Here’s the abstract of a study summarizing the reliability of findings concerning induced earthquakes. The evidence is credible.

Claims of industrially induced seismicity vary from indisputable to unpersuasive and yet the veracity of industrial induction is vital for regulatory and operational practice. Assessment schemes have been developed in response to this need. We report here an initial assessment of the reliability of all globally known cases of proposed human-induced earthquakes and invite specialists on particular cases to refine these results. 1235 cases were assessed, requiring over 1000 h of work. From the 881 cases for which scorable evidence is available, we class 87% as ‘Confidently Induced’, 10% as ‘Probably Induced’, 2% as ‘Equivocal’ and < 1% as ‘Confidently Natural’. The most seismogenic activities are fracking, research, geothermal, water reservoir impoundment, conventional oil and gas. Least seismogenic activities are construction, deep penetrating bombs, coal bed methane. 354 cases (29%) lack enough information to be assessable. Future work could include applying data mining techniques including natural language processing and AI to uncover new evidence. Future best practice for rapid assessment of cases would ideally involve an independent panel of scientists who rapidly apply a questionnaire scheme, reach consensus, and inform a response.

If you are the scholarly type, here is the PDF of the study from Nature.com.