Abstract Motivation Antibodies are widely used experimental reagents to test expression of proteins. However, they might not always provide the intended tests because they do not specifically bind to the target proteins that their providers designed them for, leading to unreliable and irreproducible research results. While many proposals have been developed to deal with the problem of antibody specificity, they may not scale well to deal with the millions of antibodies that have ever been designed and used in research. In this study, we investigate the feasibility of automatically extracting statements about antibody specificity reported in the literature by text mining, and generate reports to alert scientist users of problematic antibodies. Results We developed a deep neural network system called Antibody Watch and tested its performance on a corpus of more than two thousand articles that report uses of antibodies. We leveraged the Research Resource Identifiers (RRID) to precisely identify antibodies mentioned in an input article and the BERT language model to classify if the antibodies are reported as nonspecific, and thus problematic, as well as inferred the coreference to link statements of specificity to the antibodies that the statements referred to. Our evaluation shows that Antibody Watch can accurately perform both classification and linking with F-scores over 0.8, given only thousands of annotated training examples. The result suggests that with more training, Antibody Watch will provide useful reports about antibody specificity to scientists.