In the general case, there are no reliable independent studies that I can find, that conclude there is a measurable difference between online and offline dating-sourced marriages as far as divorce rate is concerned.
That said, there are bits and pieces of data that show, whether offline or online, that certain qualities, habits, interests, schedules, and demographic matches tend to have an effect on longterm viability of a marriage.
For example, take eHarmony.com – this is basically a "niche" dating site, where the niche is "primarily observant Christians looking for longterm relationships and marriage". Compared with the rate of offline churchgoers who marry then divorce, eHarmony's rates map very well onto this demographic – though they are lower here as well (no doubt due to their matching system helping to weed things a bit further).
So, on the one hand, they can clearly say "for our target demo, we improve the likelihood of a lasting marriage, on average"… But it's less clear that their demographic is "the pool of all singles", or "everyone looking to get married", etc. Hence, there appears to be some form of selection bias at work skewing the results.
More thorough testing and reporting needs to be done in the industry and shared publicly – but online dating companies have historically been VERY tight-lipped on data sharing, with the occasional exception like POF.com or OKCupid.com (both free sites, with much less to lose).