How to Write a Newsletter With AI the Right Way

The pushback against AI writing is earned. Most of it deserves the criticism, because most of it is a one-shot prompt: type a topic, accept whatever comes back, hit publish, repeat.

Readers can feel it, too. In business writing especially, they can't tell whether an idea is your hard-won insight or something a model invented on the spot, so they stop trusting all of it.

But the critics are describing a process, and a process can be fixed. I write newsletters with AI every week, for my own brands and for clients, and the method looks nothing like one-shotting a draft.

This is how to write a newsletter with AI the right way: an AI that knows your brand, your real experiences as the input, a negotiated outline in the middle, and your judgment at every checkpoint that matters.

Why Readers Distrust AI Writing

Every reader of business writing runs a quiet test. Did this person actually live this, or is the story machine-made filler?

A personal anecdote is a fact. It carries the same weight as a statistic, and a fabricated one destroys your credibility the moment a reader recognizes it. That's the deeper problem with one-shot AI writing. The model has no access to anything you actually lived, so it fills the gaps with plausible-sounding invention, and your reader is left doing forensics on every paragraph.

Trust is the whole asset. You publish a newsletter to build authority with an audience, and authority compounds only when readers believe the insight is yours.

So the fix starts before the writing. Give the AI something real to work with and the trust problem mostly disappears.

The Pilot Principle

The mental model I use is an airline pilot. The two moments that matter most in any flight are takeoff and landing, and that's where the pilot is fully engaged. The long middle runs mostly on autopilot while the pilot monitors.

Writing with AI works the same way. Be most engaged in the first 10 to 20 percent, where your experiences and ideas become the input and the outline. Be fully engaged again in the last 10 to 20 percent, where you edit the draft and sign your name to it. Let the AI run the middle, the structuring and drafting that used to eat your whole week.

You're still the pilot. You just stopped hand-flying the boring part.

How to Write a Newsletter With AI, Step by Step

This is the exact process I use with the select clients in Flyletter's premium service arm, and it works the same whether your AI is Flyletter, Claude, or ChatGPT.

1. Set Up a Brand Knowledge Base

Before any interview or dictation, the AI needs context it can reference on every piece: who you are, who you write for, what you stand for, and how your writing actually sounds. Without that base, even your best raw material comes back in someone else's voice.

This is the preflight work, done before anything takes off. Collect your positioning, your audience, and a few real writing samples in a place the AI reads before every draft. In Flyletter, this is the brand profile, built from your writing samples and referenced by every agent on every job.

2. Interview Yourself and Record It

The input itself is raw material only you have. With clients, I start by interviewing them about their knowledge, their experiences, and their expertise in their specific domain. I record all of it.

Do the same thing solo. Record voice notes while the thinking is fresh: the client call that went sideways, the framework you keep explaining, the opinion you keep defending at dinner. Rambling is fine. You're capturing insight, not performing it.

In Flyletter, you save those thoughts as ideas, so every half-formed insight sits queued as a future newsletter instead of dying in your notes app.

You can also run the interview in reverse and let the AI ask the questions. Connect Claude to Flyletter and it already knows your brand and your audience, so it can interview you about what's going on in your life and business, chase the interesting threads, and break the conversation into separate topics. More on that below.

3. Turn the Raw Input Into Topic Ideas

Feed the recording or transcript to the brand-aware AI you set up in step one and ask what's worth publishing. Because the input is your lived experience, every topic that comes back is grounded in something true, which beats brainstorming generic listicle angles every time. One good interview usually breaks into several topics, each one a future edition.

This is the first checkpoint. You pick which flight to take.

4. Negotiate a Beat-by-Beat Outline

Don't ask for a draft yet. Ask for an outline first, beat by beat, and then argue with it.

I go back and forth until three things are true: the flow makes sense, the points land in the right order, and my first-person perspective from the recording runs through the piece instead of sitting bolted on as a quote. Sometimes that takes two rounds. Sometimes it takes six.

Most of the piece's quality gets decided at this stage, and it costs minutes rather than hours. A wrong outline produces a wrong draft that no amount of editing can save.

In Flyletter, this stage is the content brief. You work it over with a thought-partner agent until you agree, and then the newsletter pipeline carries the brief through strategy, research, outline, and draft with a specialist agent on each job.

5. Let the AI Write the Draft

Now the autopilot earns its keep. With real input and an agreed outline, the draft that comes back is your argument, in your order, told from your perspective, because everything it needed came from you.

This is the middle 80 percent of the flight. Let it run. If the AI knows your voice well enough, the draft arrives sounding like you instead of like a press release.

6. Edit by Hand and by Conversation

The landing is yours, so get back on the controls. I edit every draft two ways: hand edits for the sentences only I would write, and conversational edits where I tell the AI what's wrong ("the second section buries the point, lead with the story") and let it rework the passage.

The draft is done when you would defend every line as your own. That bar doesn't move, because your name is on the piece.

In Flyletter, the refine agent sits next to your draft for exactly this back-and-forth, and the editor takes hand edits directly.

7. Create the Image From the Finished Piece

The featured image comes last, generated from the final article instead of the topic. A finished draft carries the piece's actual argument and mood, which produces an image that matches what you wrote rather than a generic stock-style header.

Flyletter's image agent reads the completed draft and generates a featured image matched to your brand's visual style.

Run the Whole Flow in Claude if You Want

Everything above runs end to end inside Flyletter. It also runs inside Claude, because the Flyletter MCP connector gives Claude access to your brand profile, your saved ideas, and your newsletters. Save an idea mid-conversation, develop the brief, and generate a draft in your voice without leaving the chat.

The reverse interview from step two lives here too. Ask Claude to interview you, answer its questions out loud or in text, and save each topic it surfaces straight to Flyletter as an idea. From there you can develop them one at a time, or put them on an auto-write schedule and let finished drafts arrive on their own.

Why This Is Worth Doing at All

Most people with real expertise never publish it. I see three reasons: they think they're bad at writing, they hate writing, or they can't find the time to catalog what they know.

Used this way, AI removes all three excuses without removing you. You talk, you judge, you edit. The writing labor in the middle stops being the bottleneck, so your ideas ship at the rate you actually have them, and every published piece compounds your authority with the exact audience you want to reach.

I genuinely believe the world gets better when more people share what they've uniquely learned. The backlash against AI writing is aimed at people who use it to say nothing, faster. Use it to say something real, more often.

Bottom Line

Written the wrong way, AI produces content nobody trusts and nobody needed. Written the right way, the input is your recorded experience, the outline is negotiated until it's yours, and the final edit happens under your hand.

Stay on the controls for takeoff and landing. Let the AI fly the middle. Your readers get more of your actual thinking, and you get hours of your week back.

Start this week. Record one voice note about the last lesson your work taught you, and take it through the seven steps.