i've thought of this many times.
"Wikibot, tell me more about bridges of the 19th century." <blah blah summary>
"Stop. Tell me more the Ponte delle Catene, Bagni di Lucca bridge"
<blah blah details>
Even simpler, I wanted my iPhone to read a wikipage out loud to me while I was walking home in the cold, and I was surprised that I couldn't figure it out - you can switch on the screen reader in accessibility, but it stopped reading after half a paragraph. Very annoying.
Like most things in ML, the technological exists and theoretically works, but ymmv in production.