There is a particular complaint that turns up whenever people talk about the web now: it feels smaller. Fewer places, the same places, the same shapes of page. The complaint is usually filed under nostalgia and then dismissed, which is a mistake, because underneath the feeling there is a mechanism that actually changed.
For a stretch of the web's life, one ordinary way of arriving somewhere new was to follow a link that a person had placed by hand, because they had read the thing and decided it was worth pointing at. We are not going to put a figure on how much of the web's movement worked that way. We looked for a measurement and did not find one we would repeat, and a number that merely sounds right is worse than no number. What can be shown is that the surface existed and came built into the software people published with, which is the subject of a later section. That layer had a name in one of its most common forms. It was called a blogroll, it sat in a sidebar, and it is largely gone.
This piece is about what that layer did, why it went, and what any honest attempt to rebuild a piece of it has to accept about itself. It is not an argument that the past was better. Parts of it were worse, and the reasons blogrolls disappeared are mostly boring rather than sinister.
What this is, and who it is for
This is an argument about a mechanism, written for two kinds of reader.
The first is someone who has noticed that finding an unfamiliar small thing online has become hard, and who wants an explanation more precise than a complaint about algorithms.
The second is someone building or running something small who cannot work out why nothing they do makes them findable. The short version of the answer is that the layer that used to do that job for them was never funded, and when it stopped being maintained nothing took over the specific function it performed.
Every factual claim about a third party in this piece is attached to a source we opened, listed at the end with the date we retrieved it, and where we could not source something we have said so rather than rounding the claim up to something confident. Our sourcing rules, including what counts as primary and why we refuse to repeat unsourced figures, are at our source ledger. The wider set of pieces this belongs to is at the independent web.
What a blogroll actually was, mechanically
Strip away the era and the styling and a blogroll had five mechanical properties. Each of them matters, and losing them was not equivalent to losing a design feature.
It was outbound. The list existed to send you away. That is unusual. Almost every subsequent surface that recommends things to you is built to keep you where you are, because the entity paying for the surface is measured on how long you stay.
It was standing rather than momentary. A recommendation in a post is an event that scrolls away. A blogroll was a persistent claim, sitting on every page, saying: these are the places I currently think are worth your time. It could be wrong, and being wrong was visible, because it stayed up.
It was attributed. You knew whose list it was. That gave the recommendation a shape. A link from a person who writes carefully about bicycle frames means something specific; a link from an anonymous aggregator means nothing at all, because there is no reputation attached to it that could be spent.
It was ordered by judgement. Not by traffic, not by payment, not by relevance to a query. Just by what one person rated as worth listing, with all the bias and blind spots that implies.
It was cheap at small size and expensive at large size. Twenty links is a Sunday afternoon. Four hundred links is a job. This property turns out to explain most of what happened next.
The mechanism was also cheap to acquire, because at least one widely used publishing tool shipped with it. WordPress carried a built-in Link Manager, and the WordPress.org plugin that brings it back, written by Andrew Nacin and last updated in August 2012, describes it in exactly those terms: "the Link Manager (which some use to build blogrolls)". We retrieved that plugin page on 6 September 2026. Availability matters. A discovery layer made of volunteer effort only exists if the effort required per person is close to trivial.
What it did that nothing since has replaced
Four functions, and it is worth separating them, because some have been partly replaced and one has not been replaced at all.
Transitive discovery. You arrived at one site, found a list of the sites its author read, followed one, found another list. The web was navigable by walking rather than by querying. Each step carried an implicit endorsement from a specific person who had staked a small amount of credibility on it.
Discovery without a query. You do not need to know what you are looking for to follow a link that someone else found interesting. This is the function that query-shaped tools structurally cannot perform, and we come back to it below.
Reachability without ranking. A site could be findable because one well-read person linked to it, with no requirement that it compete for a search term, produce content at volume, or be commercially significant. For anything small, this is the difference between existing and existing findably. We take that argument apart in more detail in why the small stuff got harder to find.
A standing accuracy obligation, weakly enforced. If a listed site went dark, readers noticed and the author eventually fixed the list, or the list rotted in public and readers drew their own conclusions. Nobody enforced this. It was enforced by embarrassment, which works at small scale and at no other.
Why they disappeared: the honest reasons, not just the villainous ones
The comfortable story is that a discovery layer was taken away from people. The more complete story includes several reasons that have no villain in them at all.
Maintenance burden. A curated list is not a thing you make. It is a thing you keep. Every link on it is a small standing debt: someone has to notice when it breaks, when the site changes hands, when the person you were recommending starts publishing something you would not want to recommend. The debt grows with the length of the list and never stops accruing.
