How to Make Your Work Weirder Than AI

Competent writing used to be enough of a signal. Clean sentences, clear structure, no obvious errors — that combination reliably communicated “a capable professional made this.” That signal has collapsed. AI produces clean sentences and clear structure at scale, without a person behind any of it. The result is a landscape where technical correctness has become the floor, not the differentiator, and writers who optimized for polish over voice are finding that the ceiling is lower than it used to be.

This isn’t an argument for being deliberately bad at writing. It’s an argument for being more specifically yourself in your writing — which is harder than it sounds after years of editing toward generic competence.

What “weirder than AI” actually means

The word “weird” is doing a specific job here. It doesn’t mean random, confusing, or difficult to read. It means: coming from an actual perspective, with actual judgment, that an averaged-across-all-inputs language model couldn’t replicate because no specific person produced it.

AI writing is the statistical center of a very large distribution of human writing. That center is useful. It’s clear, it’s grammatically correct, it covers the expected points in an expected order. But it has no edges. It doesn’t tell you what to reject, what the author actually thinks, where the general advice breaks down in specific conditions. It can’t, because it doesn’t have a perspective — it has a weighted average.

Newsletter writer Cassidy Williams puts it directly: a newsletter that’s “weird enough (while still being valuable)” keeps readers in a way that interchangeable content doesn’t. The weirdness isn’t a liability — it’s the proof that a person with a particular point of view made the thing.

In practical terms, “weirder than AI” means: specific names instead of generic categories, real constraints instead of theoretical ones, actual examples instead of illustrative hypotheticals, and stated preferences instead of balanced non-judgments. It means writing that could only have come from someone who has actually done the thing, talked to someone who has, or has a considered reason to think the conventional wisdom is wrong.

A five-step publishing workflow to add human signal

Step 1: Draft for utility, not for voice. Get the structure right first. If AI helps you build an outline, generate a rough draft, or check that you haven’t missed an obvious section — use it for that. This is the part of writing where AI is genuinely useful and where using it doesn’t cost you anything distinctive. The utility layer (what is this about, what does it cover, in what order) is not where your value lives.

Step 2: Specificity pass. Go through every paragraph and ask: have I said anything here that couldn’t have been said by someone who hadn’t actually done this? Replace generic categories with real tool names. Replace “some users find” with the specific person you know who found it. Replace “this can be challenging” with the specific challenge you ran into, when, and under what conditions. Specificity isn’t just vivid — it’s verifiable. It signals that someone was actually present when the thing happened.

Step 3: Taste pass. This is where you add the things AI structurally cannot: what you actually think, what you’d reject and why, where you disagree with the consensus, what you’d do differently if you were starting over. Don’t balance your opinion with a “but some people prefer X” unless that balance is genuinely useful to the reader. False balance is a form of hedging, and hedging is what the statistical center does. Stating a clear preference — “I wouldn’t use this tool for anything client-facing, here’s why” — is information. Neutral summary is not.

Step 4: Friction pass. Most competent writing answers the question “how does this work for the average case.” Friction-pass writing also answers: who does this not work for, what are the edge cases, what would you not automate, where does the advice break down. This is where you earn credibility with readers who have already tried the obvious thing and found it incomplete. It also happens to be information that AI can’t reliably generate — it requires judgment about failure modes, which requires actually having encountered them.

Step 5: AI-blandness audit. Read through the final draft looking for the specific patterns that signal AI output or over-edited human writing. Stock transitional phrases: “It’s worth noting that,” “In today’s fast-paced world,” “At the end of the day.” False balance constructions: “While X has advantages, Y also has merits.” Generic calls to action: “Consider experimenting with these approaches.” Remove or rewrite everything you’d be faintly embarrassed to have written yourself. If a sentence could appear in any article on the same topic by any author, it’s probably not doing enough work.

Practical applications for WTJ readers

Freelancers: Your proposals, case studies, and client-facing writing are where this matters most. A proposal that says “I’ve worked with three B2B SaaS companies going through ISO 27001 certification and here’s what I learned about their documentation process” is not replicable from a template. A proposal that says “I have extensive experience with compliance-adjacent content” is. The former gets responses; the latter gets filed.

Newsletter writers: The bar for staying subscribed has changed. Readers who get 40 newsletters don’t need another source of correct information — they need a specific person’s perspective on that information. The weekly-roundup format is harder to sustain than it used to be precisely because curation of publicly available links requires the curation to be distinctly yours. The opinion layer, the “here’s what I actually think about this,” is not optional anymore.

Consultants: Thought leadership content (articles, LinkedIn posts, website copy) needs to communicate what you actually think, not what a careful professional in your space would neutrally observe. If your content could have been written by any consultant in your category, it is not differentiating you. The specific failure modes you’ve seen, the advice you specifically disagree with, the thing you tell every new client before they make the mistake you’ve seen repeatedly — that’s the content.

Small teams writing documentation, guides, or client deliverables: The specificity principle applies here too. A guide that says “we recommend Notion for this client because their team is already in Google Workspace and the integration reduces friction at the handoff step” is more useful than “Notion is a flexible tool that many teams find valuable.” Specificity in internal and client-facing writing builds trust faster than thoroughness.

What weirdness doesn’t fix

Being distinctively yourself doesn’t substitute for the utility layer. A newsletter that has strong opinions about obscure topics that the reader doesn’t care about is not more valuable than a generic one — it’s just differently useless. The specificity and judgment have to be pointed at something the reader is actually trying to figure out.

Voice also doesn’t fix structural problems. A piece with a clear perspective but no discernible organization, argument, or useful information isn’t differentiated — it’s just idiosyncratic in a way that doesn’t serve anyone. The workflow above starts with utility (Step 1) for a reason: the structure has to be there before the voice can be layered on top of it.

Finally, this approach requires you to actually have done the thing you’re writing about, or to have spent enough time with people who have that your judgment is genuinely informed. The friction pass and taste pass only work if you have real material to draw on. If you’re writing about tools you haven’t used, processes you haven’t run, or problems you haven’t encountered, no amount of stylistic weirdness will cover the absence of the thing underneath it.

Source: “Being weirder than AI is practicable and valuable,” Buttondown Blog, published June 8, 2026. The core argument about AI producing middle-of-distribution content, the historical note on weird art and distribution gatekeepers, the Cassidy Williams quote on newsletter weirdness, and the concept of re-weirding yourself are drawn from that source. The five-step publishing workflow and practical applications are independent elaborations for a work-tool context.