Prior to sending the actual physical letter you could have a prediction market where anyone can indicate whether they think the physical verification will be successful. 

The act of requesting verification would be a vote YES 

Anyone could take the opposite bet (NO, this location is not legitimate). 

Users who are good at consistently identifying false information would earn profit and be encouraged to continue. 

This may generate high signal data more rapidly than your physical verification process can scale 

Another benefit is that this could fund the verification process. 

Once both sides of the bet exceeds the cost of doing the verification it could trigger. The winning side splits the winnings (losing side bets - verification cost) pro rata