It should be possible to automatically identify dubious news sources—but we’ll need a lot more data.