate a hybrid model in which certain articles are free and others are reserved for subscribers.
For a search engine, this variety is the core technical problem. A crawler is often served a different version of a page than a logged-out human visitor, partly because publishers have historically wanted their content indexed even when it is not freely readable. Search engines have used structured metadata for years to let publishers declare which parts of a page are restricted, and this kind of declaration is one obvious input for a paywall filter. But not every publisher marks their content, and marking can be incomplete or out of date.
Separately, search engines have long offered users ways to shape their own results — blocking domains, filtering by language or region, suppressing certain file types. A paywall setting is conceptually the same category of tool, applied to an attribute of the page rather than to its address.
Who is affected, and how?
Readers who use the setting get a results page with less friction, at the cost of some loss of coverage. On topics where the most thorough reporting sits behind a paywall — investigative work, specialist trade press, financial analysis — hiding those results means seeing weaker sources that happen to be free, including aggregators that summarise the original work without adding to it.
Publishers with paid models may see fewer arrivals from users who enable the filter. The size of that effect depends entirely on how many people switch it on and how much traffic the engine sends in the first place; in the case of a small subscription search engine, the absolute numbers are likely to be modest. It is not possible to state a figure here, and any specific claim about traffic impact should be treated with caution unless a publisher reports it directly.
Institutional users are a distinct case. Many readers reach paywalled articles legitimately through a university, library or corporate subscription. For them, a blanket filter removes results they can actually read, which argues for a downranking or labelling option rather than outright removal.
Where informed people disagree
The disagreement is not really about the setting’s mechanics; it is about what a search engine is for. One view holds that a search tool exists to serve the person typing the query. If that person cannot read paywalled pages, showing them is a waste of their attention, and letting them opt out is straightforward respect for their time. On this reading, the filter is no different from excluding a language the user does not speak.
The opposing view holds that search results are also a distribution channel that sustains the production of the material being searched. Making paid journalism systematically less visible, even one account at a time, weakens the incentive to fund it and pushes readers towards material produced cheaply for advertising or attention. A filter that treats a paywall as a defect rather than a business model, on this argument, quietly rewards the lowest-cost content.
There is also a practical dispute about accuracy. Because paywall behaviour varies by article and reader, any classifier will produce both false positives — free pages hidden because the domain usually charges — and false negatives. Whether a setting like this feels useful or unreliable depends heavily on how those errors are distributed, and that is something only sustained use reveals.
The practical implications for a reader
If you are considering using a filter of this kind, a few decisions matter more than the switch itself.
Prefer downranking or labelling over hard removal if the option exists. Labelling preserves the information — you can still see that a well-regarded source covered the topic — while removing the surprise of a blocked click. Hard removal is best reserved for queries where you genuinely only want free material.
Treat the setting as adjustable rather than permanent. Research-heavy queries and casual lookups have different needs, and the sensible configuration differs between them.
Check whether your existing subscriptions or institutional access are being filtered out. If a domain you can read is being hidden, per-domain controls are usually the right corrective, raising that site rather than leaving it caught by a general rule.
Finally, be aware that the filter changes what you know exists. A results page with paywalled sources removed is not a neutral picture of what has been written on a subject, and for anything consequential it is worth running the query again without the filter.
What to watch next
Three things are worth following. The first is how paywall detection is actually implemented and how accurate it proves — whether it relies on publisher-declared metadata, observed page behaviour, or a maintained list, and how errors are corrected. The second is whether other search engines adopt comparable controls; a setting that stays confined to a niche subscription product has different consequences from one that reaches a mainstream engine. The third is the publisher response, which historically has ranged from ignoring such tools to adjusting how content is marked for crawlers.
None of these outcomes can be predicted from the announcement itself. What the feature does confirm is that user-controlled filtering, rather than a single universal ranking, is an established direction for search products built around paying users.
Frequently asked questions
Does this remove paywalled sites from the internet?
No. The setting operates on what an individual account sees in its own search results. The pages remain online, remain indexed, and remain visible to anyone who has not enabled the filter. It is a display and ranking preference, comparable to blocking a domain in your own results, and it has no effect on other users or on the publisher’s own website.
How does a search engine know a page is paywalled?
Search engines can use several signals, including structured metadata that publishers add to declare which parts of a page are restricted, observed differences between what a crawler and a visitor receive, and maintained lists of known subscription domains. None of these is complete. Because access rules often vary by individual article and by reader, any detection method will misclassify some pages in both directions.
Will this stop me finding good sources?
It can. Much of the most thorough reporting and specialist analysis is funded by subscriptions, so a filter that hides it may leave you with free material that is thinner or derivative. If you are researching something that matters, it is sensible to run the query again with the filter switched off to see what you would otherwise have missed.
Is filtering paywalled results the same as bypassing a paywall?
No, and the distinction is important. A filter simply chooses not to show you links you are unlikely to be able to read. Bypassing means circumventing the publisher’s access controls to read restricted content without paying. The setting described here does the former; it does not grant access to anything and does not alter the publisher’s own restrictions.
Do other search engines offer something similar?
Mainstream search engines have generally not offered an explicit paywall toggle, though many have long supported user-level result customisation of other kinds, and various browser extensions and third-party tools attempt to label paywalled links. Whether comparable settings appear more widely is an open question, and it is not possible to say in advance how any particular engine will approach it.
What should I do if the filter hides a site I can actually read?
Most search engines that offer this kind of personalisation also offer per-domain controls. Raising or pinning the specific domain you have access to is usually more precise than switching the general filter off, because it addresses the one source causing the problem while leaving the broader preference intact. Checking the setting periodically is worthwhile, as detection rules change.
Sources and further reading
- Kagi’s own product documentation and settings pages, which describe the available search personalisation controls in detail and are the authoritative record of current behaviour.
- Hacker News discussion threads, useful for the range of user reactions and technical objections rather than as a factual record.
- Published search engine developer documentation on structured data for subscription and paywalled content, which explains how publishers declare restricted sections to crawlers.
- Press industry trade publications and journalism research institutes, which cover the economics of subscription models and referral traffic from search.
Surfaced from the hackernews signal “a search filtering setting”. AI-assisted draft, editorially reviewed.