Link rot. Links to things that no longer exist accumulate on any list left alone for long enough. The Pew Research Center's study When Online Content Disappears, published on 17 May 2024 and based on a random sample of just under a million webpages drawn from the Common Crawl archives, reports that 38% of the webpages that existed in 2013 were no longer available at the time of the study, and that 25% of all the pages it collected across 2013 to 2023 were no longer accessible as of October 2023. We retrieved it on 6 September 2026. A list of links decays at whatever rate its destinations do, and a list with many dead entries stops being useful and starts being embarrassing. The cheapest response to an embarrassing list is not to fix it. It is to delete it.
The labour was unpaid and invisible. Nobody was compensated for curating, and the work does not photograph well. The reward was reputational and diffuse. That is a fragile funding model for a layer that turned out to be load-bearing. We follow this cost structure further in the economics of not pruning.
The list became a social object. Adding a person is a favour. Removing a person is a statement. Once a list has social weight, editing it stops being a maintenance task and becomes a diplomatic one, and diplomatic tasks get postponed indefinitely.
The surface stopped being there by default. WordPress's own release documentation for version 3.5, released on 11 December 2012, records that the "Link Manager is hidden for new installs and for any existing installs that have no links", restorable only by installing a plugin. We retrieved that page on 6 September 2026. We have not established what every hosted publishing platform allows, and we are not going to assert it: if you write somewhere hosted, whether you can keep a permanent list of outbound links in a sidebar or a footer is answered by that platform's own template documentation, and it is worth checking rather than assuming. You cannot maintain a feature the software does not offer.
The traffic pattern changed underneath it. A blogroll pays its owner back when readers arrive by browsing, because a browsing reader is in the mood to keep walking. When most arrivals come from a query or a feed, they came for one thing, and a sidebar of unrelated links is furniture. The incentive to maintain it weakened even for people who still could.
Outbound links acquired an economic meaning. That links carry ranking value is stated in the founding literature of web search. Brin and Page's 1998 paper The Anatomy of a Large-Scale Hypertextual Web Search Engine calls the web's link structure "the citation (link) graph" and explains that "a page can have a high PageRank if there are many pages that point to it, or if there are some pages that point to it and have a high PageRank". Google Search Central's current published spam policies define link spam as "the practice of creating links to or from a site primarily for the purpose of manipulating search rankings" and give "Exchanging money for links, or posts that contain links" as an example, and its separate guidance on outbound links asks publishers to mark advertisements and paid placements with rel="sponsored". We retrieved all three on 6 September 2026. We are not going to say what anyone intended by any of it. The structural point is narrower: a link became a thing with a price attached, and a thing with a price attached is placed less casually than a thing without one.
What replaced them, and what that replacement optimises for
Three surfaces absorbed most of the discovery function, and it is important to be accurate about them: each is very good at the thing it optimises for, and each of those things is legitimate.
| Surface | Optimises for | What it needs from you |
|---|---|---|
| Query-based search | Returning the best answer to a question you have already formulated | You must be able to name the thing |
| Feeds and recommendation streams | Holding attention profitably within a session | You must stay |
| Marketplaces and platforms | Completing a transaction inside the platform | You must transact there |
None of these is a failure. A search engine that answers a stated question well is genuinely useful, and the fact that it cannot answer a question you did not know how to ask is not a flaw in it. A feed that holds attention is doing exactly what its business model asks. A marketplace that closes a sale is providing a real service to both sides.
What is worth noticing is that all three are inward-pointing in a way the blogroll was not. Sending a reader somewhere else and never seeing them again is a cost for all three and was the entire purpose of the fourth. That difference is structural. It requires nobody to have decided anything.
Why the replacement cannot do the same job
Four reasons, none of which is about quality.
A query cannot surface what you cannot name. If you do not know that a particular kind of workshop exists, you cannot type its name. Search is answer-shaped, and discovery of the unfamiliar is not a question with an answer; it is an encounter. That is the single function that has not been replaced.
Nobody attributable is standing behind the recommendation. A ranked result is the output of a process. There is no person whose reputation goes down if it was a bad suggestion, so there is nothing for a reader to calibrate against over time. Attribution is not a nicety; it is the mechanism by which a reader learns how much to trust a source.
A result is regenerated; a list is maintained. When a query is run again, the answer is produced fresh, so nothing carries an obligation to have checked that a specific place is still there. A standing list has that obligation baked in, which is precisely why keeping one is expensive.
The surfaces that hold attention have no reason to send it away. Again, not a conspiracy. A surface funded by time spent on it will not, over time, develop a strong outbound-linking habit, and it does not need anyone to have decided that for the outcome to appear.
What a working alternative has to accept about itself
Anyone rebuilding a piece of this layer runs into the same constraint the original ran into, and pretending otherwise is how these projects fail. The constraint is that curation cost scales with the size of the collection, not with the size of the audience.
