Page MenuHomePhabricator

Article Guidance: Explore how to best provide guidance while editing
Closed, ResolvedPublic8 Estimated Story Points

Description

As part of Article Guidance (T396029) we plan to help junior editors when they try to create an article. For the initial intervention (T414409), the main focus was on the steps that take place before the user creates the contents on Visual Editor. This ticket is focused on exploring how to support users while they are editing, and the proposed solution will be evaluated in a research study (T414812).

Problem statement and audience

Junior editors try to create a new articles on mobile, and often get their contents deleted by the community because the initial version they create does not meet the community quality standards. Example scenario: A junior editor reading the article about oasis finds that "Kitowok", a particular oasis the user knows is listed as a red link. Taps on it to create a new article and gets a blank page with no further support about which contents could be added to make a good-enough first version.

From the different situations leading to deletion, this exploration is focused on the problem of low-quality but valid articles (which could survive if they were of better quality). Rationale:

  • Entry point observations (T414535#11620008) suggest that it may be common on mobile to create articles through red links (which usually reflect topics in demand by the community), and deletion rates are still high.
  • The Article Guidance workflow (before reaching Visual Editor) is already intended as a filter: sending users away when they try to create an article that is not valid or has a high risk for notability issues (e.g., requiring biographies to exist in other wikis to allow their creation, which makes it hard to create an article about yourself)
  • Precedents like Content Translation have shown how reusing existing content created by the community, lowers the barrier for creation and results in articles that are more likely to survive the review process.

This exploration is focused on assisting junior editors while they start creating their article inside the mobile Visual Editor. The exploration revisits some of the ideas that were identified as promising in the initial research study with the additional goal of providing an integrated experience in the Visual Editor ecosystem. So that the insertion of new contents with Article Guidance and their review with EditCheck work in harmony.

Ideas explored

We identified several opportunities to provide better guidance while creating contents in Visual Editor. Some of these were explored earlier, and the initial research study showed they were promising. The specific ideas were:

  • Common paragraphs. Insert community-defined common paragraphs for users to fill the blanks and setting the expectation for the need of references (and illustrating how they are used).
  • References from other projects. Support for exploring and reusing references from Wikidata and other Wikimedia projects which may contain relevant informationon the subject of the article.
  • Verified facts from Wikidata. Quickly adding verified facts (data with references) from Wikidata.
  • Common Sections. Add community-defined common sections for the type of article right at hand. Without the need to find good examples of articles of the same type or documentation from the relevant Wikiproject.

One of the challenges we want to explore in this task is how the proposed guidance can be supported in a way that is consistent with the existing products in this space such as EditCheck. The design exploration will consider scenarios where the user experiences both features. For example, inserting some contents with Article Guidance and getting suggestions to improve them from EditCheck.

Proposed solution

Discovery

Add action. An action ("+") in the rail (always at the same line where the cursor is) invites users to add quality contents.

Use of the side rail. The side rail is used as the space where opportunities to improve the content are surfaced. This space is currently used by EditCheck to show opportunities to improve existing contents, and the proposal expands its meaning to include also opportunities to add new contents too.

Placeholder for initial discovery. A placeholder message ("Start writing or tap the +") introduces the possibility to insert content when the page is empty. This is presented as optional, allowing users to freely type their contents if they prefer.

Bottom panel with options. Selecting the "+' action leads to a bottom panel with different options to reuse relevant content for the particular subject and/or type of article: outlines with common sections and paragraphs, verified facts, and references.

Screenshot_1774532386.png (1,280×2,856 px, 188 KB)
Screenshot_1774532416.png (1,280×2,856 px, 214 KB)
Screenshot_1774532424.png (1,280×2,856 px, 225 KB)

Outline

The outline provides an overview of possible aspects to cover for the article that are often relevant for that type of article. This allows users to understand what the community generally expects for the articles of this type to cover, without having to chase similar articles to infer this structure. The contents available for each type of article are customizable by the community.

Common sections. A list of section titles with a description of their purpose is used to suggest useful general aspects to cover in the article. Sections can be used to suggest the section title or also as a container for common paragraphs.

Common paragraphs. Smaller units of content to illustrate specific knowledge pieces that may be useful to cover. This helps users to learn what to cover and how to capture it in the expected way. Selecting a common paragraphs results in new content being added to the article. Contents added are based on regular text that users can edit normally (not a template block that limits the ways to adjust it). The content for the paragraph can act as an example or instructions, and editors are expected to edit those further. To facilitate the process of completing information, optionally, placeholders can be included.

