Does boldface text really have more SEO value than plain text?
I am persuaded by the rigorous testing and skepticism of Michael Martinez of SEO-Theory that when he says emphasized/boldfaced text has some (although not huge) SEO benefit, he’s checked it. He often castigates people for dispensing SEO advice that they haven’t verified using controlled experiments (with all other variables isolated, so one knows for sure which factor is causing the result).
Since I’m guilty as charged, I conducted a few experiments. I discovered that (a) it’s really hard to eliminate all factors but the one you’re testing for, and (b) yes, it looks like boldface gives a SLIGHT seo boost.
Left: The image I used for the lens logo of each test lens, renamed with that test’s keyword.
Click “More” to read the details of my tests and results.
SEO Boldface Experiment 1:
On July 28, I made 4 dummy lenses on Squidoo with URLs: A2 A3 B3 B2 where A was one dummy keyword, B was another. They had no tags beyond their dummy names, and the Discovery Tool was OFF.
On each lens I included the word, or phrases with the word, in image filenames, alt-text, and module headers. Rather than repeating a particular phrase over and over, which is not natural, I used different phrases using the same keyword. On A3 and B2, the keywords were bolded; on A2 and B3, they weren’t. (I was making sure the number on the URL didn’t earn a slight SEO boost). I published them as close to simultaneously as I could, and then put links to them in one of my Squidbits posts in the order A2 A3 B2 B3
Preliminary Results– first link wins:
On June 29, less than 24 hours later, both the Squidbits post (interestingly, without a cached version) and the lenses were showing up in Google SERPs. It appears that they were ranked in the order of the links to them, with the first page I linked to listed first. For each Squidoo lens, the default Google search ONLY showed only one lens, and the second-ranked lens was hidden behind a “more” link:
“In order to show you the most relevant results, we have omitted some entries very similar to the 1 already displayed.
If you like, you can repeat the search with the omitted results included.”
Results repeated– first link on a page has more link juice:
In order to test this again, I went back and EDITED my previous post so that it now links to A3 A2 B3 B2 in that order. Sure enough, after a short time, A3 and B3 were listed in the main Google results, with A2 and B2 hidden behind a link. From this I conclude that the first link on a page gives more of an SEO boost than following links. This is in keeping with SEO advice I’ve read about putting the most important content higher on the page.
This doesn’t disprove that bold has any significance; it merely suggests that links have more weight (unsurprisingly).
SEO Boldface Experiment 2:
This time I was more cautious about what I linked to. I created two Squidoo lenses and four pages on my various different homepages and sites all talking about a person with a nonsense name.
On Squidoo, the paired articles were duplicate content with one bold version, one non-bolded version– and the URL this time was the nonsense word with “e” on the end of one and “a” on the other, so no numbers.
On my own website, I had two different “All about [keyword]” index pages which linked to ALL the other pages in the set. For the last doublet, I wrote a short story about the nonsense person, then mad libs it by substituting each noun for a noun, each verb for a verb, each adjective for an adjective, and so on until every word was different, but the sentence structure and number of words were exactly the same. So they register as “unique” content, but do not have different keyword densities, and more importantly, keywords appear the same distance apart.
The “mad libs” stories linked only to the Squidoo lenses, each one reversing the order of which one they linked to first. The Squidoo lenses linked only two each other. So the most links were pointing to the Squidoo lenses. Finally, I tweeted the two index lenses from two dummy Twitter accounts, again reversing the order of which I mentioned first.
It took over a month for any of these to show up in Google’s results. Finally, four of the six have appeared as follows:
Main Google results – Boldface wins:
- Squidoo lens with boldface
- Short story posted on my own site with boldface
Secondary Google results, hidden behind “show similar results” link
- Squidoo lens without boldface
- Both index lenses, and the one with boldface is listed above the one that isn’t
Not appearing in results:
- Short story posted on my own website without boldface
In all cases, the boldface lens has priority over the non-boldface lens.
In one case, Google apparently deemed that the authority of my own website was not sufficient to justify showing both results, and it only showed the boldfaced version of the pair. I wonder if it was smart enough to recognize the word-substitution trick, and treated them as duplicate content.
Conclusion: yes, Boldface did provide a slight SEO boost.
However, Experiment #1 suggests that links to a page carry more SEO weight than boldface, and that when a page links to several pages, it may give more “link juice” to the first link.
This makes testing REALLY HARD, because in order to let Google know that test pages exist, you have to announce them from something it does know, and you can’t avoid mentioning one first!
