How to Write a Newsletter with AI in 2026

Most newsletter writers use AI wrong. They open ChatGPT, brainstorm a topic, maybe generate an outline, then close the tab and write the entire thing themselves. That is like buying a kitchen full of appliances and only using the toaster.

AI can help at every stage of the newsletter workflow except the final one (hitting publish). Brainstorming, research, drafting, editing, formatting, exporting. But most writers only touch the first stage. They leave 90% of the value on the table.

This guide is for newsletter writers on Substack, Ghost, Beehiiv, ConvertKit, or WordPress. It covers the full workflow from topic to published post. It focuses on tools and techniques that exist right now, not theoretical futures.

Every newsletter follows roughly the same production cycle, whether you publish weekly or daily:

Pick a topic. Decide what to write about this week.
Research. Gather facts, quotes, data, and references.
Draft. Write the first version.
Edit. Tighten, cut, restructure, clarify.
Format. Add headers, links, images, pull quotes.
Publish. Export to your newsletter platform and send.

AI can meaningfully improve steps 1 through 5. It cannot publish for you (and shouldn't). But the way most writers use it, they only get help with step 1. That means they are doing the hard parts - drafting and editing - entirely unassisted.

The writers who produce the best AI-assisted newsletters in 2026 have a system for each stage. Here is what that system looks like.

Stage 1: Research (Not Brainstorming)

Skip the brainstorming step. If you write a newsletter, you already know what you want to write about. Your problem is not ideas. Your problem is sitting down to research and draft.

For research, two tools stand out: Perplexity and NotebookLM. They serve different purposes.

Perplexity is a search engine that cites sources. Ask it a question, and it returns a summary with links to every claim. This is useful when you need to verify facts, find recent statistics, or survey what has been written about a topic. Unlike ChatGPT, it does not hallucinate sources. Every citation links to a real page.

NotebookLM is for deep research on specific sources. Upload PDFs, articles, or transcripts. It lets you ask questions about them and generates summaries with citations back to the exact passages. This is invaluable when your newsletter references specific reports, books, or interviews.

The workflow: spend 20-30 minutes in Perplexity or NotebookLM gathering your material. Copy the relevant quotes and data points into a document. Now you have a research file. This is your raw material for the draft.

Do not ask ChatGPT to "research" your topic. It will give you plausible-sounding paragraphs with no verifiable sources. For newsletter writers who need to be credible, that is worse than useless.

Stage 2: Draft in a Markdown Editor

Here is where most newsletter writers make the critical mistake. They draft directly in Substack, Ghost, or their email platform. These editors are built for publishing, not writing. They lack focus mode, distraction-free layouts, and keyboard shortcuts for restructuring content.

Worse, none of them have AI editing built in. If you want AI help while drafting in Substack, you are back to the copy-paste workflow: select a paragraph, open ChatGPT, paste, prompt, copy the result, paste it back. Every time you do this, you lose formatting and context. As we covered in The Copy-Paste Era of AI Writing is Over, that workflow is broken.

The better approach: draft in a markdown editor that has AI built in. Markdown is the ideal format for newsletter writing because it is portable. It converts cleanly to HTML, which is what every newsletter platform accepts under the hood.

Athens is built for exactly this workflow. It is a markdown WYSIWYG editor, which means you see formatted text (headers, bold, lists, links) while writing, but the underlying format is markdown. You get the best of both worlds: a clean writing experience and a portable output format.

During the drafting stage, do not worry about quality. Write fast. Get the ideas down. You will fix everything in the next stage. The goal is to go from research notes to a complete first draft as quickly as possible.

Stage 3: AI Editing with Inline Diffs

This is the stage where AI adds the most value for newsletter writers. Not generation. Editing.

The difference matters. When you ask AI to generate a newsletter, it writes in a generic voice that your readers will recognize as machine-made. When you ask AI to edit your draft, it tightens your prose while keeping your voice intact.

In Athens, AI editing works through inline diffs. You select a passage - a sentence, a paragraph, or an entire section. You tell the AI what to do: "tighten this," "make this clearer," "cut 30%," "simplify for a general audience." The AI returns edits displayed as a diff directly in your document. Deletions in red. Additions in green. You see exactly what changed and accept or reject each edit.

This is fundamentally different from the ChatGPT workflow in three ways:

You see every change. No more guessing what the AI rewrote. Every word it touched is highlighted.
You keep control. Reject the changes you disagree with. Accept the ones you like. Your judgment stays in the loop.
No formatting breaks. The edits happen inside your document. Nothing gets copied or pasted. Your headers, links, and lists stay intact.

For newsletter writers specifically, the most useful AI edits fall into a few categories:

Tightening. Newsletter readers are busy. Every sentence needs to earn its place. AI is excellent at cutting filler words, redundant phrases, and unnecessary qualifiers. A 1,500-word draft becomes a sharp 1,100-word post.
Clarity. You know your topic deeply. Your readers might not. AI can flag jargon, simplify complex explanations, and suggest concrete examples where your writing is abstract.
Transitions. Moving between sections is hard. AI can smooth the bridges between ideas so your newsletter flows instead of jumping.
Openings. The first two sentences of a newsletter determine whether people keep reading. AI can generate five alternatives for your opening paragraph. Pick the one that hooks best.

One round of AI editing typically takes five minutes. Compare that to 30-45 minutes of self-editing to achieve the same result. You still do the creative work. You still decide what stays and what goes. But the AI accelerates the mechanical part of revision.

Stage 4: The Substack Formatting Problem

Here is a problem every Substack writer has encountered. You write your post in another tool - Google Docs, Notion, Bear, whatever. You copy it into Substack. And the formatting breaks.

