Requests for comment/Rethink redirects

This is a subpage; for more information, see the Requests for comments page.

Quick TL;DR

There are a few problems with current redirects in Wikimedia projects,
Redirects can be split into two broad categories: 1) alternative title/spelling/wording/grammar case, 2) redirects from a separate subject when a separate article on it doesn't exist.
Redirects of type 1 serve exactly the same purpose and practically duplicate information contained in Wikidata aliases, and redirects of type 2 all could be linked to Wikidata items.
Therefore, we can deprecate redirects of type 1 in favor of automatically redirecting based on Wikidata aliases, and it would solve a few problems.

Proposal

Scope

I thought of whether this RfC/essay belongs more to Phabricator or Meta, and, overall, I believe that despite proposing mostly only technical changes, they would affect many Wikimedia workflows.

Introduction

We all are (probably) familiar with the concept of redirects — very simple pages designed to redirect you from a title you entered/clicked to a title where an article is located. They come from the paper encyclopaedias, where redirect entries (called variously from "blind entries" through "direct cross-references" to "see-references") are used to direct the reader to an article from an alternative name.

In Wikipedias and other Wikimedia projects, this is not generally true. This essay focuses on main space redirects, the disadvantages of the redirects' current system, and suggests how it may be drastically improved with help of Wikidata aliases.

Why we need redirects?

There mostly are a few cases:

The article's subject is also known under another title: British Prime Minister -> Prime Minister of the United Kingdom;
The article's title can be spelled differently — words, capitalization or punctuation: Aluminum -> Aluminium, Albert einstein -> Albert Einstein, Merriam—Webster/Merriam Webster -> Merriam-Webster, Lenin -> Vladimir Lenin;
The article could be searched in the article's subject "native" language: Україна -> Ukraine, Владимир Ленин -> Vladimir Lenin;
The article's title and the redirect one are a plural/feminine/masculine form of one another (quite a rare case, but a bit different from others): Brakes -> Brake, Actress -> Actor;
There is not a separate article about searched subject: Bonnie Parker -> Bonnie and Clyde.

It is uncommon to have redirects from typos.

It is interesting that English and Chinese Wikipedias appear to be the only large projects (one of the most common redirect templates, Template:R from alternative name, is presented in only 25 other projects, which are all smaller) to have a redirect "tagging" system presented in form of templates on redirect pages, telling what kind of a case it covers.

Why current redirects are bad?

I am personally an active redirect creator in Russian Wikipedia and I've myself probably experienced all the weird quirks and problems of the redirect system. From them I can list a few:

The main one — unfortunately, almost nobody cares about them. Many pages don't have them at all, even in very obvious cases.
As double redirects don't work (and are converted to direct redirects on most projets), there is sometimes a huge problem when pages are created in place of redirects (mostly Case 5) — redirects (e.g. from alternative spelling) that were "meant" to direct to that page still continue to point to the old one. As an example, ru:Чёрная шляпа (black hat) article was recently created in ruwiki, previously pointing to Хакер системы безопасности#Чёрная шляпа — but, for example, an alternative spelling ru:Черная шляпа continued to point to it, until I manually fixed it.
Redirects aren't really articles. Yes, there are also disambiguations as part of non-article main space pages, but for them we at least have a way to properly model them in Wikidata — we don't have that for redirects (except, again, Case 5 ones).
Quite a small one, but sometimes you need to create redirects in different projects (Wiktionary, Wikibooks, etc) in the same language. Redirects mostly only depend on the language, not on the project (I don't have any examples of project dependence — can anyone present some, if they exist?).

How to make redirects good?

I thought of some system that would host redirect titles for all pages, even for not yet existing ones, would have the same rules for titles as Wikipedias do (at least Cases 1-4), would automatically track pages creation, and store titles per language, not per project. But wait... as you might have guessed by now, there already exists such a system! Meet Wikidata aliases.

(In a nutshell: on every Wikidata item (= Wikipedia subject) for every language there exist a label and a list of aliases, that are all the names the item could be looked for.)

Let's again visit the different redirects' cases:

This is exactly one of the purposes of aliases;
This is again exactly one of the purposes of aliases. There's one exception in capitalization — that on WD the case of first letter matters instead of Wikipedia, but it doesn't break anything as on Wikipedia titles are case-insensitive and the first letter automatically becomes capital;
This is what mul language is for on WD (mul label + mul aliases are included when generating aliases for any language);
This is again specifically one of the purposes of aliases;
In my opinion, this is the only case where we need main space redirects on Wikipedia — note that this kind of redirects also can be safely linked to Wikidata (using "sitelink to redirect" or "intentional ..."), so then we get as a bonus that all main space Wikipedia pages deserve Wikidata items. If we have a "double redirect" of Case 1-4+Case 5, it can also be placed as alias in this redirect's item (note specifically that it's different from the redirect rarget's item for Case 5).

Note that the only type of redirects that might have problems with aliases system are the redirects to another's article section (because it can't have a Wikidata item). Thankfully, this isn't actually a problem too — because all of these redirects (again, as far as I know — can anyone present any examples of the contrary?) belong to Case 5 or Case 1-4+Case 5 double redirect.

