Wikipedia:WikiProject AI Tools/Guide

This page provides informal suggestions for prompts you can try with LLM chatbot tools to gather ideas for improving articles, including identifying potential issues and useful sources, without using LLMs to generate article text.

The suggestions here are most appropriate for experienced editors who already write successfully on Wikipedia and are familiar with content policies and guidelines. If you choose to use LLMs, strive to serve as an ambassador for careful, constructive, and transparent use of LLMs. Before using LLMs for anything related to writing or updating articles, read:

• Wikipedia:Writing articles with large language models (guideline), which includes "the use of LLMs to generate or rewrite article content is prohibited, except for the uses as outlined below..."
• Wikipedia:Responsibly using large language models (explanatory essay)
• Wikipedia:LLM use disclosure (user essay)

Keep in mind: Review output, never copy-paste. Validate suggestions with your own research. Write in your own words. It's a tool; you're the writer.

Example prompts

Finding sources

• Prompts that WikiEdu suggests to students:
  • Suggesting sources to read: "I want to improve the Wikipedia article on [insert article title]: [link]. I have already consulted and used the following sources: [list of sources]. Could you suggest additional secondary sources — such as scholarly books, peer-reviewed journal articles, or reliable news publications — that meet Wikipedia’s reliable source standards and would strengthen the article’s coverage? Reliable sources are independent from the subject to avoid conflict of interest, and from a publisher known for fact-checking, editorial oversight, and neutrality, like academic presses."
  • Finding access to sources: "To properly cite and reference this study in my assignment, I need access to the full text of this peer-reviewed article: Yaghoobian and Kleissl, "Effect of reflective pavements on building energy use". Which academic databases or library services provide access to 2012 issues of Urban Climate?"

Copyediting and style

• "You are a professional Wikipedia proofreader. Your task is to analyze Wikipedia articles written in wikicode markup and identify issues. Spelling and Typos: Look for misspelled words, especially proper nouns, technical terms, and common words. Grammar and Style: Identify grammatical errors, awkward phrasing, run-on sentences, and violations of Wikipedia's manual of style. Factual Inconsistencies: Point out contradictory information within the article and claims that seem implausible. Do not report date inconsistencies unless they are clearly factual errors. Ignore wikicode formatting syntax (templates, references, etc.) - focus only on the actual article content. Provide specific examples and suggest corrections where possible. Organize your findings into clear categories. [Copy-paste article]" (Based on User:Polygnotus/Scripts/AI Proofreader.)
• WP:TONE: "I'm working on this Wikipedia article. Check for grammar, spelling, and readability issues, including punctuation, clarity, and inconsistency. Encyclopedic writing has a fairly academic approach, while remaining clear and understandable. Provide specific suggestions, but do not draft new article text. [Copy-paste article]"
• WP:TECHNICAL: "I'm working on this Wikipedia article. Check for any overly technical material. Articles should be written in encyclopedic style, not scholarly writing aimed at specialists. Articles should minimise unexplained jargon and not take prior knowledge for granted. Articles should be self-contained as much as possible, rather than relying on excessive links to explain unfamiliar concepts. Do not draft new article text, just list the issues. [Copy-paste article]"
• WP:LEAD: "I'm working on this Wikipedia article. The lead section is [too short / too long / too vague / too technical]. Provide specific suggestions for improving the lead. Do not draft new article text, just list suggestions. Guidance: 'The lead should stand on its own as a concise overview of the article's topic. It should identify the topic, establish context, explain why the topic is notable, and summarize the most important points, including any prominent controversies. The notability of the article's subject is usually established in the first few sentences. As in the body of the article itself, the emphasis given to material in the lead should roughly reflect its importance to the topic, according to reliable, published sources. Apart from basic facts, significant information should not appear in the lead if it is not covered in the remainder of the article.' [Copy-paste article]"

Validation and quality

• "I'm working on this Wikipedia article. Check for any major gaps in the content, including checking for any recent developments that are missing. Do not draft new article text, just list the gaps. [Copy-paste article]"
• WP:LLM: "I'm looking at this Wikipedia article that another person wrote. The use of LLMs to generate or rewrite article content is prohibited. Check for any signs of LLM-generated text. [Copy-paste article]"
• WP:RS: "I'm working on this Wikipedia article. Fact-check the content, using reliable sources such as scholarly books, peer-reviewed journal articles, or reliable news publications with a reputation for fact-checking and accuracy. Do not use Wikipedia. Do not draft new article text, just list the issues. [Copy-paste article]"
• WP:NPOV: "I'm working on this Wikipedia article. All encyclopedic content on Wikipedia must be written from a neutral point of view, which means representing fairly, proportionately, and, as far as possible, without editorial bias, all the significant views that have been published by reliable sources on a topic. Check whether this article is written from a neutral point of view. Do not use Wikipedia. Use reliable sources, such as scholarly books, peer-reviewed journal articles, or reliable news publications. Do not draft new article text, just list the issues. [Copy-paste article]"

Prompts used by LLM-powered tools

• Source Verifier: citation checking
• CNfirmed: finding and verifying sources (unfortunately finding sources doesn't work well)

See also

User essays with critical perspectives on LLMs:

• Wikipedia:LLMs are bad search engines
• Wikipedia:Why you shouldn't write articles with ChatGPT, according to ChatGPT
• Wikipedia:Case against LLM-generated articles