Google confirms that its ranking systems use query relevance, link analysis including PageRank, signals tied to helpfulness and reliability, freshness when a query calls for it, Core Web Vitals, aggregated user interaction data, and context about the searcher. That is the honest answer to "what are the confirmed Google ranking factors." What Google does not publish is a complete formula or a set of universal weights. So the bigger question isn't just which factors are real, it's how confident you can be about each one, because "confirmed" and "important" are two different claims.
This page is a reference, not a deep dive, and it's built to be a proven ranking factors list you can actually trust rather than one more copy of the same unsourced checklist. Each row below is graded by the strength of its evidence: confirmed, correlated, speculative, or myth, which is the clearest way to show real vs speculative SEO signals side by side instead of flattening them into one undifferentiated list. Where a topic has enough nuance to deserve its own piece, you'll find a link to it. Where it doesn't, you'll find a short, honest answer instead of padding.
What counts as a confirmed Google ranking factor
Before the list, it helps to agree on what these four labels actually mean, because most ranking factor content blurs them together.
A confirmed factor is one Google explicitly documents, or one a primary legal record directly shows in use. Confirmed does not mean you know its weight, its exact mechanism, or whether it applies the same way to every query.
A correlated factor comes from an observational study that found an association between some measurable property and rankings. Correlated does not mean Google consumes that exact metric, that the property causes rankings, or that changing it will reproduce the result.
A speculative claim is one the evidence assembled here doesn't establish. Speculative is not the same as disproven. It just means nobody has shown it yet.
A myth is a claim that contradicts explicit evidence, or one that treats a broad framework as a single switch you can flip.
A few more distinctions matter throughout this piece. A signal is information a ranking system uses, like a link or a term in a document. A system is something that interprets or combines signals, like PageRank or RankBrain, and these system names are not individual checkboxes on a list. A quality framework, like E-E-A-T, describes desirable traits rather than a single scored input. An eligibility condition decides whether something can even be considered, which is different from how strongly it ranks. And a third-party proxy, like a Domain Rating score, is an outside estimate, not a Google number, even when it correlates with one.
Which ranking mechanisms has Google actually confirmed
This table is the core of the page. Each row ties a claim to the documentation or record that supports it, and the boundary column exists so you don't carry the claim further than the evidence does.
| Mechanism | Tier | What the evidence shows | What it does not mean |
|---|---|---|---|
| Relevance to the query's meaning and intent | Confirmed | Google's How Search Works page and its ranking systems guide describe systems (BERT, neural matching, RankBrain) built to interpret a query and match it to useful content. | "Search intent" is a useful way to think about this, not a disclosed numerical score. |
| Query terms in headings and body content | Confirmed | Google says matching keywords are a basic relevance signal and names headings and body text specifically. | This does not confirm an ideal keyword density or a bonus for using a heading tag in a particular way. |
| Link analysis and PageRank | Confirmed | Google's ranking systems guide says PageRank remains part of its core ranking systems, though it has evolved over time. | This doesn't mean every backlink helps, or that links are always the single most important factor. |
| Helpfulness, reliability, and quality signals | Confirmed | Google's helpful content documentation says its systems prioritize helpful content and use a mix of factors tied to experience, expertise, authoritativeness, and trustworthiness. | There's no single quality score and no disclosed weight for each trait. |
| Original content and original reporting | Confirmed | The ranking systems guide describes systems meant to surface original content and original reporting prominently. | Originality isn't just a plagiarism check or a rewording of someone else's information. |
| Freshness when a query calls for it | Confirmed | Google documents query-deserves-freshness systems and contrasts breaking-news searches with stable, reference-style ones. | Not every page benefits from a newer date or more frequent updates. |
| Core Web Vitals | Confirmed | Google's page experience documentation explicitly says Core Web Vitals are used by its ranking systems. | A perfect score doesn't guarantee a top ranking, and it won't rescue irrelevant content. |
| Aggregated user interaction data | Confirmed | How Search Works describes aggregated, anonymized interaction data feeding relevance signals, and a primary legal exhibit from a February 2025 call with a Google engineer discusses clicks and a system called Navboost. | There's no universal click-through-rate formula and no confirmed link between this and a Google Analytics metric. |
| Searcher context (location, language, history) | Confirmed | Google names location, language, device settings, and past search activity as inputs that can shift results. | These are conditional, per-search inputs, not properties of your page. |
| Words in the domain name | Confirmed, limited | Domain words are named as one of many relevance signals, but Google's exact-match-domain system is designed to stop domains from getting too much credit just for matching a query. | Buying a keyword-stuffed domain is not a reliable shortcut. |
| Page-level and site-wide assessment | Confirmed architecture | Ranking works at the page level, but site-wide signals and classifiers also play a role. | Neither "only the page matters" nor "a strong site carries every page" is accurate. |
A few supporting systems, like passage ranking, deduplication, and site diversity, are also confirmed, but they're not extra checklist items you optimize for directly. Passage ranking helps Google understand sections within a longer page. It isn't proof that a specific paragraph length earns a bonus. Deduplication changes which near-identical result shows up, not a penalty you can trigger. These exist to explain why the whole ranking system resists being flattened into a single list, not to lengthen one.
