It’s much easier to write cloze deletion prompts: you can just copy/paste some source text and delete some phrase. But I find that prompts created this way tend to result in shallower understanding. After several repetitions, I’ll remember the answer, but it often feels like I’m pattern matching (Spaced repetition memory prompts should be written to discourage shallow “pattern matching”) rather than deeply integrating the idea into my conceptual network. This is probably because cloze deletions usually contain lots of extraneous information one can use as “hints.”
Other factors which may be involved:
I’m not aware of any concrete evidence on any of these issues.
On the other hand, cloze deletions are an extremely efficient way to produce new prompts, and The limiting factor in spaced repetition system capacity is writing enough good prompts. If the choice is between cloze prompts and nothing at all, the cloze prompts are definitely better! They’re also more general: Cloze deletions can be used to create typical N-sided flashcards.
Note that these observations are about the impact of cloze deletion prompts on memory. Cloze deletion prompts may work fine for Timeful text applications like The mnemonic medium keeps readers in contact with material over time: that’s the theory behind Readwise’s emphasis on them.