Most ranking-factor articles are a list of 200 things with no weighting. That’s useless. If everything is a ranking factor, you have no idea what to do on Monday morning.
This page is organized differently. It’s grouped by what you can actually control and how much it’s likely to matter, and it’s explicit about the difference between three very different kinds of claim:
- Confirmed by Google. Documented in Search Central, the Search Quality Rater Guidelines, or stated directly by Google Search staff.
- Strongly evidenced. Not confirmed, but consistent across a lot of independent testing and site-level observation. Reasonable to act on.
- Speculative. Plausible mechanism, weak evidence. Worth knowing, not worth betting a budget on.
Google does not publish weightings. Anyone who tells you “backlinks are 32% of the algorithm” made that number up. Search is a machine-learned system with hundreds of signals whose relative influence shifts by query type — a medical query and a “best hiking boots” query are not scored the same way. So treat the tiers below as priority order for effort, not as a leaked formula.
One more thing before the list. Google’s public position, repeated for years, is that you should not optimize for ranking factors at all — you should build something people want, and the systems are meant to find it. That is genuinely how the machine learning is trained, and it’s also convenient PR. The truth sits in between: the signals are real and some are directly actionable, but chasing signals in isolation, divorced from whether the page is any good, is exactly what stopped working after the Helpful Content update rolled into core ranking in March 2024.
Table of Contents
Tier 1: The things that decide whether you rank at all
Whether the page satisfies the query better than the alternatives
Confirmed by Google, in the sense that this is the stated objective of the entire system.
This is not a soft “write good content” platitude. It’s an operational instruction: pull the top ten results for your target query, read them, and identify what a searcher still doesn’t have after reading all ten. That gap is your page.
Search intent is the first filter. If the SERP for your term is dominated by comparison lists, a service page will not rank there no matter how well optimized it is — Google has already decided what format that query wants. Check the SERP before you write, not after you fail. Intent mismatch is the single most common reason a technically clean, well-linked page never ranks, and no amount of on-page work fixes it.
Google’s guidance on people-first content asks a set of questions worth answering honestly about any page you publish: does it demonstrate first-hand expertise, does it leave the reader feeling they need to search again, was it made for people or to hit a keyword. The March 2024 core update absorbed the helpful content system into core ranking and Google stated at the time that the changes were intended to reduce low-quality, unoriginal content in results by around 40%. There is no separate “helpful content classifier” to escape any more. It’s baked in.
Strongly evidenced.
Google doesn’t rank pages in isolation. It has a model of what your site is about, built from every page, every internal link, and the entities you consistently discuss. A single good page on a site with no demonstrated depth in that subject competes at a disadvantage against a page from a site that has covered the subject comprehensively.
This is why one 3,000-word article about a topic rarely outperforms a cluster of eight interlinked pages that cover the topic’s actual sub-questions. The cluster tells Google what you specialize in. The one-off doesn’t.
Practically, this means picking a subject you can plausibly own, mapping its sub-topics from real queries rather than from a keyword tool’s related-terms tab, and building pages for each with clear internal linking between them. Cover the entities properly — the people, products, standards, and concepts a knowledgeable writer would naturally mention — rather than repeating the head keyword.
how topical authority is built and semantic SEO go into the mechanics of this in more detail.
The honest caveat: “topical authority” is not a metric Google has ever named or confirmed. It’s a practitioner model that predicts outcomes well. Treat it as a useful map, not the territory.
Crawlability and indexation
Confirmed by Google.
Unglamorous and absolute. If Googlebot can’t fetch the page, render it, and index it, nothing else on this list applies.
The failure modes are boring and common: a noindex left on after a staging migration, a robots.txt rule blocking a directory, canonical tags pointing at the wrong URL, parameter-driven duplicates diluting a page across twenty variants, JavaScript that renders the main content only after a user interaction Googlebot won’t perform, or a site so thin on internal links that half of it is orphaned.
Check Search Console’s page indexing report before you spend a dollar on content. “Crawled – currently not indexed” at scale is Google telling you it fetched your pages and judged them not worth storing. That is a quality signal, not a technical bug, and no amount of sitemap resubmission fixes it.
Links from other sites
Confirmed by Google as a ranking signal, with heavily disputed weighting.
Links still matter. Google staff have described their importance as declining relative to earlier years, and internal Google documents surfaced in 2024 indicate a far more nuanced link-evaluation system than the simple PageRank model most SEO content still describes. Both things can be true: links are less decisive than in 2015 and still one of the strongest differentiators in competitive commercial queries.