Text placeholders. Placeholders represent a gap in the paragraph for the user to fill. The gap includes a description that helps to convey what is expected to be filled. When using a placeholder, the text filled becomes regular text (not a template that makes it hard to edit).

Citation placeholders. A particular type of placeholder is the "add citation" one. This helps to convey that a citation is expected at a certain place. This illustrates the purpose of citations: the system asks for the data and where it comes from. It also complements the EditCheck work by making it less surprising to encounter future warnings about missing references (since users know in advance that those were expected).

Screenshot_1774532416.png (1,280×2,856 px, 214 KB)
Screenshot_1774532433.png (1,280×2,856 px, 274 KB)

Verified facts

A list of data from Wikidata helps users to have information about the article subject at hand that would be harder to find otherwise for the less experienced users and reduces efforts to copy information over (which is especially tedious on mobile). In addition, the data added comes with references, which helps the content to be of a higher quality and helps to illustrate the role of references.

Screenshot_1774532454.png (1,280×2,856 px, 176 KB)

Screenshot_1774532513.png (1,280×2,856 px, 262 KB)
Screenshot_1774532520.png (1,280×2,856 px, 302 KB)

References discovery

A list of Eikimedia projects that contain references associated to the article subject. This helps users discover references that may be relevant to write the article. The "References" view of the panel is connected to the standard dialog to insert references, which will be evolved:

  • Discover. A new tab including references from other Wikimedia projects that users can read and insert into the article.
  • Re-use. This tab will be expanded to include the references that the user provided upfront as part of the earlier steps of the Article Guidance process.
Screenshot_1774532470.png (1,280×2,856 px, 214 KB)
Screenshot_1774532482.png (1,280×2,856 px, 333 KB)
Screenshot_1774532477.png (1,280×2,856 px, 277 KB)

EditCheck interaction

The proposed system is intended to be compatible and align with the use of EditCheck. After creating a paragraph, a check can surface issues about the tone, lack of references or any other aspects to review. This applies to both, contents inserted with Article Guidance, as well as, contents written by the user.

When a paragraph has issues, Edit Check takes precedence in the use of the side rail. That is, when a paragraph has an issue, only the EditCheck icon will show next to it. Even when the input cursor is there, the "=" icon will not show. In this way we avoid crowding the space and encourage fixing contents before adding new ones. Users can still get the option to add contents on a new line or for paragraphs without issues.

Screenshot_1774532901.png (1,280×2,856 px, 274 KB)
Screenshot_1774532989.png (1,280×2,856 px, 315 KB)

A custom EditCheck can be defined to encourage the completion of placeholders.

Screenshot_1774532904.png (1,280×2,856 px, 233 KB)

Event Timeline

Pginer-WMF renamed this task from Explore how to best provide guidance while editing to Article Guidance: Explore how to best provide guidance while editing.Jan 14 2026, 4:45 PM
Nikerabbit set the point value for this task to 8.Apr 2 2026, 10:14 AM
Nikerabbit moved this task from Incoming to Product Signoff on the LPL Hypothesis board.

A user of the pre-VE workflow shared this feedback:

the pre-filled article contains “[Country name]” repeated multiple times. Is there a way to define this once so that all instances are automatically replaced, or to have it pre-filled based on the article title?

This connects with some earlier ideas that we have not illustrated in this exploration yet: providing suggestions to quickly fill the blanks in the paragraphs. These could be based on the name of the blank and the information available, including data from Wikidata as well as previously filled blanks by the user.

Another user shared feedback around the idea of reducing repetition when filling gaps with the same purpose and reusing Wikidata contents:

Outlines contain repeated elements like [Full name], [year] or genders. I think it would be better user experience if before opening visual editor an outlined-customed form was displayed allowing to fill all these fields. I know of an old tool (in french) that offered such forms for mangas and animes articles : https://abda.toolforge.org/formulaire.html. There doesn't need to be as much fields, but at least name and gender would reduce repetitive

And when a Wikidata item is found, most of these fields could be filled using wikidata data.

I'm just curious to know why throughout this project Wikidata is being evoked. Perhaps for the smaller Wikis where traditional and scientific articles are still needed, there is a possibility that there exist some entries in Wikidata that came from the en.Wiki, but for the average new article in a day's average intake of around 500 new articles on today's en.Wiki I would be surprised if any Wikidata already exists. Fine if it can be done automatically, but that detracts from the KISS principle. and it seems to be trying to do too much.

