[Purposes] This study aims to investigate the publication status alternation of papers online first and reveal how the Regulations for Alteration of Publication Status and Content Correction of Online Publications on CNKI (Regulations for short) is implemented. [Methods] Firstly, we retrieved the withdrawal declarations of papers online first marked with "withdrawal" on CNKI from 2017 to 2019. Secondly, we downloaded them and saved the documents after coding. Then, the file directories were imported into Endnote X9 (a kind of reference management software) for statistical analysis and the texts of declarations were input into NVivo 12 Plus (a type of qualitative data analysis software) for cause classification. Finally, we analyzed the subject distribution, source journals, and causes of the "withdrawal". [Findings] From 2017 to 2019, the publication statuses of 389 articles online first have been altered, with 310 withdrawn, 74 retracted, and 5 corrected. Among them, only 278 (71.47%) state specific causes in the withdrawal statements and the causes are academic misconduct (32), authorship (25), publishing errors (82), quality problems (118), and others (21). [Conclusions] The online-first articles have a high proportion (389/182654, 0.21%) of publication status alternation and a small percentage of them (6/389, 1.54%) have strictly complied with the requirements of the Regulations, as manifested by the missing or vague causes for the publication status alternation in the withdrawal statements and even the deliberately covered-up academic misconducts. Academic journal publishers and digital publishing platforms should enhance the process management and content review of papers published online first and further improve the transparency of the causes of publication status alternation, visibility of withdrawal declarations, and traceability of the "withdrawal" process.