Diagnosis · blog post / newsletter
Why your blog post sounds like AI
Most advice about AI-sounding blog posts targets vocabulary. The tell that actually survives a vocabulary edit is the template underneath it.
A blog post reads as AI-written less because of word choice and more because of shape: a numbered-promise intro, three or five bold sub-headed sections, and a close that restates the intro. That scaffold comes from a decade of SEO content-mill writing the model trained on. Swapping an adjective leaves the scaffold standing; the fix is structural.
Why this happens to blog post / newsletter in particular
Blog posts and newsletters inherited their shape from a decade of SEO content writing that existed before any model did: an intro that names what the post will cover, three to seven sections each promising a discrete takeaway, bold or numbered sub-heads breaking up the wall of text, and a closing paragraph that restates the opening. Search-engine advice taught human writers to build posts this way because it was believed to rank and to hold attention on a page built for skimming. A model trained on that corpus learned the shape as the default answer to a request for a blog post, and reproduces it whether or not the topic calls for structure. A single anecdote gets carved into three numbered lessons because that is the container the model reaches for, not because the anecdote has three parts.
The tells that show up here
Not every AI tell appears everywhere. These are the ones that cluster on this surface, and what it is about the format that produces them.
- bold-header-list
- A blog draft turns a paragraph of real information into bold-labeled fragments, because a how-to prompt reads as a request for an enumerated structure rather than continuous prose.
- rule-of-three-list
- A numbered-promise intro defaults to three items almost without exception, not because the topic has three parts but because three is the model's default count for an unconstrained list; the checklist logs the pattern but does not score it, since a genuinely three-part topic trips it too.
- title-case-heading
- Sub-heads arrive capitalized word by word, a habit inherited from the SEO-listicle era rather than a model tic on its own, which is why it is tracked but not scored.
- throat-clearing
- A hook question gets followed by a pivot phrase that promises a coming insight before delivering an ordinary observation, filling space where a real anecdote would otherwise have to carry the opening.
Before and after
A generic AI draft on the left. On the right, the real output of running that draft through underslop, checked so it adds no name or number the draft did not already carry. The voice is The Overthinker, digressive and first-person
Ever wondered how to stay focused working from home? You're not alone! In this comprehensive guide, we'll dive deep into everything you need to know. - **Workspace:** Set up a dedicated spot away from distractions. - **Routine:** Establish consistent start and stop times. - **Boundaries:** Communicate clearly with family or roommates. By the end, you'll not only feel more productive but also reclaim your day. Let's get started on this journey toward a better home office!
I wonder if there's anyone who hasn't, at least once, stared at their own kitchen table and thought, *this is meant to be my office now*? Perhaps not a question worth asking aloud, it seems we've all been there, more or less. A few small things, I think, make the difference between a day that dissolves and one that holds its shape. A corner that belongs to work and nothing else, even if the corner is just a desk that faces the wall instead of the bed. A start time that resembles a start time, and an end time that one actually keeps, more or less. And the hard part, telling the people you live with, gently, that you are not available, even though you are right there in the next room, looking slightly pathetic. I daresay that is most of it. Not a grand transformation, not a productivity revolution, just a few small agreements with yourself, and the nerve to keep them. And perhaps, at the end of the day, you find you have reclaimed something after all.
How to fix it yourself
- Cut the numbered promise from the intro Delete any opening line that tells the reader how many points are coming or that promises an exhaustive guide. State the one true thing the post is about instead.
- Turn bold sub-heads back into paragraphs If two sub-points are really one idea explained twice, merge them into a paragraph with a real transition. Keep a heading only where a reader genuinely needs to skip ahead to a different, independent section.
- Replace the third generic point with a specific one The third item in a three-item list is usually the weakest, added to round the count out. Cut it, or replace it with the one detail from your own experience that a generic version of the post could not contain.
- Delete the closing paragraph that restates the intro End on the last real point instead of a summary. A reader who made it that far does not need the post repeated back to them.
Questions
How to tell if a blog post is written by AI?
Look at the shape before the sentences: a numbered-promise intro, three or five bold sub-headed sections, and a closing paragraph that restates the opening. That scaffold survives even after someone edits the individual words, because the structure is the tell, not any single phrase.
How to make content not sound like AI?
Remove the scaffold, not just the adjectives. Cut the numbered promise, merge sub-heads that are really one idea, and end on the last real point instead of a summary. What is left should read like one person's continuous thought, not an outline filled in.
Why does everything sound like ChatGPT now?
Because the SEO content-mill template many models trained on was already the dominant blog format before those models existed, built by human writers optimizing for search and skim-reading. A model learns that template as the default shape for a blog post, so it reproduces the same structure millions of pre-AI posts already used.
Will editing the structure hurt my SEO?
Search engines rank information and relevance, not heading count. A post with fewer, better headings that a reader actually finishes tends to hold attention longer than one chopped into filler sub-heads, which is a stronger signal than keeping a template intact.
Does this mean I should never use headings?
No. Use a heading where a reader genuinely needs to jump to an independent section. The tell is not headings themselves, it is headings added to hit a count, breaking one continuous idea into pieces it never needed.
This page explains the cause. underslop is the edit: paste up to 500 words on the free tier, and it marks the tells it could not remove. Your text is held for 0 seconds.