Parser excels at text simulation due to allowing a model world with objects and rules affecting them. The simulation is theoretically only limited to what the author can create and how much data the game can manage. Older games like Oregon Trail and Lemonade Stand are essentially small-scale world and economic simulations with random elements and could be considered choice-narratives.
What can tend to get lost in larger parser simulation is authored plot and story. Not necessarily a bad thing as a good simulation will imply plot - like how in Dwarf Fortress a random affliction making an NPC more aggressive can suggest that the character has become the town hothead/drunkard with psychological problems based on the fact his house randomly burned down a year ago…that’s all inferred by the player based on random numbers and events and can totally work. Often the trick is allowing freedom between scripted plot or world events that are brought about based on the simulation.
Building that scale of simulation is a lot of work though. One of the reasons I transitioned from writing in parser to choice-narrative specifically was because parser games are so incredibly difficult to thoroughly test due to the world model and emergent gameplay that is provided. It’s similar to the difference between a linear shooter like Half Life and an open world game like Grand Theft Auto.
That’s not to say choice-narratives need to be completely linear. They require a bit more work to give the player more options than A, B, or C, but simulationist elements can be built in to offer more agency and “game” than a simple branching narrative.