Your going to hate this answer .....
It depends on what I feel like .... its instinct more than anything. I play Gigantic / Immense, and specialise worlds into dedicated Eco/Tech/Production worlds, each of the three categories being built on a ratio of around one specialisation for four of the other specialisations.
The early ones like microrepair(Thingy - whatever its called) and Xanthium Hulls, the one's I know that the AI is going to have a go at - then I'll use up a Tile on the Manu Cap world (I tend to build a Manu Cap world asap because of this) to make sure I beat the AI to it. I usually write off the Diplomatic one, I dont bother the AI is manic about this one, I let the AI get it, I play Terran so doesnt fuss me if the AI gets it - if they are building that, gives me more time to get the others.
Outside of those two, the other relatively urgent one is Harmony Crystals, and again I might use a Tile on Manu Cap world to do that, but usually try and think ahead and use a smaller world, starting fairly early in the game, so there are circa 18/25 turns to get it - ie start it early and forget about it, it'll turn up in the end usually before the AI.
From that point, because I play Gigantic or Immense, I've usually got enough production worlds on the go to dedicate one to each of the remaining smaller ones (Ultra. Friction, Eye, etc), as long as I get them going fairly early it matters not its not on a high PQ world. That then frees up the manu cap world for the real big ones and arguably important ones such as Restaurant of Eternity - the manu world is pushing that one out whilst smaller worlds have the others ticking over in the background. The AI will always go for Restaurant, so the quicker I get Xan and MicroRepair out the way, and start Restaurant the better.
By the time Restaurant is done, I have lots of other Production worlds lurking, so they sort out any others left, and the manu world cranks out ships. Not very structured
but it works for me