Search Engine Optimization Content Writing, Step by Step

The pages ranking for search engine optimization content writing agree with each other almost completely. Put the keyword in your H2s. Keep URLs short. Add alt text. Write for readers first. Twelve tips, nine steps, six things to know, depending on which one you open.

All of that advice is correct and none of it is where the outcome gets decided.

Direct answer, then the working. SEO content writing is roughly two parts, which are choosing what to write and then writing it, and the choosing part determines most of the result while getting a small fraction of the attention. The tips lists cover the writing part thoroughly because it's teachable and safe. Below is the other part, built as a worked example on this exact keyword, including the artifact I put together before drafting this page and what it told me.

The Artifact, Built On This Page's Own Keyword

Before I write anything I make a coverage grid. It's the least glamorous file in the process and it's the one that decides whether the article has a reason to exist.

Here's the real one for this keyword, from pages I loaded on 28 July 2026.

Four columns, seven rows, and the bottom four rows are empty across the board. That's not a criticism of those pages, they're solving a different problem competently. It's the reason I had something to write.

My pipeline reported roughly 1,300 searches a month for the phrase, a difficulty score of 7, and a cost per click of $21.78. One soft result in the top ten, a Medium post sitting at position two, which is the signal I care about most.

If the grid had come back full, I'd have dropped the topic. That happens more than you'd think and it costs nothing, which is the point of building it first.

The Decision That Happens Before The First Sentence

Picking the target is the job. Everything after it is execution.

What I'm looking for isn't a low difficulty score. Difficulty scores get gamed by their own math, and I've written up two specific ways they mislead in how AI helps with SEO research, the short version being that keyword expansion drifts toward bigger adjacent topics and that navigation queries look easy because one company owns the whole page.

What I count instead is soft results. A forum thread, a Reddit post, a Q and A site, an old Medium article. Those hold a position because nothing better showed up, not because they're strong. One soft slot means a door exists. Zero soft slots and a page full of well-resourced vendors means you're not getting in on merit any time soon, whatever the difficulty number says.

The second thing I look for is a format gap, which matters more on commercial queries. If nine of the ten results are tool pages, an explanatory article isn't competing with them for the same job. It's answering a question they don't answer, for the slice of searchers who wanted an answer rather than a text box.

The Gap Statement

One sentence, written down, before drafting. If I can't produce it, I don't write the piece.

The form is always the same. This article contains X, and none of the current top ten does. For this page, X is the coverage grid above plus the documentation quotes below plus the actual volume and cost numbers, because not one ranking page has any of the three.

Vague versions don't survive. "Better and more comprehensive" is not a gap statement, it's a wish. "A worked example on the keyword the article itself targets" is one, because you can check whether it's true when the draft's finished.

This kills maybe a third of my ideas before they cost me a day. I've come to think of that as the main return on the whole exercise rather than a side effect.

What Google Actually Says About The Mechanics

Nobody in that top ten quotes the documentation, so here it is on the three points that come up most.

On word count, from Google's guidance on creating helpful content, loaded 28 July 2026. "Are you writing to a particular word count because you've heard or read that Google has a preferred word count? (No, we don't.)" The parenthesis is theirs. Every article that tells you to hit 1,500 words is telling you something the platform has explicitly denied.

On keywords, from Google's public explanation of ranking. "The most basic signal that information is relevant is when content contains the same keywords as your search query." Basic is doing work in that sentence. Using the words people search is necessary and it's the floor, not the strategy.

On what separates a page that ranks from one that doesn't, the same helpful content page asks a self-assessment question I keep taped to the inside of my head. "Does the content provide original information, reporting, research, or analysis?" Original information is the whole game. It's also the single hardest thing to produce, which is why the advice ecosystem quietly routes around it toward heading structure.

One more, since it reframes the discipline. "SEO can be a helpful activity when it is applied to people-first content, rather than search engine-first content." That's a statement about sequence. Write the thing, then optimise it, not the other way round.

The Ageing Statistic Problem

Something I noticed while reading the competition, and it's a habit worth breaking in your own writing.

The AWAI guide carries a set of numbers about smartphone users checking their devices 63 times per day, average screen time above four and a half hours, and search moments that "often last only 70 seconds", referencing Google's micro-moments framework from 2015. The Medium post cites more than half of clicks going to the top three positions, attributed to Advanced Web Ranking, and a Nielsen Norman Group figure of 20 to 28 percent of a page's words actually being read.

