As well as making it clear whether you are flying IFR or VFR, it also demonstrates the importance of knowing the VMC minima!
That's true, and though I agree with the principle of your post, I think you're getting your figures mixed up. The minimum vis for VFR in class D is always 5 km. The standard cloud clearance of 1000 ft vertically or 1500 m horizontally applies unless you're <= 140 KIAS and <= 3000 ft amsl in which case the requirement is only to be clear of cloud and in sight of the surface, but the visibility requirement is not reduced (for fixed wing). (Rule 25(2)(b) and (c))
It doesn't seem unreasonable to me for a controller to assume that an aircraft is IFR-capable when inbound to an ATIS-equipped field in controlled airspace where the vis is less than 5 km. Nevertheless, FF's original point is well made.
And anyhow, how would a controller necessarily know the met conditions in flight anyway? I don't believe the radar set has yet been invented which can tell what the visibility and cloud-base are in a particluar spot in the sky... just because the visibility is 3km or less where the controller is (and if a radar controller, then he isn't looking out of the window anyway!), it does not necessarily mean that the conditions are thus where the aircraft is - which is why this is not meant to be guessed by the controller: they have no way of knowing!
In general you are correct, but for the specific case of an arrival (or departure) in controlled airspace the requirement is to use the reported met vis at the airport. A VFR arrival is not legal if the airport is reporting less than 5000 m vis. It's not clear from FF's post if the 3 km was his estimate or from a METAR. (Rule 24(3))