What matters, in rough order:
- Relevance of the linking page. A link from a page about your subject beats a link from a high-authority page about nothing in particular.
- Whether the link would exist without you asking. Editorially placed links inside real content behave differently from footer links, directory listings, and paid placements.
- Domain diversity. Fifty links from five domains is worth less than fifteen links from fifteen relevant domains.
- Anchor text, with a strong caution — over-optimized commercial anchor text at scale is one of the more reliable ways to attract a manual action.
Third-party “Domain Authority” and “Domain Rating” scores are vendor metrics. They are not Google metrics, Google does not use them, and buying links by DR is how link budgets get wasted. They are useful as rough filters, nothing more. what backlinks are and how they actually work covers the fundamentals.
Tier 2: Real, and often the difference between position 8 and position 3
Site architecture and internal linking
Strongly evidenced, and the most under-used lever on most sites.
Internal links are the one link signal you fully control, and most sites use them badly — a handful of links dropped in a “Related Posts” widget and nothing else.
What works: important pages reachable within two or three clicks of the homepage; descriptive anchor text that says what the destination is about rather than “click here” or “read more”; contextual links inside body copy pointing from supporting pages up to the page you want ranking; a flat-enough structure that link equity isn’t stranded five levels deep.
If you have one page you need to rank, find every existing page on your site that discusses that topic and add a genuine contextual link from each. It’s free, it’s fast, and it’s the highest-return hour in most SEO engagements.
On-page fundamentals
Confirmed by Google as inputs; low ceiling.
The title tag is still one of the more direct on-page influences on ranking and by far the biggest influence on click-through. Google rewrites titles frequently — reported studies have put rewrite rates in the range of a third to over half of results depending on methodology — so write titles that are accurate and specific rather than padded, which is exactly what reduces the odds of a rewrite.
H1 and heading structure help Google and readers understand the page. They are a modest signal at best. Meta descriptions are not a ranking factor — Google has been clear on this for years — but they influence CTR, which is worth optimizing for its own sake.
URL structure: keep it short, readable, and stable. Don’t rewrite URLs for keyword reasons on an existing site; the redirect risk outweighs the marginal gain.
Structured data does not directly raise rankings. It qualifies you for rich results and gives machines an unambiguous read on what the page describes, which increasingly matters for AI surfaces. Both are worth having.
Core Web Vitals and page experience
Confirmed by Google as a ranking signal, with confirmed-low weight.
Current thresholds for a “good” experience:
Metric |
What it measures |
Good |
|---|---|---|
LCP (Largest Contentful Paint) |
Loading — when the main content appears |
≤ 2.5s |
INP (Interaction to Next Paint) |
Responsiveness — lag from input to visual response |
≤ 200ms |
CLS (Cumulative Layout Shift) |
Visual stability — unexpected movement |
≤ 0.1 |
INP replaced First Input Delay as a Core Web Vital in March 2024. If your reporting still shows FID, it’s stale. INP is a harder metric to pass than FID was, because it measures every interaction across the session rather than just the first one, and it punishes heavy JavaScript on the main thread — the tag manager sprawl, the chat widget, the seven analytics scripts.
Be realistic about the payoff. Google has repeatedly framed page experience as a tiebreaker, not a lever that lifts a mediocre page over a better one. Going from failing to passing is worth doing. Going from 92 to 98 in Lighthouse is not an SEO project. The real argument for speed is conversion rate, and that argument stands on its own.
Mobile-first indexing
Confirmed by Google, and now unavoidable.
Google completed the move to mobile-first indexing for all sites in July 2024. Googlebot indexes the mobile rendering of your page. If content, structured data, or internal links are hidden or stripped on mobile, they effectively do not exist.
This is now a hygiene check rather than a project: view your page on a real phone viewport and confirm the same content is present. Separate m-dot sites are a legacy liability at this point.
E-E-A-T, applied where it belongs
Confirmed as a rater-guideline framework; commonly over-applied.
Experience, Expertise, Authoritativeness, Trust comes from the Search Quality Rater Guidelines — the manual used by human evaluators who assess search quality. It is not a score. There is no E-E-A-T number in the algorithm, and Google staff have said so directly.
Where it genuinely matters: “Your Money or Your Life” topics. Health, finance, legal, safety, major life decisions. For those queries the raters — and, indirectly, the systems trained against their judgments — weigh who is publishing, what their credentials are, and whether the site is trustworthy far more heavily.
Where it’s over-applied: a plumber’s service page in Phoenix does not need an author bio with academic credentials and a medically-reviewed-by badge. Adding one is theater. What that page needs is a real business address, a working phone number, actual photos of actual work, and reviews that exist.
