How Much Does Semantic HTML Structure Matter for SEO?

I’ve been posting on the page semantic HTML markup several times previously:

• Here are a few tools to quickly analyze any page semantic HTML structure;
• The piece listing the best practices of using the page headings;
• And some tips on how HTML markup helps your site accessibility.

While the benefits for accessibility and usability are quite clear, there has never been any real evidence that there really is any benefit for rankings. Does Google pay any attention to the HTML markup? Another great thread over at WebmasterWorld discusses how much SEO value can the page headings add.

Questions:

• We quite seldom see sites that use page headings correctly. Plenty of the sites on top of SERPs have completely screwed HTML markup – still Google seems to love them. So does Google really evaluates HTML semantic structure at all?
• If Google still does pay attention to HTML markup, how does it treat multiple H1 tags on one page? How can Google tell which one of them is more important?

Guesses:

• H1 must be used for some relevance signals – but it’s very much a secondary or “reinforcing” signal rather than a primary signal of relevance. There is no chance for Google to use it as a primary signal because in most cases HTML markup is misunderstood and misused: thus Google must be focusing on the relevance rather than the techncal knowledge of the webmaster.
• The effect of the page HTML markup can be seen from search results snippets: for example, H2 heading often appears in a snippet “even though there are other mentions of the query term earlier (and even more frequently) on the page.”
• Google may be evaluating the “HTML markup noise” when deciding if page headings should be taken into account: it may be paying more attention to the site HTML markup if it used consistently throughout the site and each page has unique HTML structure. Thus, if the site uses invalid HTML markup (e.g. multiple H1 tags, or one and the same headings on each page), the worst what must happen is that Google will just disregard the overall structure and give words used in heading no prominence. Again, Google is believed to put more emphasis on relevance than the HTML standards compliance.

My experiments?
I have been testing different theories but that’s too early to say anything. At my own blog I use several identical H1 headings on each and every page deliberately – and doing not bad for only the one that goes first in the source code. But again, this requires more testing and I will be sharing my findings in the future.