Not necessary, because even if this isn’t typical, the argument applies to those people who do rely on the web to diagnose their medical conditions, however rare those people might be.

Necessary, because if it were not true — if people can rely on something besides scientifically valid information and still be unlikely to do themselves more harm than good — then the fact people can’t distinguish between scientifically valid and invalid information wouldn’t matter. Under the negation of (B), people who sometimes rely in invalid info because they can’t tell that it’s invalid won’t necessarily be likely to do themselves more harm than good.

Not necessary, because the argument concerns what happens to people who CANNOT distinguish between valid and invalid info. The author doesn’t need to think that people who CAN distinguish will do no harm to themselves by relying on the web.

The argument concerns people who rely on the web when trying to diagnose their medical conditions. But many people may browse the web who don’t rely on the web for diagnosis; the author doesn’t have to assume anything about those people.

The author believes that relying on the web for diagnosis when you can’t distinguish between scientifically valid and invalid info will make it likely that you’ll do yourself more harm than good. But the author doesn’t assume this kind of reliance is necessary to harm yourself. We might harm ourselves in other ways. This answer would be better if we replaced “only if” with “if.”