In magic shows the magician frequently invites members of the audience on stage to verify that the ropes are real, there really are no irregularities in the iron rings etc. But if you are an audience member not chosen, how do you know that the one going to the front is not a plant?
We can assume that the majority of the audience really is random; so all we need is for the audience member to be selected randomly.
Dice? But they may be faked. How about the last digit of some periodically changing market rate? That could work, if you happen to have a device that you trust that you can check. Don't use the wifi of the theater, though.
How about you allow everyone to shout out large numbers, one by one, and then they are modded by the number of seats in the audience, and the remainder is the volunteer who should go on stage (seat numbers have to be pre-determined). Without calculators this would be hard to game, even if the last few people shouting numbers were plants. Today it wouldn't be so hard. Of course, it would be hard to find plants smart enough to do this kind of math on the fly who would be willing to work there.
How about, the magician turns his back, and writes down a seat number. Then everyone in the audience switches seats. Then he shows the seat number, without having looked at the audience first? He could pre-arrange what seat he would choose, but it would be hard to guarantee a plant getting it every night.
Certifying Truthfulness of Institutions is another way it could be done - just have the magician make a public pledge before the show that all volunteers are genuine, non-coached random volunteers, and that if anyone can prove otherwise, he will pay them 100k. That would be quite a strong proof that the show was for real - and if the show really was fake, it'd be 100k sitting on the table for any insider to take whenever they wanted.
In probability terms, it reduces the space of all magician & helper combinations to only such combinations where the insider can resist a 100k reward. That space is much smaller.