Cuil Launches – Good for Discovery and Surprises, Bad for Search

Yesterday evening a new search engine was launched by a group of search experts (including two ex-Googlers) - Cuil is the name and the blogosphere is abuzz about weather it will kill Google. Honestly, I can never understand why we constantly expect every single new product to kill something that is strong in the particular niche already. Normally I hate that type of post titles - they sound to me like BMW is a Mercedes killer. And that makes me doubt the products even more usually - when startups fail to come up with something unique, they often choose a path of "We will kill Google".

But people still seem to be determined that a startup that has ex-Googlers on board and $33 million in funding has great chances of killing the most mainstream search engine of all. So why do some people hope that Cuil has good chances in the search field?

First of all, they launch with a very substantial index of their own - and 120 billion IS substantial, even if we compare it to Google's recently reported trillion. We may be optimistic about the index size but Techcrunch has already proved that Google beats Cuil in the number of search results, at least.

Another factor is the huge search-related experience of Cuil founders. But I strongly doubt the fact that if you know how Google operates from inside, you can actually create something better of your own.

Another difference is the way Cuil handles pages indexing and search queries - their process is better scalable and cheaper than what Google has. Unfortunately, these processes are not visible to users - and for scalability to start to matter, the first step should be reaching a significant market share.

What does make a difference is Cuil's approach to indexing web pages and categorizing them - using the relations between key words on this or that page, Cuil categorizes them to make search process easier and more comprehensive (or at least this is what they claim). Thus, if a user gets the results for a search query that are far from comprehensive, he is offered to explore the results further based on the proposed categories. But my experience with Cuil described below shows that keyword-rich pages can be far from relevant as well.

The only thing that any user will be sure to notice immediately is the totally new approach to interface - unlike in any traditional search engines where results are displayed as a list of links in line, Cuil offers a two-column or three-column interface (triggered by a user) that looks a little cluttered to me. I am not sure if it is supposed to appeal to users better but I am sure that this levels the weight of the pages in the first 10 of the output - to a certain extent at least. But what I find immediately appealing is that Cuil makes the results look better to users by adding images to the results whenever it is possible - obviously people notice visual content better so this could be a huge plus, especially for the websites where Cuil can find images for.

There is no doubt that some of the most advanced users will also see Cuil's approach to users' privacy as important and advantageous - Cuil promises never to store users' IPs so no search query can be related to any particular computer in the world. Given the latest concerns of privacy and how Google handles it, it may be a significant advantage for Cuil, especially when it comes to early adopters. The problem is I don't think that a real mainstream audience actually knows that they should care about their privacy and that Google has problems with keeping private things private.

But all of the above is mainly related to how the company described itself and its new search engine. Now that Cuil is live I wanted to see it in action myself. So to see what Cuil has to offer, I have performed a search for "web 2.0".

You know, it is hard not to fall in love with a search engine that puts Profy on the first page of search results for this rather popular search term - along with links to a number of events (Web 2.0 Summit and Web 2.0 Expo in Berlin - by the way, no idea why Berlin instead of a bigger one in San Francisco or the main site for all the Web 2.0 Expo events), the Wikipedia entry for Web 2.0 and some other blogs and directories focused on web 2.0.

But you can't fail to notice some faults immediately - for example, while Cuil returns 6,200,000 results for "web 2.0", Google is definitely more abundant with its 449,000,000 - that's 72 times more on Google. Of course, that does not mean that Google's results are more relevant and to prove that Cuil offers a number of categories to browse for more results, including News Aggregators, Semantic Web, and Buzzword (sure, we have needed someone to offer us a broad collection of buzzwords).

Out of curiosity I have clicked one of the proposed related categories - that one happened to be Google Reader. Unfortunately, it failed to produce any results at all:

So don't tell me they have more pages indexed than Google - I will not believe it until I see the same 23 million results for "web 2.0 google reader". But what surprises me here is that they suggest a category to the user that contains zero results in it. Would not it be wiser to only offer things that people can find something in?

Nevertheless, I continued clicking some of the other categories and they did offer some results for the categories - but, to tell you the truth, some of the results were truly unexpected. For example, the search on "web 2.0 semantic web" returned a CV on the personal site of Carlo Torniani that was very rich in the key words but was hardly relevant.

So my general impression is that the results sorted out by keyword richness are actually not that reliable - and if I launch a blog with the words "semantic web" repeated 10 times in every post I will probably do very good on Cuil. But the problem is that not everyone will agree that such results are actually relevant - and I think we will have to stick to tried and true Google's page rank that will at least guarantee that some people find the content on the resulting page useful to them.

And this is why I think that Cuil may have a great potential - but not as a search engine. Sure, it is different than Google, it has a visual appeal of some kind and it does determine keyword-rich pages very well. But different does not mean better - these factors do not seem to provide actually relevant results. So if you want to find useful information that many other people rely on, you'll be better off sticking to Google. But if you are looking for something really unexpected that you would never have stumbled upon otherwise - welcome to Cuil and browse some of the categories, you may be astonished about what type of content they can offer. But don't expect relevancy here - expect surprises.