Last Updated on February 28, 2021 by Filip Poutintsev
Moz.com is not a company that cares about user experiences or shines with confidence. Every time you visit their website, you have to pass the longest possible CAPTCHA test. The need for such unpleasantness is unknown because I haven’t seen any other website to demand the same, and there’s not much to steal on their server compared to online wallets and online banks for example.
Along with other SEO tools, Moz has unique Spam Score metric, which is often popularized because other SEO companies do not offer anything similar. Moz claims to uses “machine learning and a massive set of training data”; however, most of the signs it sees as spam indicators are totally irrelevant.
In this article, I will go through all signals that Moz uses to calculate Spam Score and tell which of them are wrong. I will also discuss about other problem with the Spam Score metric and in the end propose better signals to determine a website’s Spam Score.
Table of Contents
The analysis of the Moz’s Spam Score
Below you can see 27 signals that Moz uses to calculate Spam Score.
1. Low number of pages found. Our crawlers discovered only a small number of pages on this domain. This is not an inherent problem, but many spam sites have small numbers of pages, hence the correlation.
Wrong.
Many good sites have a low number of pages, for example, corporate sites that do not sell any products online. On the other hand, true spam sites usually have tons of copied articles with hundreds or thousands of pages. This indicator has nothing to do with spam.
2. TLD correlated with spam domains. This domain’s top-level domain extension (e.g. .info, .cc, .pl, etc) is one that many spam domains use.
Wrong.
It’s as easy to buy .com and .net domain as it is to buy .info or .cc domain. Spammers use all easily available domains equally.
3. Domain name length. The length of the subdomain and root domain is similar to those used by spam sites.
Wrong.
Most high authority sites indeed have a short domain name, but this is because they have been around for a long time and were able to reserve good domain name back in the 90s. However, there’s no correlation between a long domain name and spam sites. There’s a correlation between new websites and long domain names, but it’s wrong to assume that most of the new websites are spammy.
4. Domain name contains numerals. Like many spam sites, this domain name contains numeric characters.
Possible.
The user can pick any domain name they like; it has nothing to do with the site being spammy. However, it’s true that spammy sites usually don’t care about what domain they pick, so they come up with ugly domain names often.
5. Google Font API Present. This domain does not use special fonts (e.g. Google Font API). Lacking this feature was common among spam sites we found.
Possible.
While this may be true, it could also be due to poor IT skills and have nothing to do with spam.
6. Google Tag Manager. Google Tag Manager is almost never present on spam sites.
Possible.
Same here; this is usually due to poor IT skills and not an indicator of spam.
7. Doubleclick Present. The Doubleclick ad tag is almost never present on spam sites.
Wrong.
Not many websites display DoubleClick ads or any ads for that matter.
8. Phone Number Present. Spam sites rarely have real phone numbers present on their pages.
Wrong.
Most people don’t want to display their phone number to protect themselves from prank calls. Moz.com, itself does not show phone number on their website.
9. Links to LinkedIn. Almost no spam sites have an associated LinkedIn page, hence lacking this feature is correlated with spam.
Wrong.
There’s absolutely no effort in creating a free and anonymous Linkedin page, and most spammers are actually active on social media, as they use it for spam as well.
10. Email Address Present. Email addresses are almost never present on spam sites.
Wrong.
Again, most sites don’t display their email to avoid email spam. Especially this is true for truly popular websites; even Moz.com does the same thing.
11. Defaults to HTTPS. Few spam sites invest in SSL certificates; HTTPS is often a good trust signal.
Possible.
Almost all good websites indeed have an SSL certificate. However, one can get it for $10 and without submitting any documents. Thus any scammer or criminal can have easily SSL certificate on this his. Not having an SSL certificate is more about being lazy than having anything to do with spam. However, it’s true that many spam website creators are lazy and don’t bother about small details.
12. Use of Meta Keywords. Pages that use the meta keywords tag are more likely to be spam than those that don’t.
Wrong.
Using meta keywords tag is a common SEO practise and has nothing to do with spam. While some SEO experts say that meta keywords are “dead” there’s also large amount of well established website that still trust in them.
13. Visit Rank. Websites with very few visits in clickstream panels were more often spam than those with high numbers of visits.
Wrong.
There are a lot of websites that receive a low amount of visits simple because they don’t promote themselves and don’t do any SEO. This is not an indicator of spam. Spam websites usually try to get a lot of traffic, as otherwise, they will not make any profit.
14. Rel Canonical. Utilizing a non-local rel=canonical tag is often associated with spam.
Possible.
Many high authority websites use the same tag often.
15. Length of Title Element. Pages with very long or very short titles are correlated with spam sites.
Wrong.
This correlates with bad SEO skills, not spam. Surely spammers may have bad SEO skills, but so do the others.
16. Length of Meta Description. Pages with very long or very short meta description tags are correlated with spam sites.
Wrong.
This also correlates with bad SEO skills and not spam.