The Trust component is the one that carries most of the weight, and it’s the most concrete: clear contact information, honest pricing, no deceptive layouts, a real business behind the site.
Freshness — and when it doesn’t matter
Confirmed as query-dependent.
Freshness is not a universal signal. It applies to queries that deserve freshness — news, prices, “best X” comparisons, anything version-dependent, anything seasonal.
For an evergreen explainer, the publication date is close to irrelevant, and republishing it with a new date and no substantive change accomplishes nothing except training your team to do busywork. Google’s systems evaluate whether the content changed, not whether the timestamp did.
Two related habits worth dropping: auto-inserting the current year into titles via a shortcode, which produces pages claiming to be updated when they aren’t, and “refreshing” content by swapping synonyms. Update pages when the underlying facts changed. Otherwise leave them alone and write something new.
Tier 3: The AI layer, and what it changes
Partly confirmed, largely still moving.
AI Overviews and LLM-based answer engines change distribution more than they change ranking. The signals that get a page into the classic top ten are largely the same signals that get it selected as a source for a generated answer — but the traffic economics on the other side are different.
What’s reasonably established:
- AI Overviews and AI Mode draw heavily from pages that already rank well for the query and its related sub-queries. Conventional ranking work is still the price of entry.
- Content structured as clear, self-contained answers — a direct statement immediately under a question-shaped heading — is easier for a model to lift and cite than the same information buried in paragraph twelve.
- Being cited on a query where the answer is fully satisfied in the overview will cost you clicks. Multiple published analyses through 2024 and 2025 found meaningful organic CTR declines on informational queries where an AI Overview appears. The exact magnitudes vary widely by study and query type, so treat any single headline number with suspicion.
What’s speculative and sold as fact far too often:
- That there is a distinct “GEO ranking algorithm” you can optimize separately. There isn’t a documented one.
- That specific formatting tricks reliably force citation. Testing here is early, noisy, and mostly unreplicated.
- That entity mentions on third-party platforms directly cause LLM citations. Plausible mechanism, thin evidence, heavily oversold.
The defensible strategic response is unglamorous: build genuine topical depth so you’re a natural source across a whole query cluster rather than one keyword, structure answers cleanly, keep your entity information consistent everywhere it appears, and shift measurement toward queries where a click still has to happen — commercial, comparison, and bottom-of-funnel terms. our approach to AI search visibility covers how we handle this in practice.
Also worth saying plainly: if a large share of your traffic came from thin informational content that answers a question in one sentence, that traffic is structurally at risk and no optimization recovers it. The fix is a different content strategy, not a better one.
Tier 4: Dead. Stop doing these.
These persist in ranking-factor listicles — including the earlier version of this page — because listicles are copied from other listicles. Each one below is genuinely finished.
Keyword density. There is no target percentage. There never was a public one, and Google representatives have said repeatedly there is no ideal density. Modern retrieval works on meaning, not term frequency ratios. Writing to hit 2.5% produces worse copy that ranks worse. Mention the topic naturally, cover the related concepts, move on.
“LSI keywords.” Latent Semantic Indexing is a real information-retrieval technique from the late 1980s. It is not what Google uses, and Google’s John Mueller has stated flatly that there is no such thing as LSI keywords. The underlying advice — cover semantically related concepts — is sound. The label is nonsense, and tools sold on it are selling a synonym list with a technical-sounding name.
Meta keywords tag. Google confirmed in 2009 that it does not use the meta keywords tag for ranking. That was seventeen years ago. If your CMS still has a field for it, ignore the field.
Exact-match domains. The EMD update in September 2012 specifically targeted low-quality exact-match domains. A keyword in the domain provides no meaningful ranking benefit today, and a domain like bestcheapseoservicesnyc.com signals spam to humans, which is its own cost. Pick a brandable domain. Brand recognition is a more durable asset than a keyword string.
Submitting your site to search engines. Google retired its public URL submission ping tool, and the “submit to 500 search engines” services were snake oil long before that. Google finds pages through links and sitemaps. Publish a sitemap, get linked, done. If you need a specific URL indexed quickly, use the URL Inspection tool in Search Console — that’s the only submission mechanism worth using.
Article spinning. Rewriting an article with synonym substitution to produce “unique” versions violates Google’s spam policies on scaled content abuse, produces text no human wants to read, and is detected easily. This applies equally to the modern version: generating hundreds of near-identical AI pages. Google’s policy is explicitly about the purpose and quality of the content, not the method used to produce it — AI-generated content is not banned, but content generated at scale primarily to manipulate rankings is.
