Important premise: I’m a white-hatter. Don’t misinterpret this article as spammy 😛
Recently I had the privilege to meet with Gary Illyes as part of a company survey program by Google. This was a rare and special opportunity, mainly because it was not a public meeting. I was able to provide some answers to him about how we use Google’s (and other people’s) tools and products, and in exchange, with Gary’s permission got some more specific questions about Google’s algorithm answered.
He gave me permission to speak freely of any of the content of the meeting — so with that permission, along with some additional thinking and experiments, here is the resulting article:
In White Hat phraseology: “Topical Authority Borrowing”
In Black Hat phraseology: “Stealing Rankings”
1: How Google’s algorithm “actually” works
To anyone who’s studied Google’s algorithm, it’ll come as no surprise that when a SERP is generated, it is the result of a scoring algorithm which ranks all possible results for that query based on the “over 200 factors” that Google looks at. A more specific detail (that could logically be assumed) is that Google actually gives a literal score, in the form of a number, to each result in the SERP.
As you’d likely expect, the way the numbers work is that the highest scoring sits at position #1, next is position #2, etc. Gary explained (due to my curious interjection) that there is a tiebreaker algorithm for when scores are equal.
As one may guess, these fit a reverse exponential curve: position #1 generally has a much larger difference in score in relation to position #2 than positions #41 and #42. That’s why when folks do SEO they notice it’s quite easy to go from page 5 to page 4, but much harder to go from page 2 to page 1; and even harder to go from the bottom of page 1 to the top.
Scoring would look something like this:
So: SERP results are scored into numbers.
2: The blessed “canonical” link
We’re all familiar with what canonical links do. They send a signal to search engines that two pieces of content are duplicate, and they assist them to identify which piece of content is the “original”.
But — another insight from Gary — canonicals are additive in terms of SERP scores. (*Assuming I understood him right, I think I did) — this is kind of confirmed by the wording in this article: “The rel=canonical link consolidates indexing properties from the duplicates, like their inbound links”.
I.E If I have two URLs ranking say position 8 and position 9 for keyword X, and I put in a canonical link from 9 to 8, then in theory the scores of both 8 and 9 would be added together (for keyword X’s SERP), hopefully bumping the URL that was in position 8 upwards, while sacrificing the presence of the URL that was in position 9 in that SERP, AND in ALL other SERPS that the URL in position 9 could have ranked for. It truly is a big sacrifice.
3: Topical authority
There’s no real need to go into this in depth, but Gary basically mentioned that the SEO community’s understanding of “domain authority” (a metric created by Moz and widely used in domain barter and analysis) is not matched by the Google algorithm “exactly”. For Google, the idea of “topical authority” is more accurate, or in other words, a website can be authoritative for a particular “topic”, but not generally. I.E There is no such thing as “generic” or “raw” domain authority.
Now, the implications of these three are almost limitless, and, perhaps needless to say, a bit scary. For the sake of the mental exercise, let’s walk through some theoretical scenarios, and then some practical experiments that I’ve actually done which “prove” that, at least in some cases, this stuff can work and could be dangerous.
Results: My canonical experiments
Now most of the above (except topical authority) I already knew or strongly suspected before my meeting with Gary. But it’s always good to have a person who actually “knows” come around — they have much more legitimacy and credibility.
So actually I’d been running canonical experiments for some time already before bumping into Gary — here’s are some examples.
My company’s website is http://www.ewebmarketing.com.au/ — it has a blog.
I occasionally contribute content to our blog, but because I usually write the content after hours, and it’s usually the result of personal musings, if I ever donate content to the company blog I’ll duplicate the article on my personal blog (this one) and put in a canonical link from the company blog to my personal one.
What I’ve noticed is what I’ll call “Topical Authority Borrowing” — you see while my personal blog is not very authoritative in the SEO topical niche (really, it’s a small fry) — our company’s blog is far more authoritative — AND because I’ve put an article there, and a canonical link going back to MY blog, my blog has much more ability to rank for relevant keywords related to those articles. For instance, look at the SERPS:
In Australia, in incognito, my personal blog ranks very highly for all those queries. Now — two things, firstly those queries are close enough matches to the names of the articles that we could reason I would win anyway — but you’ll notice for many of them there are OTHER articles with, if not the exact same name, very similar names also featured in those SERPS. And secondly we might reason “Yeah but the content is good, it’s long form, etc. so Google might like it and rank it well anyway” — this is true, and I can’t prove conclusively that this is not the case, however my gut feel from many experiments (and knowing how my blog ranks for other things) is that:
- At least for those blog articles that are duplicated, the version on my blog ranks better than it would, were it not duplicated with a canonical link from my company. Note that often these articles are ranking over potentially more authoritiative SEO blogs, like those of businesses.
