Papers automatically tries to match all your articles to metadata it pulls from the internet. This is why Papers needs an active internet connection to perform this task in the background, and you will see matching stalls when you are offline.

We frequently receive questions about matching and about the quality of metadata being added to library entries. This warrants a short explanation of how metadata is received and handled by Papers.

What Papers does not do

First, we should cover how Papers does not extract metadata from your articles. Basically, Papers does not scan your articles and tries to guess what the title is, who the authors are, and what the publisher was. When you drag and drop articles from your hard drive into Papers, you will see metadata associated with some articles and maybe not with others. Regardless of whether you are online or Papers attempted to match. That happens when the repository from which you originally downloaded the article embedded the metadata into the PDF. This is the case for some, but certainly not all, articles.

What Papers does do

Now that we have established how matching does not work, we can briefly touch on what Papers does and this will lead us into some of the limitations of matching. Without getting into technical details, Papers pulls metadata information from an online repository. Which repository, depends on the user. The choice of repository used will also affect the quality and type of metadata that Papers can find. Google Scholar is an example of a repository that typically does not provide complete metadata. This is why some users experience missing authors or partial journal names in some of their entries after matching with Google Scholar.

The best strategy to ensure the best possible metadata being included with your articles, is to use a repository that is geared towards your field of study. For the natural sciences, PubMed is a great choice for example.

We have a knowledge base article on Matching available here.

Why some articles cannot be matched (easily)

Metadata tends to be more complete and easier to get for articles that have been published relatively recently (within the past 5-10 years). Old articles are less likely to have metadata associated with them, and unfortunately there is no easy solution for this, aside from manually editing metadata.