The Triviality of On-Page HTML Tag Optimization

I have long speculated that on-page optimization was trivial. It meshed with my understanding of how a suspicious Google engineer may treat the content of a page in relationship to its rankings. Why trust anything a webmaster says about his or her content (keywords stuffed into H1, meta, or bold tags)? Why trust anything on a page that a user won’t get to preview before visiting (anything outside the Title and Meta-Description, by-and-large)? However, despite my speculations, I lacked the data to truly start making conclusions about the usage of keywords in specific tags. Until now.

First, I am pleased to say that our micro-experimentation has found similar results to SEOMoz’s macro-experimentation. In particular, their finding that the H1 tag was no longer strongly correlated with better rankings helped reconfirm the results we were finding.

It is important to note that we use very different methodologies for performing research. I used the word “macro” in my description above because (and this is an assumption, but I believe a safe one) SEOMoz uses their ginormous LinkScape index to identify and measure relationships between ranking variables (tags, age, links, etc) and actual rankings. Virante uses a different methodology. We work on a much smaller scale, which gives us nearly complete control over potential confounding variables. If SEOMoz is taking pictures from a satellite, we are taking pictures through a microscope.

The Methodology

Our On-Page Tag Optimization experiment looks to unearth the relationships between HTML tags that are generally considered the most effective at influencing rankings. We looked at the Title tag, H1, H2, H3, Strong, Emphasize and, as a control, SPAN tags.

1. Generate a random, nonsensical keyword.
2. Generate 7 pages of nonsensical text with identical numbers of words (to exclude Keyword Density/Frequency as a variable).
3. Insert the keyword into either the Title, H1, H2, H3, Strong, Emphasize or Span tags on those pages, being careful to make sure that each keyword-tag is represented 1 time and only 1 time in each page.
4. Vary the order and location of the keyword within those tags (to exclude Keyword Location as a variable)
5. Save these pages using a different nonsensical word each.
6. Repeat several times over, to create a large body of test subjects.
7. Generate an XML Sitemap with random ordering of these pages (to exclude Order in Sitemap as a variable)
8. Use Google Sitemaps to get the pages indexed (to exclude PageRank as a variable)
9. Record rankings results.

The Results

Only a subset of the results are in, but I felt they merited discussion because they are quite telling.

Title Tag is King

Without any variation, the pages where the random keyword occurred in the title always ranked #1. This was the only HTML tag that consistently held its position and the only HTML tag that regularly and, in a statistically significant manner, positively impacted rankings.

The Rest of Them

Only the Strong Tag showed an improvement over the Plain Text (span tag) control. However, that improvement was not statistically significant. On a micro level, with PageRank out of the question, there is little to no meaningful difference in rankings brought by using the keyword in the h1, h2, h3, emphasis or strong tags.

Thoughts and Conclusions

There are a couple of things that we have to bear in mind. First, this study does not conclude that the use of keywords in these tags is without value. First, we did not test the combined value (such as using the keyword in the H1 and a Strong Tag, vs a page that only uses it in the plain text). Second, we did not test if there may be a synergistic relationship when the anchor-text pointing to a page matches keywords actually found on the page. There are many more reasons why on-page optimization is still valuable, but at best it can only be described as a minor part of the Search Engine Optimization process.