Google Ranking Factors

Keywords and phrases that appear in the page URL, outside of the domain name, aid in establishing relevance of a piece of content for a particular search query. Diminishing returns are apparently achieved as URLs become lengthier or as keywords are used more than once.

The order in which keywords appear in a URL matters. It’s been theorized that keywords appearing earlier in a URL have more weight. At minimum, it’s been confirmed by Matt Cutts that “after about five” words, the weight of a keyword dwindles.

Title tags define the title of a document or page on your site, and often appear in both the SERP and as snippets for social sharing. Should be no longer than 60-70 characters, depending on the characters (Moz Tool). As with URL, keywords closer to the beginning are widely theorized to have more weight.

The percentage of times a word/phrase appears in text. In the early 2000s, practicing SEOs were sculpting content for target words to appear 5.5%-6% of the time. Google has since improved their content analysis methods to the point that those tactics are scarcely relevant anymore. And Keyword Density, although referenced in Google Patents, is almost certainly just a simplified concept behind TF-IDF, covered next.

Think of TF-IDF, or Term Frequency-Inverse Document Frequency, like Keyword Density with context. TF-IDF weighs the density of keywords on a page against what is “normal” rather than just seeking out a flat, raw percentage. This serves to ignore words like “the” in computation and establishes how many times a literate human should probably mention a phrase like “Google Ranking Factors” in a single document that covers such a topic.

Keywords in Heading tags have strong weight in determining the relevant subject matter of a page. An H1 tag carries the most weight, H2 has less, and so forth. This tag also improves accessibility for screen readers and clear, descriptive headings reduce bounce rates according to various studies.

Keywords in bold, italic, underline, or larger fonts have more weight in determining the relevant subject matter of a page, but less weight than words appearing in a heading. This is confirmed by Matt Cutts, SEOs, and a patent that states: “matches in text that is of larger font or bolded or italicized may be weighted more than matches in normal text.”

The closeness of words to one another implies association. To anyone that’s ever wielded the English language, this won’t come as a surprise. One paragraph about your SEO work in Chicago will thus do more to rank for “Chicago SEO” than two paragraphs, with one about SEO and one about Chicago.

The ALT attribute of an image is an used to describe that image to search engines and who are unable to display the image. This establishes relevance, especially for Image Search, while also improving accessibility.

Although Google may return search results that contain only part of a search phrase as it appears on your page (or in some cases, none at all), a patent states that a higher Information Retrieval (IR) score is given for an exact match. Specifically, stating that “a document matching all of the terms of the search query may receive a higher score than a document matching one of the terms.”

It’s established by a Google patent that, when content contains an exact match of a particular search phrase, that it has a more significant impact on ranking. In the process, Google indirectly confirms that you may still rank for certain searches when a page contains a text that partially matches a search query. This is further verified by just doing a lot of Googling.

There’s a natural trend in how we write English: earlier is usually more important. This applies to sentences, paragraphs, pages, HTML tags. Google seems to apply this everywhere as well, with content that appears earlier and more visibly being given more weight. This is, at very least, a function of the Page Layout algorithm, which gives a lot of preference to what appears above-the-fold on your site.

Keyword stemming is the practice of taking the root or ‘stem’ of a word and finding other words that share that stem (ie. ‘stem-ming’, ‘stem-med’, etc.). Avoiding this, such as for the sake of a keyword density score, results in poor readability and has a negative impact. This was introduced in 2003 with the Florida update.

The anchor text of a link tells the user where that link leads. It’s an important component of navigation within your website, and when not abused, helps to establish the relevance of a particular piece of content over vague alternatives such as “click here”.

The ideal method of separating keywords in a URL is to use a hyphen. Underscores can work, but are not as reliable, as they can be confused with programming variables. Mashing words together in a URL is likely to cause words to not be seen as separate keywords, thus preventing any Keyword in URL bonus. Aside from these scenarios, just using a hyphen will not make a site rank higher.

An SEO theory manifested itself in the early 2000s called the first third rule. It noted that our language – sentences, titles, paragraphs, even entire web pages, are generally used in order of importance. Although not confirmed by Google, Northcutt’s experience with word order experiments have more frequently indicated that this is a factor.

The full name of this one is technically “fresh content when query deserves freshness”. This term, Query Deserves Freshness (often shortened to QDF), refers to search queries that would benefit from more current content. This does not apply to every query, but it applies to quite a lot, especially those that are informational in nature. These SEO benefits are just one more reason that brand publishers tend to be very successful.