Two older confirmed claims deserve a date warning. HTTPS was confirmed as a ranking signal in 2014, and at the time Google described it as very lightweight, affecting less than 1% of queries globally. That's a real, confirmed historical claim, not a current measurement, and current page-experience documentation says that beyond Core Web Vitals, most other experience factors don't directly help a page rank. Secure delivery is still a sensible baseline. It's just not a lever with a known, current ranking return. Mobile usability works the same way: Google confirms that usability matters broadly, but a specific, independently weighted "mobile-friendly bonus" with a known current size isn't established by the documentation.
What do ranking correlation studies actually prove
Correlation studies get cited constantly in SEO content, and they're genuinely useful, as long as you keep straight what kind of evidence they are.
Ahrefs ran one of the most-cited link studies, pulling the 1,000,000 highest-search-volume US keywords and calculating how ranking position correlated with several link metrics. The reported coefficients were weak across the board: Domain Rating came in at 0.131, backlink count at 0.248, followed backlinks at 0.242, and referring domains at 0.255. Ahrefs itself describes these as weak or very weak and is upfront that correlation doesn't prove causation.
Here's the useful way to read that. Google's own documentation confirms that link analysis, including PageRank, is part of how it ranks pages. That's the confirmed mechanism. The Ahrefs study adds a correlated data point on top of it: pages that rank well also tend to have somewhat more links and higher third-party authority scores, in this particular sample. It doesn't tell you that getting more links will cause your rankings to improve, because the sample skews toward high-volume US keywords, because ranking well can itself attract links (so the direction of cause and effect isn't settled), and because things like brand recognition and content quality tend to travel along with link counts rather than being separated out.
A second Ahrefs study looked at whether scores from popular content-optimization tools correlate with rankings. The researcher used 20 keywords, ran each through five different tools, and found weak correlations throughout. Worth flagging here: Ahrefs sells one of the tools it evaluated, which doesn't make the study wrong, but it does mean you shouldn't read it as a neutral third party validating the category. The honest takeaway from both studies is the same: correlation is informative context, not proof of a mechanism, and definitely not a dial you can turn with a predictable result.
Which popular SEO signals remain speculative or are myths
This is where most of the internet's "ranking factors" lists go wrong, by treating every item on a proven ranking factors list the same way they treat the confirmed ones above. Keeping real vs speculative SEO signals separate matters more here than anywhere else on this page.
Google has a preferred word count. Myth. Google's own helpful content guidance explicitly says it has no preferred word count. A page can be long because the topic needs it, not because length itself is rewarded.
E-E-A-T is a specific ranking factor, or a public score. Myth. Google's documentation is explicit that E-E-A-T itself is not a specific ranking factor; it's a mix of factors used to identify content that demonstrates those qualities, with trust weighted as the most important piece.
Search quality raters directly set rankings. Myth. Raters evaluate how well Google's systems are performing. Their ratings inform how systems are built and tuned, but they don't hand-rank individual pages.
Meta keywords and meta descriptions affect rankings. Myth, and an old one. Google explicitly denied using the keywords meta tag back in 2009, and the meta description tag was confirmed at the same time to supply snippet text only, not a ranking input.
