Google Scholar has offered OpenURL links for a while on a trial basis, and now the service is open to any library. We’ve started to tinker with it but we haven’t turned it on yet. Over the last couple of days there has been a lively discussion on the support list for our link resolver, and a few points have become clearer to me as a result.

  • Google still only provides OpenURL links for a minority of citations – apparently just those with reliable identifiers such as DOIs or PubMed ids
  • Google wants to know what we subscribe to before they provide OpenURL links for our users, so that they can distinguish links that will get you full text and links that won’t.

Our link resolver can now generate holdings lists in the format Google wants, so this isn’t a technical challenge. But, as many people have pointed out, Google wants to provide its OpenURL service in a different way from all our licensed vendors. We don’t provide a holdings list to them, so why should we for Google?

Perhaps the question should be, why don’t we provide this list to all our vendors? We know our users want full text, and many (most?) will accept no substitute. Why not make the presence or absence of full text visible at the source? That would be wonderful if it worked; but given all the unresolved problems of citation linking (e.g. title matching in the absence of an ISSN, determining coverage when the citation contains a volume but not a year or vice versa, distinguishing supplementary issues from normal ones, etc., etc.), it’s very likely that Google will make some wrong determinations, and promise full text we can’t deliver, or deny full text that we can. Our link resolver makes those mistakes too, but at least it makes them in a context where we can provide fallbacks – links to the journal level, links to chat reference, etc.

Of course, we haven’t asked our users yet which approach they’d prefer, and I’m not sure we can yet articulate the choice in a way that would make sense to them. Google thinks they know, and they know a lot about users. But their approach, it seems to me, is designed to protect Google Scholar from any appearance of failure. Don’t link if you can’t be sure of full text; and the residue of bad links will tend to be blamed on the link resolver, not Google Scholar. This will provide good Google service to Google Scholar users, but is it as good as the service we the libraries could provide?

In short, Google wants to treat Google Scholar users as their users, but there’s reason to suspect that they may be better served if we treat them as our users. Since we’re the ones providing access to the licensed resources, and thereby adding immensely to the value of Google Scholar, we ought to have more of a say in how the service is configured.

To take control, we’d have to have a library-based way of adding OpenURL links to Google Scholar. Several options exist, from my rudimentary Firefox plug-in to our Wag the Dog project and Ross Singer’s Web Localizer, and Openly’s OpenURL Referrer. All of these depend on client-side processing, so none of them are as simple as turning on Google’s service. They may still repay the effort, however, by providing consistent OpenURLs for more citations.