A Google patent states: “For some queries, older documents may be more favorable than newer ones.” It goes on to describe a scenario where a search result set may be re-ranked by the average age of documents in the retrieved results before being displayed.

Although it’s possible for outbound links to “leak PageRank”, web sites are not supposed to be dead ends. Google rewards authoritative outbound links to “good sites”. To quote the source: “parts of our system encourage links to good sites.”

Given that Google analyzes your inbound links for authority, relevance, and context, it seems reasonable that outbound links should be relevant as well as authoritative. This would likely relate to the Hilltop algorithm’s method for identifying topical experts. Google is also issuing manual actions for the reverse scenario.

This is a Bing ranking factor. Amit Singhal stated “these are the kinds of questions we ask” regarding spelling/grammar when defining “quality content”. Matt Cutts said no in 2011 as of “a long time ago”, but also that rankings correlate anyway. The first Panda update made this seem to matter a lot. Directly or indirectly, dozens of content-related factors are clearly affected by spelling/grammar.

We know that Google analyzes the reading level of content, since they created such a search filter for the results page (now removed). We also know that content mills, which Google is not fond of, are considered to be very basic, whereas academic writing was very advanced. What we don’t have, as of yet, is a concrete source or study that directly relates reading level to rankings.

Rich media, on top of drawing more traffic from in-line image and video search, has long been considered a component of “high quality, unique content”. Video appeared to be the deciding factor with Panda 2.5. Northcutt’s work has also shown a positive correlation. Currently though, there’s no official, public source signing off on this factor.

Google never rewarded meta keywords, but other search engines, scrapers, and crawlers have. Missing from the source article: the late-2000s explosion of reddit-like social bookmarking sites did leverage meta keywords. These sites would have skewed Danny’s results over time (and did skew similar experiments of ours).

A good meta description functions like a search ad. Considering how many Google Ads agencies exist almost entirely on A/B testing Google Ads ads, the marketing value here can’t be understated. Although keywords used in meta descriptions were once widely considered a direct ranking factor, Matt Cutts stated in 2009 that they’re not.

As time goes on, Google seems to do more to analyze ideas and logical entities in preference to words and phrases. It analyzes how we say things in preference to exact search queries that appear on a page. This process, in simple terms, is what’s making it possible to search for “how to cook meat”, and be returned results for steak recipes that never directly use the word “meat”.

As keyword density is now virtually a non-factor, a basic understanding of Phrase-Based Indexing tells us that if you write about content thoroughly and elaborately, you stand a far better chance of ranking compared to writing generic content that just happens to drop a lot of keywords. A clear component of one Google patent describes this as the “identification of related phrases and clusters of related phrases”.

Authorship was an experiment that Google ran from 2011 to 2014, which thrived upon bloggers using the rel=”author” tag to establish the reputation of particular authors. Google directly confirmed by the creation and demise of authorship. Eric Enge did a nice eulogy on the rise and fall of authorship on Search Engine Land.

SerpIQ once compared the length of content to top rankings and suggested that 2,000-2,500 words ranked best. Our analysis of Google Analytics across 20+ brands demonstrates that this doesn’t hold up practice. Google’s John Mueller also states that Google does not reward word count. Word count can correlate to novelty and links but is not a direct factor.

More than a decade of studies and correlation research suggests that titles that begin with a keyword usually (but not always) rank better than titles ending in a keyword. It’s easy to test and usually confirms: earlier keywords are better. But our chosen source for this suggests more. Thumback.com conducted a study where title word order changed traffic by 20%-30%. Their best-performing titles didn’t begin with a keyword, but were altered (as Google sometimes does) to do so in Google’s results.

A patent suggests that Google devalues a lot more than just identical duplicate content. Google has literally discussed methods for calling your content uninteresting. Once determining that a set of articles are related, such a process identifies which are more descriptive, unique, and/or weird (in a good way). It then rewards these “information nuggets” as part of a “novelty score”.

We know from countless sources (and even certain Search Console messages) that Google can separate user-generated content and analyzes it differently. One theory suggests that Google might look at quantities of comments on content to help rate content quality. There’s currently no documented evidence of this, and speculatively, it would also be one of the easiest factors to game.

There’s no definitive evidence (that we’re aware of) behind the HTML tag, <link rel=”hreflang”>, alone, helping you rank better. But, there is apparently a benefit for defining clearer signals behind different regional/language variations of a site. Often, multiple such signals are beneficial.

Keyword Stuffing penalties arise when abusing a once extremely effective tactic: sculpting Keyword Density to a high level. Our own experiments have shown that penalties can happen as early as 6% density, though TF-IDF (covered earlier) is likely at play and this is sensitive to topics, word types, and context.

