How to Write a Blog Post

Key takeaways

A blog post needs a question it answers, an answer up front, and every section serving that answer — and, if it is worth writing, a structure that still works when it is read again later. This post explains how to write one, and it applies the rules to itself as it goes.

Decide the Question First

Before writing a word, decide the single question the post must answer. Not the topic — the question. "How to write a blog post" is a topic; "what makes a post answer its question and get read again?" is a question. The difference decides everything you write.

The thesis is the interpretation of the question. UNC's Writing Center states it plainly: a thesis "tells the reader how you will interpret the significance of the subject matter" and should pass the "so what?" test and the "how and why?" test. If you can write your answer in one sentence, you have a contract with the reader. The rest of the post either honors that contract or breaks it.

The question for this post is: how do you write a blog post that gets read, trusted, and re-read? The answer, stated now so we are clear: answer a durable question, put the answer first, structure for scanners, say how sure you are — and then publish.

Know How People Actually Read

The first fact to accept is uncomfortable: most people will never read your post. The evidence is remarkably consistent across two decades.

Two cautions before you overreact. First, these NN/g studies are practitioner-run usability research, not peer-reviewed population statistics — the numbers are strong and consistent, but treat them as directional. Second, "people rarely read" is not "people never read": motivated readers do read deeply (Moran 2020). The honest framing is conditional attention: your reader will decide in seconds whether to commit. Structure so that a skimmed post still delivers the point.

The "8-second attention span, shorter than a goldfish" figure you see everywhere is a myth — it comes from a 2015 Microsoft marketing report, not peer-reviewed research. Skip it.

Put the Answer First

The response to conditional attention is the inverted pyramid: answer first, details after. It is not a style choice; it is a response to how people read.

The strongest quantitative support: in Morkes & Nielsen's original usability experiment, rewriting pages to be concise, scannable, and objective improved usability by +124% over the "marketese" control (concise +58%, scannable +47%, objective +27%). If your conclusion sits at the bottom, most readers never see it. Put it at the top, and every reader leaves with the point even if they stop at paragraph two.

The same rule applies inside each section. First sentence: the point. Following sentences: the support. Readers scan first sentences — the F-pattern means the left edge and the early words get the most attention.

This post did it. You read the whole answer in the opening paragraphs. Everything below is the why and the how, in decreasing order of importance.

Rating: the inverted pyramid improving web comprehension is very likely — strong convergent practitioner and eyetracking support, limited peer-reviewed replication of the pyramid specifically.

Make It Scannable

Since most readers scan, structure for the scan. Four choices do most of the work:

For a post over a thousand words, add a short summary at the top. The reader decides in seconds whether to continue; the summary gives them the map.

Write for People, Then for Experts

Plain language is for everyone — even experts. NN/g's qualitative usability studies with science, technology, and medical professionals found that even highly educated readers "crave succinct information that is easy to scan, just like everyone else." Their recommendations: sentences of 15–20 words; aim for 10–12th-grade level for expert audiences; cut redundancies and pointless modifiers (Loranger 2017).

Be honest about what this does and does not do. Readability formulas measure proxies (word and sentence length), not comprehension itself. In a randomized study of consent forms, 62% preferred the simplified 7th-grade form and 97% found it easier to read — but comprehension was statistically indistinguishable from the 16th-grade version (58% vs. 56%, Davis et al. 1998). Plain language is valued and perceived as clearer; it is necessary, not sufficient.

There is also a real cost to needless complexity. In a series of experiments, students who deliberately used longer words to sound smart were rated as less intelligent (Oppenheimer 2006) — the effect is mediated by processing fluency. Probable, with modest effects; one-author lab studies deserve that hedge.

The classics agree. Orwell's sixth rule: "If it is possible to cut a word out, always cut it out." Strunk's Rule 13: "Omit needless words." Zinsser: "Clutter is the disease of American writing." These are advice, not data — but they have held for a century.

Say How Sure You Are

If you cite research, report how confident the evidence is. This separates a post you can trust from one you cannot.

Two axes matter, and they are different: the strength of the evidence, and the size of the effect. Strong evidence can point to a small effect; weak evidence can point to a large one. Conflating them is the most common error in science writing.

