The short version: good blog posts have one argument, one real example, and at least one specific number or step-by-step in the first third of the piece, everything else is decoration. Most posts fail not because the writing is bad but because there’s nothing underneath the writing, no lived detail, no stance, nothing the writer would defend in an argument down the pub. I’ve read a lot of blog posts this year, written by people and by AI, and the good ones are rarer than you’d think.
I read a lot of blog posts. Most of them are fine, which is the problem
Fine is the enemy here. A fine blog post ticks every SEO box, hits the keyword, has an H2 for every subtopic, ends with a call to action, and says nothing that couldn’t have been generated by a competitor’s copywriter on the same day. I know because I’ve written some of them myself, back when I was chasing volume rather than usefulness.
I published 569 blog posts in 45 days as an experiment to see what AI content at scale does to a site’s visibility (I wrote up the full mess of it, indexing numbers included, in this piece). Google indexed 5 percent of them. Not ranked, indexed. The other 95 percent were fine, grammatically correct, keyword-appropriate, and invisible. That’s what fine gets you. It gets you a URL that exists and nothing else.
What good looks like, structurally
Strip away the topic and the tone, and a good blog post has the same skeleton every time. I check for five things when I edit my own drafts or a client’s:
- A single argument you could state in one sentence, not five topics loosely stitched together
- A specific number, example, or named detail inside the first 150 to 200 words, not saved for the end
- At least one section where the writer clearly disagrees with the conventional advice on the topic
- Subheads that promise something rather than just labelling a category (compare “Pricing” to “What this costs, with real numbers” for the reader, though I’d never use that particular word twice in one post)
- An ending that adds a new thought, not a summary of the paragraphs above it
That’s the whole test. No word count target, no keyword density target, none of that. I’ve seen 700-word posts do this brilliantly and 3,000-word posts fail every single one of the five points while still hitting every SEO checklist item a plugin can measure.
The paragraph that decides whether the post is any good
There’s usually one paragraph, somewhere in the middle third, that decides whether a post is worth the reader’s time. It’s the paragraph where the writer stops describing the topic and starts making a claim about it. If you can delete that paragraph and the post reads exactly the same, there isn’t one, and the post is filler with a topic attached to it.
I run this test on my own drafts before I publish. I read the piece and ask myself: where’s the sentence someone would screenshot and disagree with? If I can’t find one, I haven’t finished writing yet, I’ve just finished typing.
A real example: two posts on the same site, six months apart
In early 2025 I wrote a post about link insertions on established blogs, the kind of guest content and niche edit work that agencies buy and sell constantly. My first draft was competent and forgettable. It explained what a niche edit is, listed some pricing ranges, and told people to be careful of spammy sites. Fine. Forgettable.
Before I published it I went back and added the actual detail: how I price link insertions on this site, what I turn down, what a bad pitch email looks like versus a good one, and a real case where a client’s link placement got a page removed from a client’s own site months later because the anchor text was so aggressive it triggered a manual review. That version is the one that’s still on the site, and it’s the one people email me about. You can read the current version of that thinking on the link insertions page if you want to see the difference between “explains a topic” and “shows you the topic from inside a real business.”
Same topic, same writer, six months apart. The only thing that changed was whether I was willing to put my own numbers and my own mistakes on the page instead of general industry advice that any competitor could have written in the same afternoon.
The uncomfortable bit nobody wants to put in writing
Here’s the part most advice on this topic quietly skips over: a good blog post will annoy some percentage of the people who read it, and if yours never does, that’s a signal, not a compliment. Neutral, balanced, “on the one hand, on the other hand” writing gets shared politely and forgotten immediately. Nobody sends a friend a link because it was fair to both sides.
The posts on this site that get the most traffic, the most replies, the most people quoting a line back at me on LinkedIn, are the ones where I said something a bit unpopular. That 569-posts experiment I mentioned above got more engagement than almost anything I’d published that year, and a chunk of the response was people telling me I was wrong to have tried it at all. That’s fine. That’s the point. A post that makes nobody uncomfortable usually hasn’t said anything.
The other thing nobody likes admitting: good blog posts are mostly research and structure, not inspiration. I don’t sit down and feel creative and produce something people love. I sit down with a specific number I need to check, a client example I need permission to use, and an argument I already know I want to make, and then I write it in that order. The writing part is almost mechanical once the material is real. If you’re waiting to feel inspired before you write, you’ll produce a lot of fine posts and very few good ones.
