WikiHow Sues OpenAI, Accusing It of Unauthorized Scraping of Over 11,000 Guides to Train ChatGPT
WikiHow has sued OpenAI, accusing the company of scraping over 11,000 "how-to" articles without permission to train ChatGPT.
The lawsuit claims that OpenAI copied its copyrighted, human-written, expert-verified guides through scraping or third-party datasets without permission or payment, using them to train models and enhance retrieval generation. As a result, ChatGPT often outputs substantial content nearly verbatim, leading users to bypass the original site, diverting page views and ad revenue. The lawsuit also accuses OpenAI of deliberately removing copyright management information, violating the DMCA. The case was filed in the Southern District of New York on August 21, 2026.
Market mechanisms indicate that this incident is escalating the dispute over content copyright and AI training data. Financial and legal risks are shifting towards generative AI companies, while content platforms that adhere to licensing models benefit, and those relying on large-scale web scraping for model training face pressure. The lawsuit may encourage more publishers to seek similar remedies or licensing negotiations.
Source: Public Information
ABAB AI Insight
WikiHow's core asset is its large, structured "how-to" content library, and it accuses OpenAI of using it as training material and outputting substantial content without compensation, directly impacting its traffic and monetization model. The accusation of removing copyright management information further strengthens the claim of intentional infringement. This case continues a series of lawsuits between publishers and AI companies regarding the use of training data.
From a capital perspective, content platforms seek to prevent unauthorized use or obtain compensation through copyright lawsuits, while AI companies rely on vast amounts of publicly available web pages to build their models. This follows a similar path of other media and authors suing AI training. We are currently in a phase where the legality of data acquisition for generative AI is being systematically challenged.
This is analogous to the early disputes between search engines and content sites over scraping, but the output substitution effect is more direct.
Essentially, this reflects a change in regulation. The mechanism is that copyright law is being used to define the boundaries between "training use" and "substitutive output," forcing AI companies to shift from implicit scraping to explicit licensing or face higher litigation risks.
ABAB News · Cognitive Law
When models can directly output the essence of the original text, traffic and ads will be diverted from the source.
Unauthorized large-scale scraping will ultimately return in the form of copyright lawsuits.
The moat of content platforms is shifting from "being indexed" to "being licensed."