Domain age as a factor. Google has said domain age itself isn’t a ranking factor. Older domains often rank better because they’ve accumulated links, content, and brand signals over time. The age is a correlation, not a cause. Buying an aged domain buys you nothing on its own.
Comment counts, social share counts, and bounce rate as direct signals. Social shares are not a direct ranking signal — Google has said so consistently. Bounce rate as measured in your analytics is not a signal Google can even see. Engagement can correlate with quality; the specific metrics people cite are not inputs.
Word count targets. There is no minimum length that ranks. Long pages often rank because thorough coverage tends to require length, not because length is rewarded. Padding a 700-word answer to 2,000 words makes it worse.
How to actually sequence this work
If you’re starting from a site that isn’t performing:
- Confirm the pages are indexable and indexed. Search Console, page indexing report. Fix blockers before anything else.
- Check intent alignment on your money pages. Pull the SERP. If the format doesn’t match, the page needs rebuilding, not optimizing.
- Fix internal linking. Cheapest meaningful win available. Do it before you commission new content.
- Build depth on one topic, not breadth across ten. A complete cluster beats scattered pages. the Keyword Golden Ratio method is a useful filter for finding sub-topics you can rank for while authority is still building.
- Get Core Web Vitals from failing to passing. Then stop.
- Earn links deliberately, to the pages that need them. Relevance over volume, always.
That order isn’t arbitrary. Each step makes the next one work better, and steps 5 and 6 are largely wasted effort if steps 1 through 3 are broken.
FAQ
What is the most important Google ranking factor? There is no single most important factor, and Google has never published weightings. In practice, the strongest predictors are whether the page matches search intent better than competing results, whether the site demonstrates real depth on the topic, and whether relevant sites link to it. Relative importance shifts by query type — links matter more in competitive commercial queries, trust signals matter more in health and finance queries.
Is keyword density still a ranking factor? No. Google representatives have stated there is no ideal keyword density and no target percentage. Modern search systems evaluate meaning and topical coverage rather than how many times a term appears. Writing to hit a density number typically produces worse content that performs worse.
Are backlinks still important for SEO? Yes, though less decisive than a decade ago. Google confirms links are a ranking signal, and Google staff have described their relative importance as declining. Relevance and editorial context matter more than raw volume, and third-party authority scores like DA or DR are vendor metrics that Google does not use.
What replaced First Input Delay in Core Web Vitals? Interaction to Next Paint (INP) replaced First Input Delay in March 2024. INP measures responsiveness across all interactions during a page visit rather than only the first, with 200 milliseconds or less considered good. It is harder to pass than FID because it exposes main-thread JavaScript problems throughout the session.
Does E-E-A-T apply to every website? No. E-E-A-T comes from Google’s Search Quality Rater Guidelines and carries the most weight on “Your Money or Your Life” topics — health, finance, legal, and safety. For a local service business, the practical requirements are trust signals: real contact details, honest pricing, verifiable work, genuine reviews. Author credentials and review badges on non-YMYL pages are usually decoration.
How often should I update content for SEO? Only when the underlying information has changed. Freshness is a query-dependent signal that applies to news, prices, comparisons, and version-specific topics. For evergreen content, changing the publication date without changing the substance accomplishes nothing. Google evaluates whether the content changed, not the timestamp.
Do I need to submit my site to Google to get indexed? No. Google discovers pages through links and XML sitemaps, and its public URL submission ping tool has been retired. Submit a sitemap in Search Console and use the URL Inspection tool if you need a specific page crawled sooner. Third-party “submit to hundreds of search engines” services provide no value.
How do AI Overviews change SEO? Mostly by changing distribution rather than ranking. Pages cited in AI Overviews are generally pages that already rank well for the query, so conventional SEO remains the price of entry. The change is on the traffic side: informational queries fully answered in the overview generate fewer clicks, which shifts the value toward commercial and comparison queries where a visit is still required, and toward topical depth that makes a site a repeat source across a whole query cluster.
Where this leaves you
Most SEO underperformance isn’t caused by missing an obscure ranking factor. It’s caused by one of four things: content that doesn’t match what the query wants, a site with no demonstrated depth on the subject, technical problems keeping pages out of the index, or no one else on the web finding the site worth referencing.
Fix those in that order and the long tail of minor factors mostly takes care of itself. Chase the long tail first and you’ll spend a year optimizing alt text on a page Google was never going to rank.
If you want a specific read on which of those four is limiting your site — rather than a generic checklist — that’s the assessment we run before any engagement starts.