This factor manifests itself from logic: if a higher Keyword Density or TF-IDF is positive, at some point, a total lack of frequency/density will decrease relevance. As Google has improved at understanding natural language, this may be better described as Subject Matter Dilution: writing content that wanders without any clear theme. The same basic concept is at play either way.

Aside from a page as a whole, Keyword Stuffing penalties appear to be possible within the title tag. An ideal title tag should definitely be less than 60-70 characters and hopefully still provide enough value to function as a good search ad in Google’s results. At absolute minimum, there is no benefit in using the same keyword five times in the same tag.

Aside from a page as a whole, Keyword Stuffing penalties appear to be possible within the title tag. An ideal title tag should definitely be less than 60-70 characters and hopefully still provide enough value to function as a good search ad in Google’s results. At absolute minimum, there is no benefit in using the same keyword five times in the same tag.

Heading Tags, such as H1, H2, H3, etc. can add additional weight to certain words. Those attempting to abuse this positive ranking factor will find that they can’t simply cram as many keywords as they can into these tags, even if the tags themselves grow to be no lengthier than usual. Keyword Stuffing penalties appear to be possible simply as a function of the total space within these tags.

As a general rule, if you want a concrete answer of whether or not an SEO penalty exists, try pushing a positive ranking factor well beyond what seems sane. One easily verified penalty involves placing your entire website in an H1 tag. Too lazy for that? Matt Cutts drops a less-than-subtle hint about too much text in an H1 in this source.

While there doesn’t seem to be any penalties associated with using a word in a URL multiple times, the value added from keyword repetition in a URL appears to be basically nothing. This can be verified very simply by placing a word in a URL five times instead of just once.

Matt Cutts notes that after about five words, the additional value behind words in a URL dwindles. It’s theorized and pretty replicable that this occurs in Google as well, although directly unconfirmed. Although they operate somewhat differently, Bing has also gone out of their way to confirm URL keyword stuffing is a penalty in their engine.

Given that ALT tag text is not generally directly visible on the page, ALT tag keyword stuffing has been widely abused. A few descriptive words are fine and actually ideal, but doing more than this can invite penalties.

Given that ALT tag text is not generally directly visible on the page, ALT tag keyword stuffing has been widely abused. A few descriptive words are fine and actually ideal, but doing more than this can invite penalties.

At minimum, really long internal anchor text will not bring along with it any additional value – a devaluation. In extreme circumstances, it appears possible to draw Keyword Stuffing webspam penalties from exceedingly lengthy anchor text.

It’s theorized that just having a site that’s all links and no substance is the mark of a low quality site. This fits the narrative content quality and not ranking pages that look like too much like search results pages, but is not currently supported by a study as proof.

Matt Cutts has suggested that any style of writing that just lists a lot of keywords could also fit the description keyword stuffing. Example: listing way too many things, words, wordings, ideas, notions, concepts, keywords, keyphrases, etc. is not a natural form of writing. Too much of this sort of thing will draw devaluations and possibly penalties.

Publishing content in a manner that is in violation of the Digital Millennium Copyright Act (DMCA) or similar codes outside of the U.S. can lead to a severe penalty. Google attempts to analyze unattributed sources and unlicensed content automatically, but users can also report infringement directly, resulting in manual action.

Doorway Pages, or Gateway Pages, are masses of pages whose only value is to soak in search traffic. They do not provide value to the user. For example, creating a product page for every city name in America, each with unique keywords. This method is called “spamdexing” (short for spamming Google’s index of pages).

Google has come a long way at analyzing image, but on the whole, it’s very unlikely that text that you present in rich media will be searchable in Google. There’s no direct devaluation or penalty when you put text in an image, it just prevents your site from having any chance to rank for these words.

Just like with images, the words that you use in video can’t be reliably accessed by Google. If you are publishing video, it’s to your benefit to always publish a text transcript such that the content of your video is completely searchable. This is true regardless of rich media format, including HTML5, Flash, SilverLight, and others.

Google has come a long way at analyzing images, videos, and other formats of media such as Flash, but on the whole, it’s very unlikely that text that you present in rich media will be searchable in Google. There’s no devaluation or penalty here,

Although it’s always been better to write more elaborate content that covers a topic thoroughly, the introduction of Nanveet Panda’s “Panda” algorithm established a situation where content with basically nothing of unique value would be severely punished in Google. An industry-wide recognized case study on Dani Horowitz’s “DaniWeb” forum profile pages serves as an excellent example of Panda’s most basic effects.