Why is this better? Again, to see this, let's look on the list of problems:

Well, for this one not much would be changed — except that it's generally easier to make mass bot edits on Wikidata than mass bot creations;
This problem would be completely solved because such double redirects would be "placed" in the redirect subject's item;
Now every main space page would correspond to a potential Wikidata item;
This would also be solved by the fact that Wikidata aliases are project-independent.

One of the possible disadvantages I see is that it is uncommon to have aliases in Wikidata with parenthetical disambiguation (and same thing with labels) — I don't think there are much redirects from such a disambiguated name (as such names are highly unnatural and therefore unlikely to be accidentally accessed), instead of redirects to such a name (for which there are no problems), though, granted, I don't have much statistics on that and I'm looking for it too.

UPD: in discussion it was also noted that sometimes an alias is presented in multiple Wikidata items, but a redirect may "favor" one of them (example: Einstein -> Albert Einstein). This might also be a case when it's better to keep the local redirect.

How to implement this proposal?

I haven't thought that much about this, but here's a rough outline how I see it:

Make a change to MediaWiki that, when a title X in a wiki project Y is accessed, and the following criteria hold:
1. There is no page X in Y,
2. There is exactly one Wikidata item Z, which includes X as its label/alias in Y's language or mul,
3. Z is linked to page X2 in Y,

user visiting X would be autoredirected to X2;

Make a simple way for editors to "export" old-style redirects:

For non-Case 5 ones — deleting one locally and creating an alias in WD for the redirect target's item. The editor needs to select whether the alias goes to project's language or mul, and could it be label instead of an alias,

For Case 5 redirects — linking the redirect to corresponding WD item;

When the number of local redirects would sufficiently decrease, deprecate non-Case 5 local redirects;
Continue exporting efforts until all local redirects would be confirmed to be Case 5 redirects, and, preferably, linked to Wikidata. At this point, forbid non-Case 5 local redirects on policy level (and display a warning when creating any local redirect);
If at some point we manage to get 100% Wikidata coverage for Case 5 local redirects, technically forbid creating local redirects without a Wikidata item linked.

I would very much appreciate feedback on this plan too!

And the categories shall be next.

Discussion

This is not a voting, this is a constructive discussion. I strongly encourage you to not only post your comments below, but also mercilessly edit the above essay. Well very well (talk) 11:33, 18 May 2026 (UTC)[reply]

Good idea, sounds very well to reduce clutter on number of articles. 1/8 of the total article count on ENWP may be redirects. Ahri Boy (talk) 13:38, 20 May 2026 (UTC)[reply]

This idea is constructive, and I agree the redirect system should better integrate with Wikidata. However, there are a few issues:

Redirect and disambiguation are the "twin pillars" of authority control on Wikipedia level. Obviously some Wikidata items will share the exact same alias which needs to distinguish, the software should do nothing but detect the conflicts. We also need to keep on-wiki flexibility—for instance, we can create disambiguation pages or still create a redirect and use a hatnote on the target page.
Redirect category ("tagging") templates can sometimes provide sources for aliases, whereas Wikidata currently doesn't support references for item aliases. This potentially lead to abuse. While Wikipedia has established RfD/AfD processes to handle bad redirects, Wikidata lacks an accessible venue to discuss deleting specific aliases in specific languages. We would need a much better management and participation system.

Despite these concerns, I don't see them as deal-breakers. The ultimate bottleneck is technical: the overhead of this approach is likely too high. I am highly skeptical about the feasibility of querying Wikidata on every view for every red link just to check for alias matches. If the primary goal is easier maintenance, a more realistic approach would be using a bot to periodically sync Wikidata aliases into local redirect pages. PexEric (talk) 08:20, 24 May 2026 (UTC)[reply]

@PexEric
Obviously some Wikidata items will share the exact same alias which needs to distinguish, the software should do nothing but detect the conflicts.
Yes, that's currently written in the proposal.
Wikidata lacks an accessible venue to discuss deleting specific aliases in specific languages. We would need a much better management and participation system.
Maybe a new RfD-like page on Wikidata+some gadget on Wikipedias that would allow to open a discussion there in one click (wouldn't be too hard, I think)?
I am highly skeptical about the feasibility of querying Wikidata on every view for every red link just to check for alias matches.
Hm, I think this would be solved pretty easily just by adding to Wikidata pages table alias as another primary key (or creating another view), no? This would make overhead comparable to just opening a page based on its name. Well very well (talk) 03:02, 31 May 2026 (UTC)[reply]

I oppose abrogating individual communities' ability to make their own decisions about what should redirect where. * Pppery * _{in solidarity} 19:59, 24 May 2026 (UTC)[reply]

I think this can be an excellent tool that can relegate a lot of work of creating and supporting the redirects. While some may have reservations, I think that it is simply possible to keep old redirects if they wish for that to be so. I think that this suggestion could support and enhance a quite big chunk of the mundane function of redirect. --Duppertip (talk) 19:48, 7 June 2026 (UTC)[reply]