Google Co-op: Add Your Own Vertical Search To Google

Google said it would have a health-related announcement at today's Google Press Day -- but no, it's not Google Health. Instead, it's Google Co-op, a way for people to create specialized search engines by tapping into the main Google index or the option for searchers to pick preferred vertical search providers to show up in Google OneBox results. Yes, health information is one of the new features -- but this is more than Google Health. This is Google making a giant and somewhat perplexing leap into mass tagging.

Subscribed Links

Let's start in with the specialty or vertical search providers, what Google calls subscribed links. Many are probably familiar with how for some queries, Google will show what it calls a OneBox result at the top of the "regular" results. For example, a search on san francisco hotels brings up a section like this at the top of the page

Local results for hotels near San Francisco, CA
San Francisco Marriott - 1.0 miles NE - 55 4th St, San Francisco, 94103 - (415) 896-1600
Hyatt Hotels & Resorts: Park Hyatt San Francisco - 1.7 miles NE - 333 Battery St, San Francisco, 94111 - (415) 392-1234
Hyatt Hotels & Resorts: Hyatt at Fisherman's Wharf - 2.2 miles N - 555 N Point St, San Francisco, 94133 - (415) 563-1234

Those results are powered by one of Google's own vertical search engine, Google Maps (formerly Google Local). The new subscribed links service lets people choose other non-Google vertical search engines to show at the top of the page, if they want to.

It's a very cool idea. For example, say you are regularly searching for information about search engines and would like to know if Search Engine Watch specifically has any matching info along with searching the entire web for that topic. If we get our act together (and we'll try soon), you could make us one of your subscribed links. Then the next time you search for something where we have content, you might see our matches right at the top of Google.

Where do you find providers? Google's got a currently very small directory of them here to choose from. Preferred partners are already listed, partners that Google thinks people will be especially interested in, not those who have paid. No money is exchanging hands in either direction to be a subscribed link provider, Google says.

Digg is the only news provider listed at the moment. I subscribed to see how it works. Not too well. Perhaps not at all. Searches for Google, Playstation, Nintendo -- all topics on Digg right now -- brought nothing up. Hmm. I tried subscribing to People. Searches on Brad Pitt and Angelina Jolie gave me nothing.

Frankly, I don't think the system is working right just yet, as I'll get to further below. I also have a note out to Google about this. In the meantime, let's just pretend it's working. How do we at Search Engine Watch or anyone else get to be in that directory or a subscribed link partner in general.

Here's the guide that allows anyone to get started. I had to laugh at the intro:

The API was designed to be as easy to use as possible, and requires only basic XML skills. This guide will show you how to create subscribed links, with plenty of examples along the way.

I laughed because in short order, I was lost! Barry Schwartz, who is a programmer, still felt lost himself and said he'd through it at "one of his XML guys" tomorrow. In contrast, making a Google Toolbar Button is a heck of a lot easier. I sure wish making subscribed links were, because they are potentially going to be an important new way for people to ensure they are getting traffic from Google.

Anyone can make a subscribed link to offer on via their own site (though the developer guide doesn't go into details about this, such as how to place it). Naturally, what you really want is to be in the directory that Google itself offers. Again, the developer guide doesn't cover this. But this appears to be the submission page.

As for who gets in, Google told me that those included and featured in the directory will be based on user uptake. Get a lot of people subscribing to your results, and you'll more likely be featured to users.

Two last things on Subscribed Links:

First, another OneBox! Just how much can Google shove above the "regular" results. Google tells me that they are currently trying not to show more than two of their own -- so potentially, you might be looking at three in all on the page. You'll never see more than on Subscribed Links OneBoxes, and these will come before Google's own.

Second, if the entire idea feels familiar, you might be recalling Yahoo Subscriptions. That launched last June and is explained more in our Yahoo Search Subscriptions Brings Premium Content Into Web Search article.

I've no idea how popular the service is, but I'm guessing not much, given that most people seem never to have heard of it. Unlike the Google system, the number of partners is very small and there's no API allowing anyone to jump in. Instead, you've got to go the contact form route. I suspect Google's system will be far more popular, since it should have a much wider range of providers.

Labels, Google Health & Vertical Search

A second part of Google Co-op is the ability to allow people to label URLs into different topic areas. You mean tagging! Google still prefers the term label, while I'm still a hold out for saying categories. But whatever the name, it's not like the idea of tagging you might be used to at other places. This is industrial-strength tagging.