I'm just curious to know why throughout this project Wikidata is being evoked. Perhaps for the smaller Wikis where traditional and scientific articles are still needed, there is a possibility that there exist some entries in Wikidata that came from the en.Wiki, but for the average new article in a day's average intake of around 500 new articles on today's en.Wiki I would be surprised if any Wikidata already exists. Fine if it can be done automatically, but that detracts from the KISS principle. and it seems to be trying to do too much.

Thanks for the feedback. Happy to provide some context. We identified Wikidata as a potentially useful resource from multiple angles. Sharing some thoughts:

  • Wikidata may contain relevant information and references about the article subject. For cases where a Wikidata item exists, Wikidata represents a potentially useful set of information and references that can be reused. Reusing this information can be convenient to both reduce effort (particularly relevant for mobile-only users, who would be excluded otherwise), and illustrate the role of references (helpful for newcomers.
  • Encouraging a fact-first process. When a topic is not covered either on Wikipedia or Wikidata, some users may find it beneficial to start by creating the Wikidata item first. Starting with the facts (and the references supporting them) may be a useful way to check whether it is an item worth documenting in the Wikimedia universe before documenting it on Wikipedia where additional considerations take place. Contributing to Wikiedata facilitates the work for editors in other languages to reuse the efforts of finding the info and relevant references.
  • While it may be more relevant for non-English Wikipedias, Wikidata info can be useful for many cases in English too. English has several million articles more than other Wikipedias. This creates a big opportunity for topics with existing content that could be reused when covering them in other languages. Due to the coverage difference, one may think that English Wikipedia already covers most of the topics covered in other languages, but research shows that the coverage may be lower than expected: "English, contains only 51% of the articles in the second-largest edition, German". So there may be a significant number of items covered on Wikidata that can become potential articles on English Wikipedia. Note that the research mentioned is from 2014, when we started working on Content Translation. In addition, even for topics that already exist on Wikipedia, reusing Wikidata contents can be useful to expand them.

Hi,

I realize this project has been entrusted to a department primarily concerned with content translation, and and after having read the research paper by Hale, I'm left wondering whether the main focus of this project is to grow the corpora through cross-language article alignement (perceived need), or to help new users create acceptable articles on truly encyclopedic topics (established need).

Hale posits some interesting hypotheses about multi-lingual article writers but the question it raises is whether this has significant impact to require immediate attention and if the cost/benefit prediction will provide the expected growth. Hale's statistics are 10 years old, that’s a lot in Wikipedia time. Hales own references are to books and papers that are even older, and are primarily about cross-cutural linguistics. Three of the cited sources are to his own works.

There are other pressing priorities among the communities to ensure that the quality of articles is the most important consideration rather than the pure growth that is of more interest to the WMF; the effort by so many volunteer authors of articles of true value is to add knowledge first and not necessarily out of a need to grow the corpus.

Many German and French articles for instance just cannot be translated for use in the en.Wiki because their references are of a standard that is too low for the criteria of the WMF’s flagship project - as I am aware myself as a translator.

Rapid changes in IT (eg. AI created articles, and spam) are producing serious challenges that require quality control at levels where existing systems and their volunteers can cope. At the moment, TheEnglish Wikipedia’s control of new articles is close to collapse. (https://en.wikipedia.org/wiki/Wikipedia_talk:New_pages_patrol/Reviewers). We know why less than 10% of the 800+ New Page Patrollers are prepared to review them, but we do not know why these reasons are not of interest to the WMF departments that could address them. Like many others I do emphasise that a closer collaboration with the volunteers of an appropriate background and drawing on their experience would drive the project forward towards addressing both goals.

Regards,

Chris
(user:Kudpung)

Hi,

I realize this project has been entrusted to a department primarily concerned with content translation, and and after having read the research paper by Hale, I'm left wondering whether the main focus of this project is to grow the corpora through cross-language article alignement (perceived need), or to help new users create acceptable articles on truly encyclopedic topics (established need).

Thanks for your input @Kudpung. The goal of the project is definitely to "help new users create acceptable articles on truly encyclopedic topics". We just launched an early version of the feature as an initial experiment on Simple English Wikipedia and a few articles have been created already, as well as some cases where potentially invalid ones have been prevented form creation.

The availability of Article Guidance in Simple English Wikipedia also allows English Wikipedia communities to try the feature in a more realistic environment and point to specific that may need improvement as the feature gets ready to better support the English Wikipedia needs.

Hale research was quite useful when we started Content Translation over ten years ago. That was in the context of translation. When comparing article creation with and without translation, those are very different processes, but there are some insights and learnings that may still be useful about supporting a guided process. In this case, I was only pointing at the value of Wikidata for new articles in some cases.