Caveat: The dummy lenses did have one link to them outside my control: while I gave them no tags shared by other lenses, and I turned off the Discovery Tool, I couldn’t stop their being indexed by their Topic (Experiment #1 was filed under SEO, Experiment #2 under Biography). I think they were buried so far down in their Topic listings that the only pages linking to them would’ve had zero value to Google, but I can’t be absolutely sure.
Try setting up your own experiments! Just remember, there are a ton of variables determining how Google crawls and rates, and it’s really hard to eliminate all factors but one in order to test that one!
Nice experiment, There is no other way to find new techniques in SEO rather than doing such a small researches. Thanks for the research and report.
Good example of how to do some basic testing, so you’re moving in the right direction, but part of the scientific process includes questioning your own conclusions.
Before you form that hypothesis, ask yourself how many ways a particular outcome of an experiment might be explained.
For example, your conclusion for experiment 1 says: “This doesn’t disprove that bold has any significance; it merely suggests that links have more weight (unsurprisingly).”
Agreed. It seems to suggest that links have more weight than bold-face — but I can think of several ways to challenge that conclusion as you don’t seem to have isolated repetition as a potential factor.
However, a little above your experiment’s conclusion you draw another inference: “From this I conclude that the first link on a page gives more of an SEO boost than following links.”
Is there any logical reason for this to be so? If you think there is, can you articulate it? If you cannot develop and articulate a logical explanation then, lacking a hypothesis, your conclusion is irrelevant (to the facts).
You should be able to look at the behavior and infer some sort of explanation (even though it may be wrong). That inferred explanation should then lead you to more testing.
I can easily show that first links on pages don’t have any SEO value — I have pages on several of my own sites where the anchor text for the first links is not being passed to other pages, and yet anchor text from links further down the page IS being passed.
If it were really as simple as “the first link counts the most”, then there should be no secondary links that pass more value than first links — OR we should have an explanation for why there are exceptions to the apparent rule.
In other words, in science one rarely comes up with a hypothesis that terminates a line of investigation. Search indexes are evolving mechanisms. Their behaviors are constantly changing.
One of the most unprovable hypotheses in science is the hypothesis that an evolving system exists in an unevolving state. That is, maybe at one time only the first link counted but now it doesn’t; or maybe at one time other links counted more but now the first link counts the most; or maybe something else is happening now but at some time in the past one of the two prior claims was true; or maybe all the claims have been true at one or more times in the past and will be true at one or more times in the future.
In fact, it can easily be shown that there is no statistical significance to differentiating PageRank-like value between links on a page but that there may other significant aspects to link order on a page (such as the order in which the links are passed into crawl queues).
Michael! Gracious, apologies for not seeing this sooner! I’m surprised and pleased to see you here, rather than in discussions of Arda where our paths crossed years ago.
Thank you so very much for the detailed reply. Much to ponder.
When guessing, “the first link listed may hold more weight,” I was thinking not so much logically, as by analogy, which betrays my classical studies training. First, I noted that search engines list links in SERPs from (they hope) most to least relevant. Second, when giving a list of links, people are savvy to the fact that the first link attracts the most clicks, so they may list the one they consider most important first. Third, as link lists get updated, newer links tend to get added to the bottom — not always, but it’s the easiest way to edit and add new links. For all those reasons, I could see that search engines might give a little more weight to the first link in a series.
But that was a flailing guess based on only one isolated case. You’re right; I’d need a lot more testing to prove it. And I haven’t really done enough to be sure about the boldface. Last week, I tried switching the boldface from one test page to the other in each of my doublets. Google’s seen the change but has not switched them in the SERPs. Drat.
The trouble is, this really takes a heck of a lot of time to do properly; nor am I particularly good at this kind of experimentation. I knew it would be difficult to isolate all factors except the one I was trying to test; it’s even harder than I thought! It’s so tempting just to assume that you and Matt Cutts and the other Experts know better than I. It is a lot easier for me to play textual critic, comparing and analyzing your posts and Matt Cutts’ sound bites et alia like a classicist trying to sort between the skewed accounts of the Catalinarian Conspiracy in Cicero and Sallust. (In fact, it seems to me that even real SEO experts do their share of literary criticism rather than scientific testing, trying to interpret what Matt Cutt means or reading other sorts of verbal tea leaves).
I will keep puttering away at these experiments, but I fear at this rate it will take a year before I’ve compiled enough meaningful data on even the simplest of factors (boldface). By that time, of course, every major search engine will have changed its algorithm.
I’m always about one step from deleting this blog as bogus information, but hopefully I’ve made it clear that I am an amateur trying to understand SEO, rather than a professional.