OkCupid did an experiment telling users they were compatible when its scores said they were not.

These pairs carried on with conversations, apparently not noticing their supposed incompatibility. I wonder how many of them formed relationships and how those worked out.

It's possible that this experiment constituted lying to the users, but I don't know whether that is true. It depends on what promises the site makes to users about how it judges compatibility, which are facts I don't know. If "flip a coin" is as good a method as its usual one, perhaps it satisfies the promises.

However, if a dating site's claim to being useful is its method for estimating people's compatibility, and that method turns out to be little better than chance, it seems to imply that the site is pointless.

More information about the experiment.