Buying Google Ads improves organic rankings. Myth. Google's How Search Works page states plainly that advertisers get no special treatment in organic results.
Correct structured data guarantees a rich result or a ranking boost. Myth as a guarantee. Structured data makes a page eligible for certain search features; eligibility is not the same as display, and neither is the same as a general ranking increase.
A high third-party authority score is Google's internal score. Myth as an equivalence claim. Domain Rating and similar metrics are outside estimates built by other companies. They can correlate with rankings without being anything Google actually uses internally.
Raising your raw click-through rate always raises rankings. Speculative. Interaction data is confirmed as part of the system, but a universal formula where a higher dashboard CTR number produces a predictable ranking gain isn't established by the evidence.
There's a universal ideal dwell time. Speculative. The primary legal record discusses historical time-before-return information, but it doesn't disclose a target duration marketers can aim for.
Posting more frequently raises rankings on its own. Speculative. Freshness systems are confirmed for queries that call for them, but a general bonus for publishing cadence isn't.
Domain age alone provides a bonus. Speculative. Older domains often have more accumulated links and reputation, which can confound this claim; no primary evidence here shows age itself as an independent factor.
Keyword density has a universal ideal percentage. Speculative. Keyword relevance is real; a specific percentage target turns a relevance signal into a repetition quota that isn't supported.
AI-generated content is automatically penalized, or automatically advantaged. Myth in both directions. Google's guidance says appropriate use of automation is fine and gives no special gains either way. What gets penalized is content, AI-assisted or not, produced mainly to manipulate rankings without helping anyone.
Why dated evidence and current documentation can disagree
A lot of ranking-factor confusion comes from mixing a true historical statement with a false current one, or the reverse.
Helpful content is a good example. It launched as its own system in 2022, then became part of Google's core ranking systems in 2024. Treating it today as an unchanged, standalone checklist misses that its architecture and terminology have shifted.
Page experience is worth being especially careful with, because Google's own current documentation holds several ideas at once: there's no single page-experience signal, Core Web Vitals are used by ranking systems, other page-experience aspects don't directly help a page rank, and yet overall experience is still worth improving because it affects whether people stick around once they arrive. A perfect score on every metric isn't necessary and won't save thin or irrelevant content.
AI search adds another layer worth separating from ordinary web ranking. Google says established SEO practices remain relevant to AI Overviews and AI Mode, and it describes a process called query fan-out, where related searches help surface supporting material. But AI Overviews and AI Mode can use different underlying models and techniques, so what gets pulled into one won't always match the other, and neither is a guaranteed extension of your normal organic position. Ranking well in standard search is a reasonable foundation for AI citation, not a guarantee of it, and the two are worth tracking separately rather than assuming one follows automatically from the other.
What should a marketing team actually prioritize
None of this is a disclosed algorithm you can reverse-engineer. It's a set of documented mechanisms you can act on with reasonable confidence, which is a more honest foundation than chasing a checklist.
Match the page to the reader's actual question before anything else. Relevance is the most consistently confirmed mechanism on this whole list, and it's also the one teams skip past fastest in favor of tactics.
Provide information that's useful, reliable, and genuinely original, rather than a repackaged version of what's already ranking. This is where content length decisions should come from the question, not a word count target.
Make important content easy to find and understand, structurally and in plain language. This has more to do with keyword placement and page clarity than with hitting a density number.
Use links and site organization because they genuinely help readers and search systems understand your content, not because you're chasing a backlink quota.
Improve real performance and usability, but treat a perfect Core Web Vitals score as a reasonable baseline, not a ranking lever on its own.
Build credible, consistent author and brand signals over time instead of treating any single page element, like an author bio, as a guaranteed boost, and keep an eye on third-party authority metrics as directional context rather than a Google score.
Measure your actual traffic and business outcomes separately from tool scores, dwell time estimates, or AI citation counts. Each of those is a different kind of signal, and conflating them is how teams end up optimizing for the wrong number.
What would change any of these grades is a new, explicit Google statement, a dated primary record showing a mechanism in use, or stronger causal evidence for a narrower version of a claim. A single new correlation study shouldn't promote anything straight to confirmed on its own, and that's really the whole point of keeping a proven ranking factors list separate from a speculative one instead of merging them for the sake of a tidier page.