For example, with Yahoo My Web 2.0, I can tag any page with any words I prefer. The system is really designed for me to tag on a one-by-one basis. If I do a search, see something I like, I can click the Save button, add a tag, some notes and have that individual page stored for easy recall.

Yes, I can import many pages and assign them all tags en masse. But that doesn't seem to be the case for most people. The system currently has only 1.1 million pages tagged, hardly double the amount I recall it having not long after launching last year. If there were massive tagging imports, I'd expect the number to be higher.

In contrast, Google's label system is initially designed as a more mass tagging system for those who want to create vertical search engines. Google's now rolled out a number of these:

Let's dive into the health area. Sure, call it Google Health if you want -- though Google says a more full-fledged Google Health is coming and definitely doesn't call this Google Health itself. Whatever you all it, this health thing lets you search against pages that have been labeled with the help of contributors such as the Mayo Clinic or the Harvard Medical School as being health-specific.

Ideally, it means that I should be able to do a search and get back only stuff related to health issues. Here's an example. Say I search for cold on regular Google. The first link is for the musical band Cold, and the third link is for Cold Stone Creamery. That's great place for ice cream, but the only health connection is that it might make you fat! Midway down, Macromedia shows up because of its ColdFusion product, then there are two links on the Cold War followed by two links on Cold Mountain.

For regular Google, this variety is fine. Who knows what you want when you search for the word cold? It could be any number of things. But for a health search, you want to get rid of all that junk. Google Health's labels ideally should do this. But go there, then search and what happens? Pretty much nothing. The off-topic stuff I mentioned is still there!

I suspect there's a bug in the system right now. Google Co-op didn't go live when announced, and then it slowly came up. I'll check on this, and the better test will be in a day or so, especially when some of these new topical areas are pitted against existing verticals in various areas. But conceptually, hopefully you'll understand what's happening. In each of the topic areas above, either contributors have helped label content or Google's worked behind-the-scenes to get some of these going.

Keep in mind that for any top level label/topic/category, there are also sublabels/subtopics/subcategories. So for health, you'll see further narrowing options such as:

  • Symptoms
  • Treatment
  • Drug Uses
  • Support Groups

What about for the more individual user that wants to label? Good luck. Here are a bunch of instructions. They make my head spin. Maybe it will spin less when I read it many more times. But compared to tagging elsewhere, it's a nightmare.

Heck, it's a nightmare compared to trying to make a vertical search engine period versus a place like Rollyo. Want to make a health vertical search engine there? Give it up to 25 web sites and you're done.

Google told me it does want to make the process easier for anyone to take part, so hang in there, if the programming stuff makes you feel excluded as it does me.

The downside to Rollyo, compared to what Google's doing, is that you won't have all the refinement and subcategories. But I find it difficult to understand how well these will work, anyway. There's no controlled vocabulary for new people coming in. Moreover, it seems like some of these refinements could be done through clustering. My Yahoo My Web Tagging & Why (So Far) It Sucks article goes into some depth about these types of issues from when Yahoo rolled out My Web 2.0. They seem just as applicable to the new Google label system.

When Yahoo launched My Web 2.0, my gut felt like we were making a big step backwards, using humans to do stuff where technology actually could work. I wrote similar things when Google Base went up, pushing people into tagging content when it might not be necessary.

Don't get me wrong. I want humans involved in the search process. If anything, I've also written about how the growth of crawlers and automation has pushed human help sadly -- and perhaps harmfully to relevance -- to the side.

Certainly it's a big positive that Google's letting humans more in the door this way -- a huge jump for the service that has pretty much looked to technology to solve everything, as it acknowledges.

"We've never given our users this much control and access into our system," said Marissa Mayer, Google's vice president of search products & user experience. "We have an advantage with machines and how we crawl, but if we can turn our users into a network, that will yield better results."

Still, I think the idea of humans sharing and swapping what they like such as with Yahoo My Web or via the recent change with Google Reader might be the better way to go rather than manually tagging up millions of pages of content. But we'll see how it goes. As for the idea of subscribed links -- I've got a big thumbs-up for that part of Google Co-op.

Postscript: Google Blogoscoped has a nice guide to how Google Co-op also works for publishers, though most of the examples shown are more static than dynamic data draws.

About the author

Danny Sullivan was the founder and editor of Search Engine Watch from June 1997 until November 2006.

To contact current Search Engine Watch editorial staff, please click here.