The complicating factor I realized, is that while DOIs are much more stable and reliable, the ACS page in this case is probably not Open Access, so many users wouldn't be able to see it if I change it to the DOI.
Summary: DOI: More likely to work over a long term (5, 10 years). Less likely to be available to the general public. Direct linking to a free source (Authors website for example): More likely to be available to the public now. However, if other information is not recorded (As in the above) then if that link breaks we are unlikely to be able to reconstruct the link.
I'd in favour of requiring a DOI with each link to an academic paper, in addition to any free link. What do the rest of you think?