Where AI changed the standard, not lowered it
AI can now write a structurally competent blog post in about ninety seconds, hitting every one of the mechanical checklist items: subheads, keyword, meta description, a tidy conclusion. That’s exactly why the mechanical checklist stopped being the bar. If every competitor’s content can be structurally competent for free, structural competence is worth nothing to a reader or to a search engine trying to work out which page deserves to be cited.
I built an interactive calculator instead of another AI-written blog post for exactly this reason, because the thing that made it worth visiting wasn’t prose anyone could generate, it was a tool that did something a reader couldn’t do for themselves in ninety seconds. The same logic applies to writing: if an AI tool could produce your draft with the topic swapped out and nobody would notice, you haven’t written a good post yet, you’ve written a template with your name on it.
This is also why the “best tools” listicle format is having a strange second life right now, not because listicles are inherently good writing but because AI answer engines are pulling from pages with clear, specific, checkable claims rather than vague ones. I’ve written about how that shift works for anyone publishing lists or guest posts in this piece on getting named in AI answers, and the underlying principle is the same one running through this whole post: specificity is what gets cited, generality is what gets skipped.
A five-minute checklist for your next draft
Before you hit publish, run through this. It takes about five minutes and it’s caught more bad drafts of mine than any editing tool I’ve paid for.
- Can you state the single argument of this post in one sentence? If it takes three sentences, you’ve written two posts badly instead of one post well
- Is there a specific number, name, date, or figure in the first 200 words? Not “many businesses struggle with this,” an actual figure
- Is there one paragraph a reasonable person could disagree with? If not, add one, or admit the post has nothing new to say
- Would this exact post exist if you removed your own experience from it? If yes, it’s generic and replaceable
- Does the last paragraph say something new, or does it just repeat the introduction in different words?
I’d add a sixth for anyone doing this as an actual business rather than a hobby: check whether the post is worth the time it took, because if you’re building a content business, the maths matters more than the muse. I’ve laid out real, current figures on what blogging pays in this piece on what you can realistically earn from blogging, and the short version is that volume of fine posts loses to a smaller number of good ones almost every time, both for traffic and for income.
What this means if you’re the one being paid to write these
If you’re freelancing rather than writing for your own site, this same standard is what separates the writers earning £40 to £150 a post from the ones earning ten times that for something that on paper looks like the same word count. Clients who’ve been burned by fine, forgettable, AI-flavoured content are actively searching for writers who can produce the version with a real example and a real opinion in it, and they’ll pay for it because they’ve already tried the cheap version and watched it sink without a trace. If you’re hunting for that kind of work, I’ve written a longer breakdown of where those better-paying freelance writing jobs sit, because they’re rarely on the job boards everyone else is refreshing.
The market has already sorted itself into two tiers. One tier produces fine content quickly and gets paid accordingly. The other tier produces good content slowly and gets paid, and gets read, and gets cited, for years afterward. The skeleton I’ve laid out above is how you move from the first tier to the second, and it has nothing to do with talent and everything to do with being willing to put your own specific, checkable, occasionally unpopular material on the page instead of a tidy summary of everyone else’s.
The write for us finder on this site runs the searches for you and lists the domains accepting guest posts in your niche.
Frequently asked questions
How long should a good blog post be?
Whatever length it takes to make one argument with real evidence, no more and no less. On this site that averages around 2,000 to 2,500 words for a proper standalone guide, not because of a target but because covering a topic with real examples and numbers takes that long. A 700-word post with one sharp point beats a padded 3,000-word post every time.
What’s the biggest mistake people make when trying to write a good blog post?
Writing to cover a topic instead of writing to make an argument. Covering a topic produces a summary of what everyone else already said. Making an argument produces something worth reading, sharing, or citing, because it takes a position a reader can agree or disagree with.
Can AI-written blog posts be good?
Structurally, yes, an AI can build a competent skeleton in seconds. Substantively, no, not without a human feeding it real examples, real numbers, and a real opinion, because the model has no lived experience of your business, your clients, or your mistakes to draw on. Every good post I’ve named in this piece needed a human’s specific detail sitting inside it.
Does a good blog post need a controversial opinion?
It needs at least one claim a reasonable reader could disagree with, which isn’t the same as being deliberately controversial for clicks. If every sentence in your post is agreeable and safe, you’ve probably written a summary rather than an argument, and summaries get skimmed once and forgotten.
