The first results of the experiment are in, and they’d seem to indicate that Google places no importance at all on heading elements relative to any other text in a page.
In order, the current results (with duplicate hits included) are:
- embedded
h2 - full content of a normal paragraph
- embedded
h1 - normal text embedded in a paragraph
- split between
h1andh2
Note, however, that the missed h1 test–probably the most relevant of any of them–is not yet included. I don’t hold high hopes that it will fare any better, though.
I’ll leave this to age for a few more days and see if things change at all, then try a new iteration. Suggestions are welcome, as always.
This makes me wonder how Google decides which page comes out on top…
At first I thought it’d weight the links at the beginning of the document more, but that doesn’t seem true… Weird.
I suspect you might have abstracted it out too far. Google’s not simplistic about results – if it was, it would be far too easy to game. A phrase that appears in an <h1> but nowhere else? Smells a bit spammy. A phrase that appears in the <title>, an <h1> and embedded in text? That’s probably what the page is about.
It’s a very interesting question, but not very easy to answer: you probably need to compare something more like a simple but real semantic page to random crap markup and search engine spam markup.
Very good points, Phil. I’ll try some additional combinations the next time around, and some real(istic) text instead of the lipsum nonsense. Different hierarchy depths might be interesting too.
There are two things I did notice about the ordering, Dark. First, doing a phrase search drops the last result, which I suppose is to be expected. Second, ordering the words differently in the search only makes a difference when the middle term is placed first or last–in other words, when the search words are in the order they appear on the page. The only result that moves is the final one though.
I think we should do some more testing before we truly dispair. There are still very real reasons to use the heading tag, noteably the use of headings within screen reading software.
Absolutely, Matthew. The goal was just to test Google’s handling of the header tags, not to say they’re useless.
By the way, I’d have tried other search engines too, but they all seem to have migrated to “register to add your URL” or “pay us to add your URL”, and frankly I’m not willing to go through the bother/cost of either.