Google's "good writing" content filter

Summary: The web pages actually at the top of Google have only one thing clearly in common: good writing. The usual SEO sacred cows and bugbears, such as PageRank, frames, and JavaScript, are less important, if they even matter at all.

I was recently struck by the fact that the top-ranking web pages on Google are consistently much better written than the vast majority of what one reads on the web. Yet traditional SEO wisdom has little to say about good writing. Does Google, the world's wealthiest media company, really rank web pages based primarily on arcane technical criteria such as keyword density, link text, or even PageRank?

Apparently not.

Most Common Website Content Success Factors

I took a close look at Google's top five pages for the five most searched-on keywords, as identified by WordTracker on June 27, 2005. Here's what I found.

The web pages that contained written content (a small but significant portion were image galleries) all shared the following features:

* Updating: frequent updating of content, at least once every few weeks, and more often, once a week or more.
* Spelling and grammar: few or no errors. No page had more than three misspelled words or four grammatical errors. Note: spelling and grammar errors were identified by using Microsoft Word's check feature, and then ruling out words marked as misspellings that are either proper names or new words that are simply not in the dictionary. Google almost certainly has better access to new words than the dictionary, with its database of billions of web pages. Supposed grammatical errors that did not in fact violate style rules were also ignored. Google would certainly be less conservative than a grammar checker in evaluating popular stylistic devices such as sentence fragments.
* Paragraphs: primarily brief (1-4 sentences). Few or no long blocks of text.
* Lists: both bulleted and numbered, form a large part of the text.
* Sentence length: mostly brief (10 words or fewer). Medium-length and long sentences are sprinkled throughout the text rather than clumped together.
* Contextual relevance: text contains numerous terms related to the keyword, as well as stem variations of the keyword. The page may contain the keyword itself few times or not at all.

SEO "Do's" and "Don'ts" that Don't Really Matter

A hard look at the results slaughters a number of SEO bugbears and sacred cows.

* PageRank. The median PageRank was 4. One page had a PageRank of 0. (Note that the low PageRank would seem to discount the idea that these pages owe their ranking completely to numerous incoming links.)
* Frames. The top two web pages listed for the most searched-on keyword employ frames.
* JavaScript-formatted internal links. Most of the websites use JavaScript for their internal page links.
* Keyword optimization. Except for two pages, keyword optimization was conspicuous by its absence. In more than half the web pages, the keyword did not appear more than three times, meaning a very low density. Many of the pages did not contain the keyword at all.
* Sub-headings. On most pages, sub-headings were either absent or in the form of images rather than text.
* Links: Most of the web pages contained ten or more links; many contain over 30, in defiance of the SEO bugbears about "link popularity bleeding." Moreover, nearly all the pages contained a significant number of non-relevant links. On many pages, non-relevant links outnumbered relevant ones.
* Text content: a significant number of pages contained little or no text. These pages were almost all image galleries (there was one Flash movie), with the images being photographs of the subject covered by the keyword.
Originality: a significant number of pages contained content copied from other websites. In all cases, the content was professionally written content apparently distributed on a free-reprint basis. Note: the reprint content did not consist of content feeds. However, no website consisted solely of free- reprint content. There was always at least a significant portion of original content, usually the majority of the page.

Recommendations

* Make sure a professional writer, or at least someone who can tell good writing from bad, is creating your site's content, particularly in the case of a search-engine optimization campaign. If you are an SEO, make sure you get a pro to do the content. A shocking number of SEOs write incredibly badly. I've even had clients whose websites got fewer conversions or page views after their SEOs got through with them, even when they got a sharp uptick in unique visitors. Most visitors simply hit the "back" button when confronted with the unpalatable text, so the increased traffic is just wasted bandwidth.
* If you write your own content, make sure that it passes through the hands of a skilled copyeditor or writer before going online.
* Update your content often. It's important both to add new pages and update existing pages. If you can't afford original content, use free-reprint content.
* Distribute your content to other websites on a free-reprint basis. This will help your website get links in exchange for the right to publish the content. It will also help spread your message and enhance your visibility. Fears of a "duplicate content penalty" for free-reprint content (as opposed to duplication of content within a single website) are unjustified.
* In short, make sure the bulk of your investment in your website is devoted to its content, rather than graphic design, old-school search-engine optimization, or linking campaigns.