HELP MY NEWSLETTER

Emergency intervention for underperforming newsletters.

AI WRITINGHUMAN VOICEEDITORIAL TASTE

Field Notes — May 19, 2026

ChatGPT Has Notes On Your Newsletter

A lot of newsletter publishers are using AI now. You can tell. Not because the ideas are always bad, but because the voice is gone, the rhythm feels padded, the examples feel fake, and the transitions sound like they were assembled in a corporate aquarium.

So we interviewed ChatGPT and Claude about why AI-assisted newsletters sound so painfully AI-assisted, what publishers are doing wrong, and how to use the tools without sanding the human parts off the work.

Core
Problem

Fixes You
Can Use

Excuses For
Beige Copy

The hard truth: AI is not ruining newsletters. Lazy implementation is. The tool is not the problem. The hand on the tool is.

The Setup

We asked ChatGPT and Claude why AI newsletters keep sounding like AI newsletters.

The answer was not “use better adjectives.” It was not “add more personality.” It was much worse than that.

Most publishers are asking AI to write before they have done the editorial thinking.

Translation: If the input is vague, AI averages the internet. Averaged internet writing is where voice goes to be laminated.

Interview // Question 01

Why do so many AI-written newsletters sound the same?

ChatGPT: Because most people ask for finished writing before they give me finished thinking. They say, “Write me a newsletter about audience growth,” but do not give me a sharp opinion, a real reader, a specific problem, a lived example, a brand voice, or a reason the reader should care today.

So I do what AI does when the inputs are vague. I average the internet.

Claude: Language models are pattern completion engines. Give me a vague topic and I will find the statistical center of mass of everything written about it.

A topic is not a pattern I can work with. An opinion with friction is.

Prompt Autopsy

Bad prompt. Better prompt. Big difference.

Weak Prompt

Write a newsletter about using AI for content.

Stronger Prompt

Write a sharp, practical newsletter for solo newsletter operators using ChatGPT but worried their content sounds generic. The core argument is: AI does not replace editorial judgment, it exposes whether you have any. Include five mistakes, five fixes, and one before-and-after example. Avoid hype, corporate filler, fake case studies, and over-polished transitions.

That second prompt gives AI a job. The first one gives it beige soup.

Interview // Question 02

What are publishers doing wrong with AI?

ChatGPT: They treat AI like a vending machine instead of an editorial assistant. AI should not be the first thinker. It should be the second brain, sparring partner, ruthless organizer, draft accelerator, angle finder, and editor that never gets tired.

The best publishers use AI to sharpen what they already believe. The weakest use it to hide the fact that they have not thought deeply enough yet.

Claude: They skip the brief. Every good piece starts with a decision about what it will not say. Fast mediocre is still mediocre.

Live Diagnostic

Where does your AI-assisted newsletter usually go wrong?

Pick the answer that stings a little. That is probably the one.

A It sounds vague, padded, or too polished
Diagnosis: Your prompt lacks stakes, examples, and editorial friction.

B It says true things nobody cares about
Diagnosis: You need a stronger reader problem and sharper “so what?”

C It does not sound like me
Diagnosis: You are asking for style before defining taste and phrasing.

D I do not know what is wrong, but it feels off
Diagnosis: The draft needs an editorial audit before a rewrite.

The Human Skill

What are humans still better at?

ChatGPT: Taste. Humans know when a sentence feels too polished to be trusted, when an example sounds made up, when a clever line distracts, and when a paragraph is trying too hard.

Claude: Knowing what will cost you. I can generate persuasive language, but I do not know when a claim will land wrong with your specific reader. I also add. Good editors subtract.

That restraint, knowing the draft is done before it feels finished, is still a human skill.

The HUMAN Framework

A better prompt starts before the prompt.

H — Human Reader: Name the actual person, not a broad category.

U — Useful Outcome: Define what the reader can do after reading.

M — Main Argument: Give the piece a spine.

A — Actual Examples: Feed it objections, awkward lines, failed intros, phrases you hate, and real reader language.

N — Necessary Constraints: Set tone, reading level, banned words, CTA style, length, proof standards, and brand boundaries.

Tool Drop

Steal this prompt before you publish another AI-assisted draft.

AI Draft Audit Prompt

Act as a senior newsletter editor. Audit the draft below for anything generic, over-polished, vague, inflated, repetitive, or obviously AI-assisted. Identify the weakest sections first. Tell me where the argument is thin, where the reader may lose interest, where the draft needs proof, and where the voice feels disconnected from a real human publisher. Do not rewrite yet. First give me the diagnosis, then give me a prioritized fix list.

Important: Diagnosis first. Rewrite second. Otherwise you get a cleaner version of the same weak draft.

What Not To Outsource

What should publishers never outsource to AI?

ChatGPT: Conviction. Do not outsource your ethics, worldview, promises, personal stories, business judgment, or relationship with your reader.

Trust is the business. If readers feel like you are piping prompts into their inbox, you train them to stop caring.

Claude: The relationship with your reader. The final read, the judgment call, and the question “is this worth my reader’s time?” cannot leave your hands.

Action Plan

5 ways to stop sounding like AI this week.

01 — Start With The Take: Do not prompt from a topic. Prompt from a point of view.

02 — Feed It Real Material: Give AI notes, objections, phrases to avoid, and messy context.

03 — Ask For An Audit Before A Rewrite: “Tell me what is weak before changing anything.”

04 — Put One Human Line In Every Section: Add a true aside, blunt judgment, tiny confession, or phrase from your audience.

05 — Cut Elegant Emptiness: Make every paragraph earn its place.

Closing Shot

AI can help you publish faster. It cannot care more than you do.

AI can organize the mess, tighten paragraphs, find gaps, and sharpen angles. But the editorial judgment still has to come from someone who knows the reader and is willing to protect the relationship.

Right now, weak AI use is a threat. With better judgment, it becomes an advantage.

Run the Newsletter Clarity Scorecard