Google Ranking Factors That Actually Matter

Which ranking factors carry the most weight?

Intent-matched content, E-E-A-T signals, authoritative and topically relevant backlinks, and Core Web Vitals. In practice, the query's presence in the title tag, the trustworthiness of the domain, mobile-friendliness and page speed account for a large share of what separates position 3 from position 13.

Content and intent match

Does the page answer what the searcher meant, in the format they expected? The largest single lever, and the one you fully control.

E-E-A-T

Experience, expertise, authoritativeness and trust. Not a score you can read anywhere - it is inferred from authorship, citations, reputation and the company you keep in the link graph.

Relevant referring domains

Distinct sites, weighted by topical relevance. Ten links from your category beat a hundred from unrelated directories.

Core Web Vitals

LCP, INP and CLS, now aggregated into a composite performance score - failing one metric costs more than it used to, even when the other two pass.

What about the factors people argue over?

Exact-match domains, keyword density, meta keywords and word count are not meaningful factors. Domain age is a proxy for accumulated links, not a factor in itself. Treat anything expressed as a magic number with suspicion.

The reliable test is mechanism: can you explain how the signal would help a search engine return a better result? Keyword density fails that test - a page repeating a phrase is not more useful than one that says it once and answers the question. Referring domains pass it easily: an independent site choosing to link is evidence of value that is expensive to fake.

The meta description is a special case. It is not a ranking factor, but it is the text under your title in the results, so it strongly affects whether anyone clicks - and engagement is a factor. Write it as ad copy, not as a summary.

How much do Core Web Vitals actually matter?

They are a real but secondary factor: rarely enough to outrank better content, reliably enough to lose positions when they are bad. The composite scoring introduced in 2026 means a page failing one of the three metrics is penalised more than the old pass-two-of-three behaviour allowed.

The practical reading is that Core Web Vitals are a hygiene factor with a cliff. Going from good to excellent buys you very little. Going from failing to passing can be worth several positions on a competitive query, because you were being held down rather than pushed up.

Server response time deserves separate attention. It is not one of the three metrics, but a slow time to first byte drags LCP directly and eats crawl budget, which for a large site is its own ranking problem.

Do AI Overviews use different factors?

Mostly the same ones, weighted differently. Citation in an AI Overview correlates with topical authority, clear structure, a direct answer under a question-style heading, page-level trust and accurate structured data. Content that is easy to quote out of context is far more likely to be quoted.

The practical consequence is a formatting discipline rather than a separate strategy. Lead with the answer, keep each key passage self-contained, use headings that state the question, and make sure your structured data accurately describes the page rather than overstating it.

The site-level half of it is unchanged: authority and trust still gate whether you are a candidate at all. Formatting improves your odds among candidates; it does not make you one.