What Google actually ranks on: speed, templates, and the parts that don't matter

Confirmed inputs: helpful content, Core Web Vitals, mobile usability, HTTPS — documented as considerations. Not inputs: which template, which builder, design taste; no builder field exists. Violations: doorway pages, cloaking, scaled content abuse, documented in the spam policies.

Google's page experience guidance is the source for the first group and is careful about its own strength: it describes a set of considerations rather than a single signal, and says great page experience does not override having content that is genuinely useful.

The middle group is the one agencies exploit, in both directions — selling a rebuild to escape a template, or selling a template as sufficient. Neither is a documented ranking input.

No. There is no builder field in any ranking system, no penalty for a template and no bonus for hand-written code. What templates do is make three other things likelier, and those three are documented factors.

Duplicate and near-duplicate URLs. Templates generate archive, tag and pagination routes that serve the same content at several addresses. Google's guidance on consolidating duplicate URLs is the fix — canonicals — and it is a configuration problem rather than a template problem.

Performance headroom spent before your content arrives. A theme that ships a slider, three font families and a page builder has spent the budget in the first category above.

Thin pages at scale. A template makes it trivial to publish forty pages that each say very little, which is a content problem the template merely made easy.

So the honest answer: the template is not the defect; it is a multiplier on configuration. Nothing in the documentation stops a well-configured template from outranking a badly configured custom build.

Yes, as a confirmed input, and the phrasing in the documentation is worth reading closely because the industry quotes the first half and stops.

The thresholds are numeric and published: LCP 2.5 seconds or less, INP 200 milliseconds or less, CLS 0.1 or less, at the 75th percentile of real page loads. Field data, not a lab score, which is why a perfect synthetic test and a failing report are not a contradiction.

Where speed does not move a position is beneath a page that answers the question better. Making a thin page fast produces a fast thin page. That is a common way a performance budget gets spent for nothing: a rebuild bought for speed, on a site whose real problem was that it never said anything a searcher needed.

Two things that are not the same as ranking and get sold as it: a Lighthouse score, which is a lab measurement of one load on one machine, and a "100/100", which is not a ranking input at any value.