Duplicate content that appears on another site can bring about a significant devaluation even when it’s not in violation of copyright guidelines and properly cites a source. This falls in line with a running theme: content that is genuinely more unique and special against a backdrop of the web as a whole will perform better.

Similar to when content duplicated from another source, any snippet of content that is duplicated within a page or even the site as a whole will endure a decrease in value. This is an extremely common issue and can creep up from anything ranging from too many indexed tag pages to www vs. non-www versions of the sites to variables appended to URLs.

This was introduced as the “Bad Neighborhood” algorithm. To quote Matt Cutts: “Google trusts sites less when they link to spammy sites or bad neighborhoods”. Simple as that. Google has suggested using the rel=”nofollow” attribute if you must link to such a site. To quote Matt again: “Using nofollow disassociates you with that neighborhood.”

If a page contains the meta tag for “robots” that carriers a value “noindex”, Google will never place it in its index. If used on a page that you want to rank, it’s a bad thing. It can also be a good thing when removing pages that will never be good for Google users, and elevate the average experience on visitor arriving from Google.

This can appear two ways: if a page contains the “robots” meta tag with the value “nofollow”, it will imply that the rel=”nofollow” attribute is added to every link on the page. Or, it can be added to individual links. Either way, this is taken to mean “I don’t trust this”, “crawl no further”, and “do not give this PageRank”. Matt does not mince words here: just never “nofollow” your own site.

Google has vocally taken action against affiliate sites that provide ‘no additional value’ in the past. It’s in the guidelines. There’s much SEO paranoia that surrounds hiding affiliate links using a 301 redirect in a directory blocked by robots.txt, although, Google can view HTTP headers without navigating. A number of affiliate marketers have reported reasonably scientific case studies of penalties from too many affiliate links, therefore, we rate this as likely.

Generally speaking, Google wants users to land on content, not other pages that look like listings of potential content, like the Search Engine Results Page (SERP) that such a user just came from. If a page looks too much like a search results page, by functioning as just an assortment of more links, it’s likely to not rank as well. This may also apply to blog posts outranking tag/category pages.

Machine-generated content that’s based upon user search query will “absolutely be penalized” by Google and is considered a violation of the Google Webmaster Guidelines. There are a number of methods that could qualify which are detailed in the Guidelines. Once exception to this rule appears to be machine-generated meta tags.

If Google might have reason to confuse your site with a phishing scheme (such as one that aims to replicate another’s login page to steal information), prepare for a world of hurt. For the most part, Google simply uses a blanket description of “illegal activity” and “things that could hurt our users”, but in this interview, Matt specifically mentions their anti-phishing filter.

A Google patent exists surrounding stale content, which is identified in a variety of ways. One such method for defining stale content basically just surrounds being old. What is unclear is whether this factor harms rankings on all queries, or simply when a particular search query is associated with something Google refers to as Query Deserves Freshness (QDF), which means exactly what it sounds like.

While Google does index and return X-rated content, it’s not available when their Safe Search feature is turned on, which is Google’s default state. It’s therefore reasonable to consider that unmoderated user-generated content or one-time content that inadvertently crosses a certain line may be blocked by the Safe Search filter.

It’s theorized that, even though Google now tells us that they don’t use meta descriptions in web ranking, only for ads, it may still be possible to send webspam signals to Google if there’s an apparent attempt to abuse the tag.

Since 2009, Google has said that they don’t look at meta keywords at all. Despite this, the tag is still widely abused by people who don’t understand or believe that idea. It’s theorized that because of the latter fact, this tag may yet serve to send webspam signals to Google.

Google should single out problems appearing in the user-generated portions of your site and issue very targeted penalties in such a context. This is one of few circumstances where a warning may appear in Google Search Console. We’re told these penalties are usually limited to certain pages. We’ve found that WordPress trackback spam appearing in a hidden DIV is one way that this penalty can creep up undetected.

Obviously, if you write in a language that doesn’t belong to your target audience, almost no positive, on-page factors can work their charm. Matt Cutts admits that improperly isolated foreign language content can be a stumbling point both for search spiders and for users. To not interfere with positive ranking factors, Google needs to be able to interrelate content on the page as well as sections of a site.

Using Babelfish or Google Translate to rapidly “internationalize” a site is a surprisingly frequent practice for something that Matt Cutts explicitly states is a violation of their Webmaster Guidelines. For those fluent in Google-speak, that usually means “it’s not just a devaluation, it’s a penalty, and probably a pretty bad one”. In a Google Webmaster video, Matt categorizes machine translations as “auto-generated content”.