How long should a blog post be, and why that question fails

The short answer is that there is no number, and it is not an evasion: the question measures the thing that decides nothing. Google does not count words, and nobody has shown a document where it does. What does happen is that long articles tend to rank better, and not because they are long, but because a query that demands a complete answer produces long text in whoever answers it well. Confusing the consequence with the cause is what fills blogs with 2,500-word articles where 600 would have fit.

How long should a blog post be to rank?

There is no minimum length. The right length is the one that covers the query and its follow-up questions without filler: roughly 300 to 600 words for a single fact, 800 to 1,400 for a definition, and 1,500 to 2,500 for a procedure or comparison. Google does not measure length, it measures whether the page answers.

Quick answer

• There is no minimum. Length is set by the query, not by a rule.
• The correlation is real and read backwards. Long does not cause position; intent causes both.
• The real minimum is covering the question and its follow-ups. When it is covered, you stop.
• Filler has a cost, and in the age of AI summaries it is a bigger cost than before.

Where the myth comes from

There are studies showing that top-ranked results have, on average, more words than the worst-ranked ones. The figure is real and the conclusion hung on it does not follow.

Think about it from the other side. Competitive queries tend to be the complex ones too: "how to migrate a site without losing rankings" needs more text than "how many grams in a pound". The sites competing on those queries invest more, cover more and write more. Length is a symptom of the query type and of the effort invested, not an ingredient.

Proof that the relationship is not causal is available to anyone: take a 600-word article that ranks and stretch it to 2,000 with filler. It does not rise. In the best case nothing happens; in the worst, the answer that was on top gets buried and the extraction is lost.

What the real minimum is

There is one, and it is not measured in words: the query covered, plus the three or four questions the reader asks immediately after.

An article on what a service costs that gives a range and stops is incomplete, not because it is short but because whoever asks the price then asks what is included, why it varies, and how to compare two quotes. Covering those follow-ups might take 900 words or 2,000, depending on the topic, and the resulting number was never the goal.

The practical way to find them is free: the "people also ask" block on the results page itself, and the Search Console queries report if the article already exists. Both sources, with the method to work them without paid tools, are in free keyword research. That change of criterion has a non-obvious advantage: a list of questions can be finished, and a word count cannot. You know the article is complete because no reasonable question is left unanswered, and that moment arrives sooner than the fear of coming up short suggests.

And a maximum, which almost nobody mentions

Filler always had a credibility cost. Now it has a measurable one.

If the answer to the query is in the ninth paragraph, behind an introduction about how important the topic has become, there is an AI summary that will give it first and without citing you. The structure that survives puts the answer on top, in forty to sixty words that stand on their own, with the development below. That does not shorten the article: it changes the order, which is what mattered. How that block is written is in how to write SEO content, why the summary reaches it first is in what AI Overviews are, and what makes a block quotable by those engines is in how to get cited by ChatGPT.

There is a second cost, and it is opportunity. A thousand words of filler in each article, over fifty articles, is fifty thousand words that did not go to covering fifty more queries.

What Google has said, in its own words

The argument above stands on its own, but the quote is worth having, because this query needs more than reasoning.

John Mueller, of Google, in August 2019: "from our point of view the number of words on a page is not a quality factor, not a ranking factor", reconfirmed on 16 September 2026. And on the specific practice of copying the length of whoever ranks first: matching the word count of a well-ranked article will not make your pages rank first, the way collecting a pile of USB chargers will not get you to the moon. Google's own guidance on creating helpful content closes it too: among its self-assessment questions it asks whether you are "writing to a particular word count because you've heard or read that Google has a preferred word count", and answers itself, "No, we don't."

Now the interesting part, which is what happens on the first page of this very query. Checked on 16 September 2026, the first results open by saying there is no fixed number, and then give one: that first-page articles average 1,400 to 1,900 words, that an informational post wants 1,000 to 2,000, that the minimum to rank is 300.

That contradiction is not carelessness. The correct answer cannot be sold and a number can, so the article says the true thing in the opening and the profitable thing in the body. The reader keeps the number, which is the only actionable thing they were handed, and starts filling until they reach it. Of the three, the worst is the 300-word minimum, also the most repeated. It comes from no Google documentation, corresponds to no known threshold, and has circulated from article to article for years. An eighty-word product page ranks perfectly well if it answers the query's intent; what does not rank is a page that answers nothing, at 80 words or 3,000.

When a short article is the right answer

When the query is exhausted in a sentence. A fact with its source and the context to read it can be complete in 400 words.

When a long article on the topic already exists. Stretching a short piece to "compete" with it usually produces an internal duplicate, and a query already covered does not need a second page: the trap is the content brief that specifies a length instead of the questions to answer.

When the goal is to convert, not to inform. A long service page rarely sells more than a short, clear one.

When there is nothing more to say. It is the most common reason and the least accepted.

Mistakes that repeat

• Writing to hit a number. Filler shows and buries the answer.
• Reading the correlation backwards. The query produces the length, not the other way around.
• Copying the length of whoever ranks first. They may be on top despite it.
• Putting the answer in the ninth paragraph. The AI summary gives it first and without citing you.
• Stretching a short article that works. You risk what already performs for a number.
• Borrowing a band without its format. The range comes from the structure, not the reverse.

Data and transparency

The John Mueller quote that word count is not a quality or ranking factor is from August 2019 and was reconfirmed on 16 September 2026; the USB-chargers-to-the-moon comparison is his as well. Google's helpful-content self-assessment question about writing to a preferred word count, and its "No, we don't", come from Google's public guidance, checked on 5 September 2026. That the first results for this query open by denying a fixed number and then supply one (averages of 1,400 to 1,900 words, 1,000 to 2,000 for an informational post, a 300-word minimum) is an own observation from 16 September 2026, and those figures are recorded as examples, not adopted as reference. The 300-word minimum comes from no Google documentation.

That Google does not use text length as a ranking factor is a position its spokespeople have stated publicly and repeatedly; what exist are third-party studies observing a correlation between length and position, and correlation is not the mechanism. None of those studies is cited with a specific figure on purpose: they circulate in several versions, their conditions do not transfer, and using them here would reproduce the error this article describes. The length bands by query type in the table are operating judgment from work across a portfolio recording more than 300 million impressions a year in Search Console, not a published figure. Verified as of September 2026.

Primary source, opened 16 September 2026: John Mueller on word count, via Search Engine Journal.

What this changes

The word-count question survives because it is the only one in all of SEO with a comfortable answer, which is why it gets asked so much.

Trading it for another one costs more and works. Instead of how many words the article needs, which questions it has to answer so that nobody searches again. That list comes out of the results page in ten minutes, admits no filler because each item is a specific question, and by the time it is finished the word count already came out on its own. It was never the goal, it only looked like it because it is the only thing you can count without reading.

Frequently asked questions