What Is a Ranking Factor?

Google confirms very few ranking factors by name, with a date, through its own channels. Most of what circulates under the label "ranking factor" is third-party interpretation, not a statement from Google. These are the ones that are actually confirmed, with source and date:

• HTTPS: Google made HTTPS a ranking signal on August 6, 2014, according to its own Search Central Blog (then the Webmaster Central Blog). The post described it as a "lightweight signal," carrying less weight than content quality, and at the time affecting fewer than 1% of global queries.
• Mobile-friendliness and mobile-first indexing: Google announced on February 26, 2015, that mobile compatibility would start influencing rankings, effective April 21 of that year, the update the industry nicknamed "Mobilegeddon." Since 2019, Google also indexes the mobile version of every page by default instead of the desktop one, under its mobile-first indexing approach.
• Core Web Vitals: part of the "page experience" signals since 2021, alongside HTTPS and the absence of intrusive interstitials, according to Search Central's official documentation on Core Web Vitals.
• Helpful, people-first content: Google announced it on August 18, 2022, rolling out from the 25th of that month. Since 2024 it's no longer a standalone update but part of the core ranking systems, per the official guide "Creating helpful, reliable, people-first content."

Notice what these have in common: each one carries an announcement date, an identifiable official source, and, in HTTPS's case, even an impact figure Google stated itself. Outside this short list, Google has consistently avoided confirming individual factors with that level of detail. When an article claims Google "confirmed" a factor without citing a date or its own source, it's most likely repeating an industry assumption, not an actual statement.

A correlating signal is a trait that shows up more often among well-ranking pages in a given study. That doesn't mean Google uses it as a factor: it could stem from some other shared cause, or simply be statistical noise in the sample being studied. The distinction matters because a confirmed factor is a lever you can pull with a reasonable expectation of effect. A correlation is, at best, a lead worth investigating further.

The industry's most-cited example is the "200 ranking factors" list. Its origin is a remark by Matt Cutts, then head of Google's webspam team, who mentioned in 2009 that the algorithm ran on "over 200 variables." Google never published that list with names and weights. The 200-item, numbered versions circulating today are compiled by SEO tool vendors from their own observations, not from any Google source.

There are dated, quotable counterexamples too. In January 2014, Matt Cutts himself denied, in an official Google Webmasters video, that Facebook likes or Twitter followers were part of the ranking algorithm. In June 2019, Google's John Mueller replied on Twitter that "domain age helps nothing" when asked whether an older domain ranks better. Both signals keep showing up in third-party "factor" lists even though Google denied them explicitly, by name and date.

Load speed suffers a similar mix-up. Google confirmed Core Web Vitals as a signal, not some generic "seconds to load" metric on top of them. Treating page speed as an extra factor, separate from and additional to Core Web Vitals, blends a genuinely useful diagnostic metric with a ranking signal Google actually defined with precision.

A recent case shows why it's still tempting to treat any correlation as a confirmed factor. In May 2024, internal documentation from Google's "Content Warehouse" API leaked onto GitHub, exposing over 14,000 attributes spread across roughly 2,600 modules. Google confirmed the documents' authenticity on the 29th of that month, but warned against "making inaccurate assumptions about Search based on out-of-context, outdated, or incomplete information." The leak shows Google works with far more signals than it confirms publicly, and that seeing a field's name in an internal database says nothing about how, or whether, that field actually weighs into the final ranking.

Selling a correlation as if it were a confirmed factor promises more control than actually exists. If an agency or a tool claims "optimize these 200 factors and you'll climb the rankings," it's promising a cause-and-effect relationship that not even Google has verified for most of those items. The client buys a feeling of certainty, not a real guarantee.

That incentive isn't unique to classic SEO. Something similar happens with visibility inside generative AI answers. An Ahrefs study from May 26, 2025, covering 75,000 brands, found only a weak correlation of 0.218 between raw backlink counts and presence in Google AI Overviews, while unlinked brand mentions correlated far more strongly, at 0.664. A Semrush study from October 16, 2025 found a much higher correlation for Authority Score instead, 0.65, measured against a different sample with a different method. Both studies caution against reading their own numbers as causation, and they don't even agree on the order of magnitude. The same pattern repeats across both fields: the less real certainty sits behind a number, the more profitable it becomes to sell it as though the certainty were there.

Whoever benefits from blurring the line between factor and correlation is usually whoever is selling the "optimization" service, not whoever is buying it. A long, numbered list of factors looks more rigorous than a short, honest one, even when the short list is the only one backed by a verifiable, dated source.

None of this means correlation data is worthless. A weak or contradictory correlation is still a lead worth following, just not a promise. Telling a client that backlinks "might" relate to AI visibility, without guaranteeing a ranking effect, sets a more realistic expectation than selling the same correlation as a sure lever.

• Before calling something a "factor," look for the primary source: Search Central documentation, an official Google blog post, or a named, dated statement from someone at Google, not an article citing "studies" without linking to any.
• If the source is a correlation study, check the sample size, the statistical method, and whether the authors themselves flag that correlation doesn't imply causation, the way Ahrefs does in its own AI-visibility reports.
• Be wary of any figure quoted without a date or a link to the original source. In a field that shifts every few months, an undated number is often already stale by the time you read it.
• When two credible sources disagree, as happens between the Ahrefs and Semrush studies on backlinks and AI visibility, name both figures with study and date, instead of citing only the one that's more convenient.
• Check developers.google.com/search periodically for changes: Google stamps that documentation with a visible last-updated date at the bottom of each page.

• Presenting the "200 ranking factors" list as if Google had published it, when its origin is a loose 2009 figure Google never spelled out in detail.
• Recommending "optimize domain age" or racking up social media mentions as if they were ranking factors, despite Google explicitly denying both.
• Treating load speed as a factor separate from Core Web Vitals, instead of understanding that Core Web Vitals are the actual form in which Google measures and uses that speed.
• Citing a correlation study, say between backlinks and AI mentions, as proof of causation, without mentioning that the study's own authors warn against exactly that reading.
• Ignoring a source's date: repeating a 2009 figure or factor in 2026 without checking whether Google updated it, retired it, or folded it into a broader system.