- This is an extension, and purely speculative, but it is possible that my blog generally has gleaned some “topical authority” in SEO from this number of canonical links from http://www.ewebmarketing.com.au/ — this would have even more implications.
Implications : The scary
There are many implications of this, let me name just a few. Keeping in mind, we are in the realm of conjecture (AKA “educated speculation”) with some of these being more plausible than others:
Hacking and stealing rankings (for example, eCommerce)
The first, most obvious implication of this is in pure black hat, malicious hacking. From a gut feel I’d give this a 7-8/10 chance of succeeding.
Imagine you and I are both eCommerce store owners and compete directly with each other.
I find a hacker somewhere in Ukraine and commission him to hack your store. But I’m not after defacement or downtime, oh no, something much more sinister and menacing — I want your Google organic rankings.
So he successfully hacks your store, and I give him a canonical map. He could carefully place canonical links in, at a highly granular and targeted level, into certain URLs on your site that rank well and point them to the corresponding URLs on my site that sell the same products or contain the same content. Assuming a few things (like Google comes back and crawls them, the content is sufficiently similar for the canonical to be accepted, etc.) — viola, not only have I potentially stolen YOUR rankings, but I have ADDED them to my own, boosting my site potentially higher in SERPS than either of our sites could achieve independently. Note Google takes canonical links as “a hint that we honor strongly”
You know the scary part of this?
- People buy and sell businesses (and eCommerce stores) all the time. It’s perfectly plausible, from Google’s point of view, that I bought your business and put the links in legitimately.
- The sheer undetectability of this is also quite scary. Anyone without an SEO eye would not spot this. Even folks with an SEO eye would have to read the page source of the canonical source pages to find the links. (I.E They would have to be actively looking for THIS exact thing, or using a crawling tool that reports this and happened to bump into it) Want to make it even better concealed? — be very picky and choosy about which canonicals you actually put in place in order to obfuscate your efforts — so it becomes even harder to detect. Just pick a handful of useful keywords (I.E URLs that are ranking favourably for said keywords) but not ALL URLs — and you have yourself a dangerous tool to steal rankings, almost undetectably.
This is wild speculation, and may just be very unlikely, but imagine that you build and highly optimise several web pages on different domains for a certain keyword — each different from each other, but still similar enough to effectively target that keyword. Then — you pick one to rank, and put in canonicals from all of the others to that one. IF scores are additive in SERPS for canonicals, then the “ranking power” of each of those URLS might be added together to form a higher score than any would individually:
Hence the idea of “ultra on-page” — or optimising several pages in a non-duplicated fashion. This I’d give more of a 3/10 chance of succeeding — I’d have to try it to see the results. The challenge with experiments of this nature is that the canonical URL itself might, in and of itself, be good enough to rank first anyway. It’s hard to run these experiments without actual knowledge of SERP scores.
Visualise a website into a kind of triangle — the homepage is at the top, the category pages in the middle, and the end-pages at the bottom in higher volume. Now transform this triangle into three-dimensional dish, like a satellite dish. The surface area at the bottom are the lowest tier pages, the point of focus is the homepage and category pages.
When you do a search for something, say “shoes”, all possible results are in there somewhere, deep within the SERPS. If you have a website that sells shoes, likely the homepage has the word “shoes” as its main target, then categories are likely things like “men’s shoes”, “women’s shoes”, etc. And lower down you have specific shoes (“blue Converse size 12 men’s shoe”) — when you do a search for something like “shoes”, ALL of those results of your entire website are “in there somewhere”, but it’s likely the homepage is ranking higher with higher score.
Now — implement the satellite dish — put canonical links in ALL the lowest tier products pointing to either the category pages or just the homepage, and all scores from all of those lower pages will be added to the score of the homepage (or category pages), like a SEO satellite dish, or “Canonical Dish”.
I’ve never tried this before. It is actually weird to the point of almost being silly. I’d expect not a massive boost even if Google did accept the canonicals. But … thousands of products with minimal score… added to a homepage or category page… It’d be interesting to see the results. In any case, you’d be sacrificing the ranking of all lower tier products — people would not be able to search for them directly anymore.
But you’d probably do better to stop playing silly SEO tricks and build a better website with your time 😛
Warning: Don’t NOT try this at home
We don’t have time to try every SEO idea in a sterilised experiment. But if anyone does try any of these — I’d seriously warn them not to do this for anything valuable — especially not your clients.
To borrow the slogan from Google: “Don’t be evil”. Try this in a safe environment if you want to — don’t be surprised if you get nuked by Google later on — but don’t be silly and put other people’s businesses at risk with this stuff.