None of those are invented and I'm not accusing anyone of anything. The issue is that as a reader you cannot tell how old any of them are, because the year isn't attached, and a behavioural statistic about phone use from a decade ago is being deployed in an article about how to write this year. Numbers decay. Search behaviour in particular has been rearranged more than once since 2015, and a figure that was carefully measured then can quietly become decoration now.

The rule I use is dull and it works. Every number in a draft carries the source and the date I loaded it, in a scratch file, before drafting starts. If I can't produce both, the number doesn't go in the article, and if the source won't load I write that instead of quoting from memory. I've published pieces with visible holes where a figure should be because a page returned a 403 to me, and I'd rather do that than pass along something I half remember.

There's a second benefit that took me longer to notice. Writing with dated primary sources produces specifics, and specifics are most of what makes a page worth reading and most of what makes it hard to copy. The recycled-statistic version of an article is trivially replaceable, because everyone recycling the same three numbers ends up with the same three paragraphs, whereas a number you loaded yourself on a Tuesday and dated is yours in a way that a scraped one isn't.

On-Page, Ranked By How Much It Matters

The tips lists give equal weight to every item. They aren't equal.

Things that genuinely matter. The title, because it's both a relevance signal and the thing a human decides to click. Whether the page answers the query directly and early, since a reader who has to scroll to find out whether you're on-topic often doesn't. Whether the content contains information not available in the other results.

Things that matter a bit. Headings that describe their sections honestly. Internal links to related pages, which help people and help crawling. A URL slug containing the words.

Things that are hygiene. Alt text, which you should write for accessibility reasons that have nothing to do with search. Meta descriptions, which influence clicks rather than rankings and get rewritten by Google anyway. Keyword placement in the first paragraph, which happens naturally if you're actually writing about the topic.

I hold that ordering loosely and I'd expect an experienced technical SEO to argue with the middle group. What I'm confident about is the shape, which is that the top group is worth hours and the bottom group is worth minutes, and the tips lists present them as a flat checklist where each item feels equally like progress.

Writing Against A Page Full Of Vendors

This situation comes up constantly and none of the ranking pages address it, so here's how I handle it.

When the top ten is product pages, you have two options and only one of them works. The losing option is writing a better product page. You don't have the domain strength, and even if you did, the searcher already has nine to choose from.

The winning option is answering the question those pages structurally cannot. A tool page can't tell you when not to use the tool. It can't publish the limits. It can't compare itself honestly to the free alternative or admit the job is better done another way. Every one of those is a real question that real searchers have, and the incumbents are disqualified from answering them by what they are.

That's not a clever trick, it's just the format gap applied. It also means the resulting article tends to be genuinely more useful, which is a rare case of the incentive pointing the right way. For the mechanics of finding these situations at scale, SEO automation covers what I automate and where the API limits bite.

Saying How It Was Made

Short section on something the tips lists skip and Google mentions.

The helpful content guidance says "It's helpful to readers to know how a piece of content was produced", and pairs it with a question about whether authorship is self-evident to visitors. That's a nudge toward disclosure, and it's aimed at exactly the situation most people writing SEO content are now in, which is that a model was involved somewhere.

My own position, held without much confidence that it's the universally right one, is that the honest thing to disclose is the process rather than the tool. Where the numbers came from, when they were loaded, what I couldn't verify. Naming which model helped with a paragraph tells a reader nothing useful. Telling them a source returned a 403 and so there's a gap where a figure should be tells them quite a lot. The order-of-operations side of working with a model is in how AI fits into a writing process.

What This Doesn't Fix

Ending with the limit, because I've watched myself confuse these two things for longer than I'd like to admit.

Choosing well and writing well gets you a page that deserves to rank. It does not get you a page that ranks, and the gap between those is filled by things that happen off your site over months, which is a completely separate discipline covered in getting a site to the top of Google. I've had a site sit at an average position of 47 with pages I still think were good. Good wasn't the constraint.

What the process above does buy you is that when a site does become visible, it becomes visible for something people actually search for, rather than for a phrase you liked the sound of. That's worth real money and it's not the same as traffic. For what the surrounding tooling costs, the writing tools breakdown has the arithmetic, and it's less than people assume.