Headers lose their levels. Lists collapse. Links disappear or get mangled. Bold and italic formatting sometimes transfers, sometimes doesn't. Block quotes turn into regular paragraphs. You spend 15 minutes reformatting a post you already formatted once.

This happens because Substack's editor does not support markdown. When you paste rich text, it goes through a conversion layer that strips some formatting and corrupts other parts. The same problem exists with Ghost, WordPress (in its block editor), and most email platforms.

People have tried workarounds. Pandoc to convert markdown to HTML. Custom scripts. Google Docs as an intermediary (because Substack handles Google Docs paste better than most). None of these are reliable, and all of them add friction to the publishing step.

The actual solution is to export clean HTML. Every newsletter platform accepts HTML, even if they do not advertise it. Substack has an "import post" feature that accepts HTML and preserves formatting correctly. Ghost has a native HTML card. WordPress has always supported HTML in its code editor.

Athens exports to HTML directly from markdown. Because the underlying format is structured markdown, the HTML output is clean and semantic. Headers are <h2> and <h3> tags, not styled <div> elements. Lists are proper <ul> and <ol> elements. Links preserve their href attributes. No inline styles. No garbage markup.

The workflow: finish editing in Athens, export to HTML, paste into your platform's HTML editor. Total time: 30 seconds. No reformatting. No broken links. No lost headers.

For more detail on the formatting problem and how to solve it, see How to Get ChatGPT Output into Substack Without Breaking Formatting.

The Full Workflow, Start to Finish

Here is the complete process, with approximate times for a 1,200-word newsletter:

Research in Perplexity or NotebookLM (20-30 minutes). Gather facts, quotes, and data points. Save them to a research doc.
Draft in Athens (30-45 minutes). Write fast. Do not edit while drafting. Get the full post down. Use your research notes as scaffolding.
AI edit in Athens (5-10 minutes). Run 2-3 rounds of AI editing. Tighten the prose. Clarify the weak sections. Sharpen the opening. Accept the good changes, reject the bad ones.
Final read-through (10-15 minutes). Read the post once from top to bottom. Fix anything that feels off. This is your voice check. The post should sound like you, not like a machine.
Export and publish (2 minutes). Export to HTML from Athens. Paste into Substack, Ghost, or WordPress. Add your subject line and preview text. Send.

Total time: roughly 70-100 minutes. Compare that to the typical workflow without AI editing, which takes 2-3 hours for the same quality. The time savings come almost entirely from the editing stage, where AI does in minutes what would take you half an hour.

A good newsletter has a point of view. It has opinions that come from experience. It has anecdotes and observations that only the author could provide. AI cannot generate any of this.

AI cannot tell you what to write about. It does not know what your readers care about this week. It does not know what happened at the conference you attended, or what trend you spotted in your industry, or what book changed how you think about your field.

AI cannot replace reporting. If your newsletter involves interviews, original data, or first-hand accounts, no AI tool can substitute for doing the work. It can help you write up what you found, but it cannot find it for you.

AI also cannot maintain your voice over time without your active involvement. If you blindly accept every AI suggestion, your writing will drift toward a generic mean. The writers who use AI editing well are the ones who reject 30-40% of suggestions. They use AI as a sparring partner, not a ghostwriter.

The best newsletters in 2026 are not written by AI. They are written by humans who use AI to write faster and edit sharper. The difference is obvious to readers.

Tools for Each Stage

Here is a quick reference for the tools mentioned in this guide, plus alternatives:

Research: Perplexity (for sourced search), NotebookLM (for analyzing specific documents), Elicit (for academic papers).
Drafting and editing: Athens (markdown WYSIWYG with AI inline diffs), iA Writer (clean markdown editor, no AI), Ulysses (polished writing app, no AI editing).
Grammar and style: Grammarly (real-time grammar), Hemingway Editor (readability), LanguageTool (multilingual).
Publishing: Substack (for independent newsletters), Ghost (for self-hosted newsletters), Beehiiv (for growth-focused newsletters), WordPress (for blog-newsletters).

For a broader comparison of AI writing tools suited to bloggers and newsletter writers, see Best AI Writing Tools for Bloggers and Newsletter Writers.

Markdown is the hidden advantage in this workflow. It is what makes the export step painless instead of painful.

When you write in Google Docs and paste into Substack, the rich text carries hidden formatting. Google Docs exports styled HTML with inline CSS, font declarations, and nested span elements. Substack strips most of it, sometimes incorrectly.

When you write in markdown and export to HTML, the output is clean. A header is a header. A list is a list. There is no hidden formatting to strip or corrupt.

Markdown is also version-control friendly. You can track changes over time, diff your drafts, and maintain a clean archive of every newsletter you have published. Try doing that with Google Docs or Substack's editor.

For newsletter writers who are curious about why markdown is gaining traction with AI tools specifically, we wrote about this in detail: Why Markdown is AI-Native.

Getting Started

You do not need to change your entire workflow at once. Start with one stage.

If you currently draft in Substack, try drafting your next newsletter in Athens instead. Use the AI editor to tighten one section. Export the result to HTML and paste it into Substack. See if the quality is better and the formatting holds.

If you currently use ChatGPT for brainstorming, try using Perplexity for research instead. Notice how having sourced references changes the quality of your draft.

If you currently self-edit for an hour per post, try running one round of AI editing first. Accept the suggestions that improve your writing. Reject the rest. See if you finish faster.

The newsletter writers who produce consistently great work in 2026 are not the ones with the best AI tools. They are the ones with the best process. AI just makes that process faster.

The Newsletter Writer's Workflow