Doubling readers costs almost nothing. Doubling entries doubles the recurring work forever. This means:
- It will be small. Not small at the start and large later. Small as a permanent property, unless the recurring work gets funded. We are not claiming that no funding model for sustained third-party checking exists anywhere, because that is not a claim anybody can support. We are saying we could not name one that does not end with the checked party paying, and payment by the checked party is the arrangement that empties the check of meaning. If a model exists that survives that objection, we would like to know about it.
- It must publish its bar. A collection whose inclusion rule is private is indistinguishable from one that sells positions, and a reader has no way to tell the difference from the outside.
- It must decay on purpose. Removing things is the work. A collection that only ever adds is a collection that will be wrong within a couple of years, and the wrongness will be invisible until someone clicks.
- It must be dated. A recommendation without a date is a claim about the present made by someone who may last have looked years ago.
- It must be free of position payments. The moment a place in the list can be bought, the list stops being evidence of anything and becomes an advertising product wearing the vocabulary of curation.
- It must be honest about being partial. A small curated collection is a sample of one person's attention. Presenting it as coverage of a category is a lie that is easy to tell and hard to detect.
The other people working on this run into the same wall, which is why they are all small too. We survey that honestly in why every attempt stays small, and the longer background is in the longer history.
What we are attempting, and what it costs
This site is a register of small independent businesses that a named person has looked at, with the date of the check published next to the entry, and a published bar for what gets in. The register is where the entries live and the stumble shows you one at a time.
The important constraint is that the method and entry bar govern every entry; the rules are not rewritten to fit a preferred candidate.
This article supplies no dated re-check log, so it makes no cost claim. A conforming measurement would have to publish the observed work and period; the presence of a method alone is not a result.
We are not continuing anything and we are not anyone's successor. Earlier services that did discovery work existed, some of the people who built them are still working, and describing ourselves as their inheritors would be both untrue and rude. This is a small attempt at one specific job: keeping a short list true.
If you would rather look somewhere other than here, that is a reasonable preference, and where people still find independent work is a list of routes, most of which are not us.
In Plain English
A blogroll was a list of links that one person maintained by hand, pointing at other people's sites. It let you find things you did not know to search for, because you were walking through somebody's judgement rather than querying a database.
They mostly disappeared for unglamorous reasons: keeping a link list true is recurring unpaid work, links break, removing someone from a list is socially awkward, and the software stopped putting the feature in front of people by default.
What took over is good at other things. Search answers questions you can phrase. Feeds hold your attention. Platforms complete sales. None of them is built to hand you off to a stranger's small website and forget about you, and there is no villain required for that to be true.
Anything that rebuilds a piece of this will be small, because the work of keeping a collection accurate grows with the collection and never stops. Small is the honest outcome, not a stage before scale.
Sources
Each entry below is recorded in the source ledger under the identifier shown.
- S-036 · WordPress Link Manager — Link Manager (plugin by Andrew Nacin), WordPress.org Plugin Directory. https://wordpress.org/plugins/link-manager/ Retrieved 2026-09-06. Supports: WordPress shipped a built-in Link Manager, described by the plugin's own text as the feature "which some use to build blogrolls".
- S-035 · WordPress 3.5 release notes — Version 3.5, WordPress.org Documentation. https://wordpress.org/documentation/wordpress-version/version-3-5/ Retrieved 2026-09-06. Supports: WordPress 3.5 was released on 11 December 2012, and from it the Link Manager is "hidden for new installs and for any existing installs that have no links".
- S-010 · Pew link rot study — When Online Content Disappears, Pew Research Center (Athena Chapekis, Samuel Bestvater, Emma Remy and Gonzalo Rivero, 17 May 2024). https://www.pewresearch.org/data-labs/2024/05/17/when-online-content-disappears/ Retrieved 2026-09-06. Supports: 38% of webpages that existed in 2013 were no longer available at the time of the study, and 25% of the pages collected across 2013–2023 were no longer accessible as of October 2023, from a random sample of just under a million pages taken from Common Crawl.
- S-012 · PageRank paper — The Anatomy of a Large-Scale Hypertextual Web Search Engine, Sergey Brin and Lawrence Page, Stanford University. http://infolab.stanford.edu/~backrub/google.html Retrieved 2026-09-06. Supports: links are treated as citations, and a page's rank rises with the number and rank of pages pointing at it.
- S-032 · Google spam policies — Spam policies for Google web search, Google Search Central. https://developers.google.com/search/docs/essentials/spam-policies Retrieved 2026-09-06. Supports: the published definition of link spam, and that exchanging money for links is given as an example of it.
- S-031 · Google outbound link guidance — Qualify your outbound links to Google, Google Search Central. https://developers.google.com/search/docs/crawling-indexing/qualify-outbound-links Retrieved 2026-09-06. Supports: publishers are asked to mark advertisements and paid placements with
rel="sponsored".
Try the stumble and see one independent business at a time, with the date somebody last checked it printed on the page.