The Open Source Initiative, a nonprofit organization that promotes and protects open source software, has published its annual ranking of the most viewed open source licenses in 2025, reflecting the preferences and priorities of developers, organizations, and open source communities worldwide.
At the top of the 2025 rankings is the MIT License, maintaining its long-standing position as the most sought-after open-source license. With approximately 1.53 million pageviews and 925,000 unique visitors, the MIT License’s permissive terms and minimal restrictions remain highly attractive for projects ranging from personal open source repositories to large-scale commercial products.
Is this LLM BS?
MIT and Apache 2.0 remain the most widely used open source licenses
After reading the abstract, I was interested in what the article says about that. And the only thing I found is a section about page views.
With approximately 1.53 million pageviews and 925,000 unique visitors, the MIT License’s permissive terms and minimal restrictions remain highly attractive […].
How can you jump from a statistics about license viewing to that it is the most widely used?
No, I did not miss anything because the author goes on with
Following the MIT License, the Apache License 2.0 ranks second in interest, drawing 344,000 pageviews
Well, you could also count packages in most Linux repos. You would reach the same conclusions.
Or, you could look at licenses on GitHub. The same story is repeated there.
I take it this collides with your assumptions?
The problem is not the conclusion*. The problem is that the method used to reach it is terrible.
As you say, you could look at license popularity on GitHub, and the author should have done something like that, but even those statistics have to be interpreted carefully:
Each data point corresponds to the rank of a license based on the count of unique developers who uploaded code to a repository subject to the terms of that license during a given quarter.
In other words, this measures how many developers are commiting code under each license, and thus is more of a reflection of the popularity of software under each license, rather than the licenses themselves.
Perhaps a more meaningful measure would be how many (unique) repositories are created with each license, since a developer commiting code to a repo does not mean that they favor the license of that repo. I couldn’t find numbers for 2025, but amusingly these totals from 2020 suggest that no license is the most popular license, followed by MIT and then Apache
* The majority of my own code is MIT, by a large margin-
Manjaro is more widely used than ubuntu based on page hits on distrowatch
“Most widely used licenses” is something that we can actually measure by counting (relevant) repos.
Page hits may be used to measure “popular” licenses since popular is subjective.
Licenses are chosen by devs, not users. License viewers are also users, not only devs. There are more users than devs. A fraction of users could distort the measure. At best page hits are a proxy but not a definite measure.
This whole article is hokey.
But assuming permissive licenses lead copyleft licenses, I think it’s unfortunate.