Abstract: Systems and methods are provided for automatic resource replacement in web pages. In one embodiment, a method comprises using a machine learning model to select a replacement resource for a broken resource, in response to a presence of the broken resource in a first web page. The machine learning model may be trained to identify the replacement resource based on similarity in characteristics. In this way, resource errors may be mitigated in real-time and user experience of accessing web pages is improved.