Where you are (physically, or virtually) are index points for information relevant to that place. An obvious example is any mapping software that uses where you are (via GPS, Cell triangulation, or Google's index of wifi MAC addresses) and positions you on the map. Combine with Street View or Microsoft's Street Slide and you can see where you are looking from pictures posted by others. Merge this with the image from the outward facing camera, and you now see the reality, augmented by overlays from sources you select. Simple examples from my recent vacation in Colorado include: an overlay looking at the Rocky Mountain National Park peaks from Trail Ridge Road that names the peaks, perhaps provides elevation and distance information. Try the same thing at Mesa Verde National Park and you could have text or audio of about the ancient Puebloian cliff dwelling you are viewing from a simple "Junior Ranger" view to an in-depth discussion from experts (anthropologists or the modern Puebloian perspective.) We are a short time-frame away from overlaid video sequences that can animate history in historical sites, or even fantasy stories that operate with 'boots on the ground'.
Folks like Blair MacIntyre, at Georgia Tech have been doing research for a while, and are looking to establish standards that open the door for platform independent AR content. It looks like W3C may take up this cause building on KML (ARML, KARML) -- with likely competing commercial interests that could delay things for a few years.
While the devices that will make this environment "essential" (see Vernor Vinge's Rainbows End) may include goggles/headsets or implants; in the short term we can get significant impact with the next generation pocket device (as cell phones, GPS units, cameras, tablets, et al converge) ... an interesting question is what components will be needed in said device?
An initial list - camera, display screen, GPS, wifi/bluetooth (optional phone link -- cell phones are an expensive channel for AR aps (IMHO) and won't work in many interesting locations), sensors for positioning (you are here, but where are you looking?), sound output ---- and of course the inner workings that will make it click. My A500 is a bit big for portable use in this context. You need memory to store the anticipated content elements given that you may not have online access as you move though a park or remote situation. I can envision parks adding bluetooth or wifi transmitters, as they now have 'cell' phone ones and/or have had AM radio points of interest to provide local content in some situations.
We need to consider how services evolve for this as well. Wikitude is more of an augmented map than an overlay on reality, but it does give us a sense for location based browsing. Many of the elements near my location have no value ... which is like web search unfortunately. If you want to track a given author/source, or look for history or geological or some other characterization of content -- this is not the tool. Just as we have web sites that link a set of logical pages together, similarly there is a value in having a linked AR facility as well.