Also think about the way algorithms (like WolframAlpha) interpret the structure of the questions. Like some of the other commenters, switching some words around makes WolframAlpha fail.

It might be interesting to come up with a methodology for question structure that is harder for algorithms to interpret...?