17. Length of Meta Keywords. Pages with very long meta keywords tags are often found on spam sites.
Wrong.
Again, this also correlates with SEO skills, not spam.
18. Browser Icon. Spam sites rarely use a favicon; non-spam sites often do.
Possible.
It’s very easy to create a favicon, and many CMS platforms come with a favicon by default. However, it’s true that many spam website creators are lazy and don’t bother about small details.
19. Facebook Pixel. The Facebook tracking pixel is almost never present on spam sites.
Wrong.
Like with Linkedin, it’s very easy to add a Facebook pixel to the website, and the majority of spammers use social media actively for spam.
20. Number of External Outlinks. Spam sites are more likely to have abnormally high or low external outlinks.
Possible.
Many high authority sites link to no one, and at the same popular news publications have hundreds of thousands of outgoing links. More important than the number of links, is what type of websites they are linking to.
21. Number of Domains Linked-To. Spam sites are more likely to have abnormally high or low unique domains to which they link.
Possible.
Many high authority sites link to no one, and at the same popular news publications link to tens of thousands of different domains. More important than the number of links, is what type of domains they are linking to.
22. Ratio of External Links to Content. Spam sites are more likely to have abnormal ratios of links to content.
Possible.
It’s not clear what Moz means by “abnormal ratios”. Is it an extremely low or high amount of links? More important than the number of links, is from what type of websites they are coming from.
23. Vowels/Consonants in Domain Name. Spam sites often have many sequential vowels or consonants in their domain name.
Possible.
The user can pick any domain name they like; it has nothing to do with the site being spammy. However, spammy sites usually don’t care about what domain they pick, so they come up with ugly domain names often.
24. Hyphens in Domain Name. Spam sites are more likely to use multiple hyphens in their domain name.
Possible
The user can pick any domain name they like; it has nothing to do with the site being spammy. However, spammy sites usually don’t care about what domain they pick, so they come up with ugly domain names often.
25. URL Length. Spam pages often have abnormally short or long URL path lengths.
Wrong.
This also comes to bad SEO technique and is not about spam.
26. Presence of Poison Words. Spam sites often employ specific words that are associated with webspam topics like pharmaceuticals, adult content, gaming, and others.
Correct.
27. Uses High CPC Anchor Text. Spam sites often employ specific words in the anchor text of outlinks that are associated with webspam topics like pharmaceuticals, adult content, gaming, and others.
Correct.
The Verdict
As we see, most signals picked by Moz do not exactly characterized a spammy website. The reason for this is how Moz picked these signals in the first place.
“Spam Score represents the percentage of sites with similar features to the site you’re researching, which we’ve found to be penalized or banned by Google.”
Moz is not even trying to figure out if those signals have anything to do with spam? Correlation does not imply causation. Every SEO expert knows the majority of reasons why the website gets punished by Google. Instead of adding random signs that seem to correlate, why not do thorough research on the matter?
A real-life example of deductive fallacy used by Moz could be this: People with name “Jeremy” are more likely to be criminals than people with other names. So is it correct to discriminate against a random person who is called “Jeremy”, and because of this, and give him a “Criminal Score %”?
Instead of collecting signs that websites punished by Google have and then just adding them blindly to the list, Moz should have done some research to determine if these signs are actually connected with spam at all. Moz, knows that they messed up with their Spam Score, so there’s a disclaimer on their website:
“A high Spam Score for your site, or a site you’re looking at, doesn’t mean this site is necessarily spammy.”
Well, if the Spam Score cannot its job at identifying spammy sites correctly, what’s the point to have such a metric in the first place?
Other issues
Another massive problem with Spam Score is that it’s only updated every six months! Yes, just twice a year! Most SEO tools update their data every day, but here you have to wait half a year! This makes Spam Score metric not only useless but also completely outdated; 6 months on the Internet is eternity. Moz should be ashamed for charging money for such garbage.
How to calculate Spam Score correctly?
There are many other signs of a spammy website that Moz did not mention.
- Reciprocal links
- Broken links
- High amount of links from low-quality websites
- Plagiarized content
- Obvious web design flaws
- Not optimized for mobile
- Frequent page errors or downtime
- Domain or the IP is blacklisted
- Presence of malware on the website
- Suspicious source of traffic (high amount of direct traffic or referral traffic from suspicious websites)
- Domain authority compared to actual amount of organic traffic the website receives (spammy websites might have lot of backlink, but very little no organic traffic)
- Poorly written content (spelling mistakes, grammatical errors)
- Unnatural text to image ratio (only text with no images, or only images with very little text)
- Black hat SEO techniques
These are pretty obvious signs of a spammy website and actions for which Google can punish the website, thus it’s strange why Moz did not include them.
In the end, it’s actually a shame that such an important SEO metric is so poorly done that it loses its utility. Perhaps another SEO company will introduce a replacement for it and will provide more truthful data.
What do you think about Moz’s Spam Score?
Did I forget any other signs of a spammy website?
Please share your thoughts in the comments!
