Word count is not a ranking factor.

Where the number came from

Someone took the top ranking pages for a large sample of queries, measured them, and found the winners were longer on average than the losers. That study was real and it has been repeated. The mistake was in what happened next: a measurement of what winning pages look like got rewritten as an instruction for how to win, and the average length of a result became a minimum to hit.

Google has said plainly and repeatedly that there is no ideal article length and no word count threshold. There could not be one. The best possible answer to "what is the capital of France" is three words, and the best possible answer to "how do I migrate a site without losing rankings" is a few thousand. A single number that was correct for both would be describing nothing.

What the correlation is actually measuring

The correlation is not fake, it is just pointing at something else. Pages that cover a subject thoroughly tend to be longer, because covering a subject thoroughly takes words. They also tend to answer more of the surrounding questions, attract more links, and satisfy more of the people who land on them. Length is the by-product of all of that, sitting downstream where it is easy to measure.

So the honest version of the finding is: comprehensive pages rank well, and comprehensive pages are usually long. Reverse it into "long pages rank well, so make pages long" and you have kept the only part with no causal power. You can add a thousand words to a page in an afternoon. You cannot add understanding to it that way, and understanding was the thing doing the work.

What padding actually costs

Writing to a target has predictable symptoms: a preamble explaining why the topic matters, a history section nobody asked for, a definition of a term the reader already knows, and the actual answer somewhere around the seventieth percent of the page. Every one of those is a reason for someone to leave before reaching the part they came for.

It costs something on the machine side too. Search engines and language models both reward a page where a clear answer sits in a passage that stands on its own. Burying that answer under several hundred words of throat clearing makes it harder to extract and harder to quote, which is now a live cost rather than a theoretical one. That is a large part of what LLM SEO is actually about, and padding works directly against it.

The one place length genuinely matters

Thin content is a real problem, and this is where the instinct behind the word count comes from. A page with almost nothing on it, offering nothing that is not already elsewhere, will struggle, and Google may well decide it is not worth indexing at all. That is the situation the crawled, currently not indexed status usually describes.

But notice the diagnosis. The problem with a thin page is that it has nothing to say, and the word count is just the visible sign of that. Adding words without adding substance treats the symptom and leaves the disease. The fix is either to give the page something genuine to offer, or to merge it into the page that already does, which is more often the right call.

The general lesson

This is the same shape as keyword density: a number that is trivial to compute, easy to put in a brief, easy to grade, and disconnected from the thing it is standing in for. Both survive because they turn a hard editorial judgement into an easy numeric one, and both produce worse pages the moment they become the goal.

Write until the question is answered properly, then stop. If that takes 300 words, the page is finished at 300 words. The only length target worth having is the one the subject sets.