Is it runtime efficient to use machine learning to render webpages?

I am working on an app that's supposed to take a webpage and embed it with graphics as a part of a game app, and there I need to parse those webpages and render graphics manually, however I considered an option if it is possible to simplify this task with unsupervised machine learning. Normally I will need to do tokenisation, grammar parsing and semantic analysis, however I wonder, will it be more efficient at the runtime if I trained a ML model that takes HTML, CSS, JavaScript and the graphic input and train it to produce the graphic interface instead of computing all of that manually? The data is easily accessible, so if I opt for machine learning, will it be able to process the data faster than if I was interpreting it?