On a hunch I ran a test game as Attila on a Duel map vs two Greece AIs. It seems Alex will always liberate another civ's cap. It worked 1, 5 and 10 turns after capturing as well as 4 times in a row. I suppose it's because his Diplomacy flavor is highest of all civs, so I'll definitely include him next time I try the Liberty strategy. Maria might be a good AI for it as well.
It isn't reliable. Liberation is based on many factors. Civs will raze or liberate cities they don't want, or if they can't support the happiness. An AI will almost never liberate another civ's *original* capital. But it's not reliable! In the past, Gandhi has almost always liberated non-capitals. In my most recent game, (where I wanted him to liberate) he razed instead. (And puppeted the capital when I tried that)
Luckily this is a quick G-Minor. I wouldn't bank on this strategy, unless you're willing to restart a couple of times. Also, I'm not sure if it would be counted against you in this case, but trading an AI a city and immediately recapturing it is banned in HOF play. So, if he *didn't* liberate it, and you were forced to recapture, it might screw you, delaying your cultural victory significantly.
However, I'm pretty sure that rule is only there to prevent gross exploits in Deity so it might not be considered an exploit here. (The exploit where you can trade cities for INSANE gold, DoW, recapture, trade again to another AI, etc)
Personally, I think the liberation of an AI for the express purpose of winning a cultural victory SHOULD be banned in HOF play. I've done this in the past in non-HOF play on deity, to get a sacred sites win,
![Smile :) :)](/data/assets/smilies/smile.gif)
but I've never submitted a game to HOF where I did it. (to my knowledge)
And this should include cities originally captured by an AI IMHO. For non-cultural victories, it shouldn't be considered an exploit. But it's against the flavor/spirit of the rules IMHO to rely on a civilization being wiped out and reborn to get the cultural victory.
However, it can be so challenging to pull off that I'm not sure it's worth banning. Because, as I said, it's very unreliable.