How long should a blog post be?

Ask five people and get five answers, all of them confident. Here is what the data actually says — including the part that contradicts the last decade of advice.

The number everyone quotes, and where it came from

In 2014 a company called SerpIQ published a study showing that top-ranking pages averaged around 2,000 words. It was cited everywhere. It became, more or less permanently, the thing people say when someone asks how long a blog post should be.

The study was not wrong. The conclusion drawn from it was. Top-ranking pages were long because they tended to cover more of a topic and answer more of what a searcher wanted — not because length itself was doing anything. A decade of content strategy was built on mistaking the symptom for the cause, and a great many perfectly good 1,200-word articles were inflated to 2,500 in the belief that this was helping.

Google's own Gary Illyes has put it about as plainly as anyone could: "There's no ideal length for blog posts."

Which is true, and completely unhelpful if you are sitting in front of an empty document trying to decide whether this thing is a 900-word piece or a 2,200-word piece. So here are the numbers, and then a way to use them.

Different studies land in different places — you will see figures from 1,152 to 2,450 depending on who crawled what and when. That spread is itself informative. If there were a magic number, the studies would converge on it. They do not, because there is not one.

What changed, and why it matters now

Here is the part that genuinely undercuts the received wisdom, and most length guides have not caught up with it yet.

Ahrefs examined 560,346 AI Overviews and 174,048 pages cited in them — the largest look at this question so far. The average length of a cited page was 1,282 words. That is barely longer than the average ranking page, which is unremarkable. What is remarkable is the rest of it.

More than half the pages being cited are short. And the statistical relationship between how long a page is and whether it gets cited is not weak — it is essentially absent.

Which makes sense once you think about what an AI Overview is doing. It is not rewarding a page for being thorough. It is looking for a passage that answers a specific question cleanly enough to quote. A tight 700-word piece that answers one question properly is a better candidate than a 3,000-word omnibus where the answer is buried in section six.

So there are now two games. Traditional ranking still mildly favours comprehensiveness. Citation favours clarity and structure. The good news is that they are not in conflict — a long piece built out of clearly-answered questions serves both. A long piece built out of padding serves neither.

Length by what the reader came for

This is more useful than any single number, because it starts from the reader rather than from a target. Someone searching "what time does the shop close" and someone searching "how to restructure a marketing team" want very different amounts of your writing.

The floor: how short is too short

There is no official minimum. Google has never published one and says repeatedly that it does not have one. But there is an observable floor in practice, and it sits somewhere around 300 to 500 words — below that, a page starts looking like what Google calls thin content, and it rarely surfaces for anything competitive.

The exceptions are real and worth knowing: news items, definitions, reference pages, anything where brevity is the service being provided. A 200-word page that answers a factual question completely is not thin. A 200-word page that gestures at a big topic is.

The honest test is not the count. It is whether a reader arriving with a question leaves with an answer. If they have to go somewhere else afterwards, the page was too short regardless of how many words it contained.

The ceiling: where longer stops paying

Long content does have one durable advantage, and it is a real one: pieces over 3,000 words earn substantially more backlinks — Backlinko has put the figure at around 77% more on average. People cite the definitive thing. Being the definitive thing usually takes room.

But the returns flatten, and then they reverse. A 6,000-word post rarely outranks a 3,000-word post on the same topic unless the extra 3,000 words contain something genuinely new. And padding is not neutral — readers bounce faster from bloated pages, and that behaviour is itself a signal. You can actively damage a good article by making it longer.

The useful way to hold it: length is a consequence of having something to say, not a strategy for appearing to.

How to pick a number for your piece

This takes about fifteen minutes and beats every rule of thumb on this page, because it uses the only evidence that matters — what is already working for your exact query.

1. Search your keyword in a private window, in the country you are writing for. Ads off, your own site ignored.
2. Take the top ten organic results and paste each one's body text into a word counter. Skip navigation, footers and sidebars — you want the article, not the furniture.
3. Note the average, and note the outliers. A thin page ranking on brand strength alone tells you nothing about what you need to write. A forum thread ranking tells you something else entirely — that the searcher wanted opinions, not an article.
4. Aim within about 20% of that average, then ignore the number and ask the better question: what do all ten of them fail to cover? That gap is your reason to exist on the page.

If the top ten average 1,600 words and you have 900 words of genuinely superior material, publish the 900. If you have 2,400 words of genuinely new material, publish that. The average is a sense-check, not an instruction.

How to tell padding from depth

Everyone agrees padding is bad and almost nobody can spot it in their own draft, because it never feels like padding while you are writing it. It feels like being thorough. A few reliable tells:

• The definitional opening. "Before we discuss X, let's define what X is." If your reader searched for X, they know what X is. This paragraph exists because it was easy to write.
• The restated point. Long pieces often make the same argument three times in different costumes. Read your H2s in sequence — if two of them are the same idea, one of them is padding.
• The section that could be a sentence. Four hundred words establishing something everyone already accepts. Cut to the sentence and spend the words on what is contested.
• The delete test. Remove a paragraph and reread around the gap. If nothing is missing, nothing was there. This is brutal and it works. Our guide to word count covers the mechanics of cutting in more detail.

Questions people actually ask

Related: the complete word count guide, essay word counter, LinkedIn post counter — the one platform where longer measurably wins, X character counter.