Review: Out of the gate, Cuil is ambitious but flawed

After not much fanfare but plenty of anticipation, a new search engine called Cuil has launched. In spite of all the noise about the founders’ pedigree and the company’s index size, the only thing that makes a search engine tick is the relevance of its results across several dimensions. I spent an hour testing the service last night and found it has made an ambitious and interesting start out of the gate.

When evaluating a search engine, several interesting factors come into play. Does it do a good job with obscure terms? Will it satisfy the average user for popular terms such as “Las Vegas”? Is it susceptible to spammers? (No prizes for guessing: The answer is almost always a resounding yes.) Overall, I’ll try to focus this review on features that are interesting as well things that are blatantly broken in this early version.

(Disclosure: I work at Kosmix and manage the Kosmix RightHealth product. Kosmix is focused on a product that is very different from a general purpose web search engine, and isn’t intended as a direct Google — or Cuil — competitor. Also, these opinions are my own and do not represent my employer.)

First, I tried a popular query: “Las Vegas”. The point of this test is to make sure the search engine doesn’t make spectacular errors on terms for which plenty of web pages are available. So I pulled up Cuil’s Las Vegas page. Most of the results returned are reasonable, and if you use Google as the gold standard in terms of result relevancy, the first result, LasVegas.com, is on the money.



The one thing that Cuil misses sorely is any recognition of travel or local search intent. It fails to understand that a user searching for “Las Vegas” might be interested in driving or flying there. Google, on the other hand, will show you a handy map with a zip code box for driving directions.

Next, I tried a very obscure term: “Apocrine Hidrocystoma of the Eyelid”. The point of this test is to ensure that the search engine can serve results for the potentially infinite set of queries that can be thrown at it. After all, search terms similar to the one above are tried by users every day and are the source of Google’s prowess in relevance. Cuil’s results are quite reasonable.

Finally, we tried a term that is notoriously susceptible to spam: “Cialis”. Cialis is a treatment for erectile dysfunction and there are plenty of folks online trying to sell it to you and me on the cheap. Cuil fails to protect me from these guys. It has been speculated that Google prefers editorial content – such as that produced on Wikipedia – over commercial content, and it shows clear as day in this query. If you want to purchase Cialis using Google, you’ll probably have to rely on the ads over the search results.

Apart from the relevance issues, Cuil has an interesting take on user interface. Results are presented in a three-column grid format. The engine also tries to associate an image with the result. Sometimes, the image is the site’s logo — presumably, the presence of images is supposed to anchor the page and make it easy to read and scan. More interestingly, Cuil tries to present exploratory terms in the right column under a box titled “Explore by Category.” The idea is to allow users to make connections to the query term at hand in an easy-to-read format. While the feature is useful, I found the accordion menu treatment to be jarring and unnecessary.


There is another thing missing that’s a little more subtle. Google Universal Search, which launched last year, tries to blend search results from Image Search, Blog Search, Local Search, News Search as well as general Web search. For example, given the recent salmonella outbreak, a search for “salmonella” on Google turns up relevant news results. Similarly, a search for “pizza” will consistently return a map with several local pizza delivery services and their phone numbers. Cuil fails to do any of these things but can be forgiven for the simple fact that plain old web search is hard enough to pull off right. Blended/universal search can follow.

Overall, I recognize that any real comparison with Live Search or Google Universal Search is unfair. However, the point of this review is to highlight the fact that Cuil does a very reasonable job in this nascent stage. While it is nowhere near replacing any of the major search engines as a default option, I hope that it will become a worthy underdog over the next two years or so.

Next Story: More video game publishers embrace Comic-Con show
Previous Story: Is Apple thinking about ditching Intel chipsets?

Bookmark and Share

Tags:

Photo of Saumil Mehta

About the Author, Saumil Mehta

  • Lar
    I was trying to find a website to help me identify the song a melody comes from, so I tried "hum a few bars". Live Search and Google both pointed me to an O'Reilly blog that talked about musipedia.org. Cuil only returned 10 pages, most of them in Danish (or some other foreign language).
  • Machefsky
    I give it a failing grade all around. Try a simple search that is most relevant to Cuil (or whatever its name is).

    "size of the world wide web"

    The results are total nonsense. Hence, an "F".

    This thing needs lots of work.

    /Ira
  • @Lar: Yeah, foreign language filtering is another thing that search engines need to handle correctly. I didn't even want to get into it in the main post since spam filtering is more critical by any means :)
  • Clustered search results, but more readable than Kartoo. I like it.
  • Uhhh where did you see clustering? I saw no clustering whatsoever! Can you send link please? Thanks!
  • When I search on finance, say, the results come back in groups/clusters in the gray bar e.g. Personal Finance Software, Personal Finance Company. Link? There is a good tutorial on Latent Semantic Indexing at miislita.com , very detailed but still clear:
    http://www.miislita.com/information-retrieval-t...
  • Oh, I was looking for a Cuil link where I could see it, but now I understand what you mean. Yeah, top level categories, I've seen Clusty do it in the past and there are others who have tried. Its hard to get people to click on tab refinements though!
  • I'm not impressed. When I loaded the website early this morning, I had high hopes. But throughout the course of the day, my high expectations have soured into disappointment.

    First, the results are not relevant at all. It never retrieves what I am looking for, in the way that Google can.

    Two, the images are not necessarily associated with a website. I looked up my website 'Velvet Blues'. And instead of a logo or screen shot, I found an image with Spanish text on it, which came from God knows where! There's nothing about Cuba on my website.

    Three, you've ignored the downtime. Somehow, this search engine which was made to serve users, didn't adequately prepare for traffic.

    All in all, they made a poor showing, which I blame on a last minute scramble to reach their deadline... If you take a look at Digg or NYT, you'll find that over 90% of the comments are negative. In less than 24 hours of their launch, they've only been successful at one thing: alienating a good portion of their audience.
  • VelvetBlues,

    Relevance of results will be ongoing; you can't expect ANYONE to come out of the gate and match any of the majors, let alone Google.

    I do agree with you that the images that are pulled up are frequently irrelevant. Worse still, beyond "anchoring" the page I have NO idea what value they serve to the user.

    I do think though that people tend to be overly critical and evaluate search products along irrelevant angles (looking at the number of results per SERP across search engines, for example, which is a pretty lame way to approach the comparison).
  • I did a search for "financial content", the industry that I'm in. The first result was a description for TickerTech.com next to an image of our ad for FinancialContent.com which linked to TickerTech.com.

    So Cuil freely mixes text and images from different sources. I suppose more permutations would mean more results...
  • Jack
    I tried a typical techie query of ORA-07445 on both google and cuil. Cuil gives me zero results and google gives me 17,500.. Why bother using this engine even if its not universal ?
  • edhardy622
    British law student sues Abercrombie-Fitch for disability discrimination.
    http://www.abercrombieonsale.co.uk