"Cuil" New Search Engine: Is it Any Good?

The big news online today is all about Cuil, a brand-new search engine started by some of Google's finest. It uses primarily keyword and semantic technology to rank and return pages, rather than the Google core technique of looking at linkage - who links to your page and who you link to - and thus looks like one of the first true Web 2.0 search engines.

But is Cuil really cool? Because it may be able to scarf up a hefty share of Google's current search traffic - and because the fate of Yahoo is in the air these days - it just might be.

Then again, it could be a total dud.

Checking Out Cuil

The first thing you notice at Cuil.com is the black screen with nifty rounded widgets, instead of the stark white and pointy-cornered Google. It's pretty clear that Cuil is setting itself up to be the anti-Google.

The second thing you notice is that when you type anything into the Cuil search box, it pops up a list of suggestions, some good and some really odd. For web marketers and anyone interested in search term hits, this could be really important; more people will select one of the suggestions native to Cuil than will go ahead and type the rest of whatever they were typing in there. This will skew keywording somewhat, and is definitely something to take into account when doing your SEO.

The third thing I noticed - and I hope you don't - is that after I'd performed a couple of searches, Cuil returned a "high server load" error message, advising that I try the search again. Fagh. If they continue with that, how many people are going to stick with it? I suspect part of the problem is in the way results are returned. Instead of Google's relatively simple text return, Cuil gives you an image and quite a lot of text. Combine that with the fact that everyone's checking them out today, and you end up with a problem.

In search returns, Cuil uses the ten-items-per-page rule just like Google. But instead of using a straight-down-the-page technique, it shows them three across. (Why they don't do a top 12 so as to have nice neat columns, I don't know.) Watch the text! I'm not quite certain where they're drawing preview text from, but it looks like some people are getting preview text they may not have wanted, like top-line menus and text that ought to be hidden under images. I'm putting that down as something to keep an eye on when I go deeper into how Cuil works.Look around on the Cuil returns page. At the top are tabs with other search results for variations on your initial search -- an interesting idea, though most of what I was seeing on the first page showed up on the other tabs as well. On the top right instead of a listing of advertisers is a dynamic menu listing other related search topics; these might be good terms to include when looking for LSI keywords.

Another odd thing: there was a picture returned with one of my articles that - well, I just have no idea where it came from or why Cuil used it. Maybe this was an Associated Content database issue, but it makes me wonder - where does Cuil cull images from for documents that don't have a native image? Because I'm sure not every document has an image, and every return I looked at did have one. Hm.

How My Searches Ran

How to Read Palms: top 10 (or thereabouts) in Google, nowhere to be found in Cuil. This is disconcerting, as this is probably my best-LSI'd article. You'd think a search engine based around semantic indexing would put this one way up at the top - but no. I thought it was the long-tail problem (see next search), but Cuil bolds your search term when it pulls a page, and it had the whole thing bolded. Again, hm. It did look like returned result were heavy on the niche sites, so a generalized site like Associated Content might take a hit when being compared to, say, a specialty palm-reading site.

How to Move to Hawaii: top 10 in Google, had to use quotes to find it in Cuil. Got some weird results - initially, Cuil seemed to focus on "how to move" instead of the whole search term. Long-tail keywords may not work properly in Cuil. Also, I got a mysterious picture in the preview - instead of my pretty picture of the beach, it was some mustachioed black dude sitting at a conference table or computer. I have no clue where that came from.

I attempted to run other searches, but that's about when Cuil decided to be very un-kewl and the servers went high load.NOTE: it looks like high load is cuilspeak for "your search term is too long for me to parse right now, you dummy." I searched alternately long-tail and short-tail terms; the short ones came up every time, but if I had more than three significant words, I got the obnoxious error message.

Cuil: Recommend or Not?

I don't know. It seems to give a more concise set of returns than Google does these days, probably because it focuses on the most-likely-needed pages and not on the "broad and deep" thing Google is currently messing around with. However, I wasn't able to search enough on it to make a fair judgment. I'll probably used both for searching right now, then decide later.

What really scares me about Cuil is if it catches on, the whole LSI/SEO game is changed. All y'all people who put in hordes and hordes of hours on linkage - that won't make a spit of difference to Cuil. Those who moved to semantic, you'll be in better shape - assuming your LSI makes sense to Cuil. Here's hoping their systems improve immensely.

I predict, too, that Cuil will catch on. They have some big name players running it, and some high-profile deep-pocket backers funding it. It may not capture much market share initially, but I would be very surprised, gauging from my initial results, if it's not doing quite well in a couple of years. Look for someone on SearchEngineWatch to deconstruct it in the near future for more clues on how to write in a way that helps pick up increased traffic from Cuil.