Ranking factors that hold up under testing

Three tiers of evidence, and why sorting matters

Ask five SEOs what the ranking factors are and you'll get five confident, different lists. Not because SEO is unknowable — because the field mixes three very different classes of claim and prints them in the same font.

Sort every claim into one of these tiers before acting on it. A ten-second habit that saves entire quarters.

Tier 1 — what Google has actually confirmed

Google's own "How Search works" documentation groups its systems into five buckets: the meaning of your query, the relevance of pages, the quality of content, the usability of pages, and context such as location and settings. Public, stable, and a better mental model than any list.

Underneath it, these specific things have been confirmed at one point or another.

• Relevance of content to the query. The oldest and largest. Google matches meaning, not strings — which is why synonyms do the work keyword repetition used to.
• Links from other sites. PageRank is Google's own patent and links have been confirmed repeatedly. What changed is weight, not existence. See where links still decide rankings.
• Page experience, including Core Web Vitals and HTTPS. Google describes these as a tiebreaker between comparable pages, not a way to overtake better content.
• Mobile usability. Google indexes the mobile version of your site. Content missing on mobile is missing from the index.
• Helpful, people-first content. Now part of the core ranking systems rather than a separate update, aimed squarely at pages written for search engines.
• Spam systems. SpamBrain and the link spam systems are confirmed and documented, and mostly neutralise bad signals rather than punish sites.
• Freshness and context. Freshness applies on queries that deserve it — news and definitions are treated very differently. Context means country, language and location. Two people typing the same words have always seen different results.

Tier 2 — strong signal, never confirmed

This is where competent SEOs live and where honest ones hedge. Everything here shows up consistently in testing, but Google either declines to confirm a direct signal or describes it as a byproduct of something else.

Tier 3 — the folklore that won't die

Several of these are still sold as line items on Indian SEO proposals. If one appears on yours, ask what evidence supports it and watch the answer arrive.

• Keyword density. There is no target percentage and there never was one. Writing to a density figure produces pages that read like they were written to a density figure.
• Domain Authority and Domain Rating. Third-party metrics from Moz and Ahrefs that approximate link strength. Google doesn't see or use them. Useful for comparing two link profiles, useless as a goal — see why DA is a vanity metric.
• The meta keywords tag. Google confirmed it ignores this in 2009. It's still on plugin checklists.
• Bounce rate from Google Analytics. Google has said it doesn't use Analytics data for ranking. Bounce rate is a business metric, not a search signal.
• Word count. No ideal length exists. Long pages often rank because thorough answers need length, not because length is rewarded.
• Social shares as a direct factor. Google has said they aren't one. Social drives distribution, which earns links, which are a factor. The chain matters, the shortcut doesn't.
• Submitting your site to 500 search engines. A 1998 service still invoiced in 2026.
• LSI keywords. Latent semantic indexing is a real 1980s technique Google has said it doesn't use. Related terms matter; the named service doesn't.

Where the "200 ranking factors" number came from

From Google itself, in the 2000s, as a rough figure used in talks to convey that ranking was complicated. It was never a list, and Google dropped the number for good reason: ranking now runs on machine-learned systems where one "factor" may have hundreds of variants, may apply only to certain query types, and may not be separable from the rest.

So a page titled "the 200 ranking factors, updated for 2026" is repackaging a two-decade-old talking point. It ranks because people search for it.

The reframing that helps: stop counting factors, start asking what problem Google is solving on this specific query. Most ranking questions answer themselves at that point.

The mix changes with the query

This is what makes universal lists useless. Google weights signals differently across query classes and has said so. Optimising a local clinic page the way you'd optimise a national SaaS page is a category error, not a tactical one.