When you give a number, give the uncertainty around it. An effect size without a confidence interval is a headline, not a measurement. And translate it: "a d of 0.3" means nothing to most readers; "a learner using the method would outperform another roughly 58% of the time" means something.

Use a consistent scale for certainty — established, very likely, probable, speculative, inconclusive. This post uses it throughout. It takes one sentence, and it lets you re-read your own post without second-guessing it.

Writing Is Thinking

Here is the reason this method matters beyond the reader — the part that makes blogging worth it for the author.

Writing is a generation task, not a consumption task, and generation is what memory rewards. In the classic experiment, people remembered generated items better than read items across every measure — cued and uncued recognition, free and cued recall (Slamecka & Graf 1978).

It is also retrieval practice. In Roediger & Karpicke's landmark study, students who restudied scored better after 5 minutes (81% vs. 75%) — but after a week, the testers won: 56% vs. 42% (d = 0.83), despite reading the material far fewer times. And the restudy group predicted they would remember more. The thing that feels productive is the least productive; the effortful version is what sticks.

The broader principle is desirable difficulties (Bjork & Bjork): conditions that slow apparent learning often optimize long-term retention. Spacing and testing are the best-documented examples. Julia Evans describes the blogging version from practice: writing "really forces me to think about whether the lessons I think I learned actually make sense," and her blog functions as memory — "if I want to remember what I learned, I can just look it up."

Rating: that blogging specifically improves the author's memory is probable — the mechanisms are well supported (generation, retrieval, explanation), but no single experiment tests "blogging improves the author's memory" directly.

Structure for Rereading

The structure above makes a post answer its question and survive the scan. The third thing — getting read again — is mostly a matter of what the question was in the first place. Will this question still be worth answering in a year? A post about a durable problem gets re-read because the question persists. A post about a trend or a version number does not, no matter how well it is structured.

Gwern, whose site is the archetype of this, states the audience bluntly: "I am attempting to explain things to my future self, who is intelligent and interested, but has forgotten." He publishes long evergreen essays with explicit confidence metadata — the Kesselman list (certain to impossible) and importance tags — which is exactly why his essays get re-read for years.

Two choices make rereading more likely, and both are structural:

Cut What Doesn't Serve the Question

Every section must serve the question. If it does not, cut it — no matter how interesting it is.

The reverse outline finds the problem: after drafting, summarize each section in one line, in order (UW–Madison Writer's Handbook). Then test each line against the question. Sections that drift off-topic are the scope creep that sinks posts.

There is a test for this. Ask of every paragraph: so what? If the reader cannot say why it matters to the question, the paragraph does not belong. UNC puts it bluntly: omit evidence that does not serve the question.

Answer It at the End

The conclusion must answer the opening question explicitly. The reader who skips straight from title to conclusion should be able to reconstruct the full answer. The University of Wisconsin handbook describes the good conclusion as one that "revisits your main idea with new insight" — not a summary, but a return to the question that adds something.

The Missing Rule: Publish

The evidence and the writing craft only matter if the post ships. This is the part with no study behind it — it is practitioner consensus, and it is worth stating as strongly as the research.

Simon Willison, who blogged every day for a full year and has published over 12,000 posts: "My number one tip for blogging is to lower your standards! Aim to hit publish while you are still actively unhappy with what you have written, because the only alternative is a huge folder full of drafts and never publishing anything at all. Nobody will ever know how perfect the thing you intended to write would have been."

Julia Evans pairs it with an evidence-based insight: the fluency illusion means polished-feeling drafts overstate themselves. "Posts must be 100% correct" is a myth — write "I think," fix it later. Dan Luu keeps a public, dated list of his own major errors, and treats publishing with stated uncertainty as the default, not a failure.

The empirical note that makes this easier: your flaws are mostly invisible to readers, because they do not read most of what you write.

Answer One Question

Every rule in this post serves one habit: decide what question you are answering, then make every sentence pull toward that answer.

Structure is not a template. It is the visible shape of a single decision. A post that knows its question writes itself; a post that does not cannot be saved by formatting.

This post asked one question and answered it. You can verify the structure worked: read the opening paragraph, read the conclusion. Both give the same answer — answer a durable question, put the answer first, structure for scanners, say how sure you are, and publish. That is the test. If your post can survive reading only its start and end — now and in a year — the structure held.

Sources