Substack / Essay Writing Beginner

25 ChatGPT prompts for Substack writers to draft raw, conversational essays

You opened this because you want a Substack that sounds like a person, not a content machine. I’m going to give you 25 ChatGPT prompts for Substack conversational essay work that I’ve stress-tested on real newsletters - plus the 4-part anatomy, a 7-day workflow, and the small handful of mistakes that quietly kill voice. We’ll get into it.

Substack crossed 5 million paid subscriptions in March 2025, according to The Hill and the Substack Wikipedia entry, and the founders raised another $100M at a $1.1B valuation in July 2025 (NYT, via Wikipedia). That’s a lot of writers chasing the same readers, which is exactly why sounding like yourself has become a real economic edge.

Pull quote: “Trust is not about information and its quality alone. It’s about the messenger.” - Jasmine Sun, The independent writer’s advantage in the age of AI, On Substack, June 2, 2026.

Let me show you how to actually use ChatGPT for that, without letting it sand off every weird, recognizable thing about you.

Quick answer (TL;DR)

The 25 prompts below are grouped into five jobs a Substack essay needs done well: voice discovery, opening lines, storytelling, opinion, and revision. Each prompt is multi-line, copy-paste ready, and built for ChatGPT (works just as well in Claude, Gemini, or Mistral). Don’t paste them and ship the output - paste them, then rewrite the output in your own words. That’s the whole game. The platform’s own June 2, 2026 essay on the independent writer’s advantage in the age of AI makes the same point: AI can summarize, but it can’t see stuff, can’t feel stuff, can’t break news (Jasmine Sun / On Substack, June 2026).

Why most Substack essays sound like LinkedIn

A “conversational essay” is a first-person piece of writing that sounds like a smart friend talking at a kitchen table, not a brand publishing a press release. The reason so many Substack essays don’t sound conversational isn’t bad writing. It’s that the writer handed the whole job to ChatGPT and shipped whatever came back, complete with the “Let’s dive in” intro and the “In today’s fast-paced world” second paragraph.

Three numbers that explain why this matters in 2026:

5,000,000+ - Substack paid subscriptions as of March 2025 (The Hill)
50,000 - creators earning money on Substack, with 50+ clearing $1M/year, as of October 2025 (Wikipedia / Substack entry)
$1,100,000,000 - Substack’s valuation after its July 2025 raise (NYT, via Wikipedia)

When the top of the platform is rewarding voice - see Scott Galloway’s Prof G Media or Lena Dunham’s Famesick tour - the median essay is going to read more like a press release by default. The only way out is to be obviously, stubbornly a person.

That’s what these 25 prompts are built for.

The 4-part anatomy of a conversational essay

Every conversational essay on Substack, whether it’s by Heather Cox Richardson, Anne Helen Petersen, or Emily Sundberg’s Feed Me, has the same four-part anatomy. The prompts below are organized around these parts.

The open - one or two sentences that drop the reader into a moment, not a topic.
The turn - a small personal story or scene that earns the right to the next idea.
The stake - one clear claim the writer is actually willing to defend.
The handoff - a final paragraph that returns to the reader (“If this resonated, hit reply”) instead of a generic “follow for more.”

A good ChatGPT prompt names which of these four parts it’s working on, gives the model your voice, and gives it a hard constraint (e.g., “do not start with ‘I’”). I’ll show you what that looks like in the next section.

Voice-discovery prompts (Prompts 1–4)

Voice discovery is the work of pinning down what your writing already sounds like so ChatGPT can imitate it without flattening it. Don’t skip this section. The writers who sound like themselves on Substack - Casey Newton’s Platformer, Kevin Kelly, Esther Perel on Where Should We Begin? - do this work in public by reposting their own past sentences.

Prompt 1 - Voice audit from past writing

Purpose: Reverse-engineer the actual voice fingerprint from a body of work you’ve already published.

You're an editor, not a writer. I want a "voice audit" of my writing.

Below are 5–10 paragraphs I've actually published on Substack or in past
emails. Do NOT rewrite them. Instead, read them and tell me:

1. My 5 most consistent sentence rhythms (e.g., "short setup, long
   payoff", "fragments stacked on top of fragments")
2. My 5 most-used sentence openers (literally the first 3–5 words)
3. The 3 emotions I default to (bored, amused, furious, tender, etc.)
4. The 3 emotions I almost never use
5. My 2 signature moves (e.g., "ends paragraphs with a question",
   "uses parenthetical asides to undercut the main point")
6. My 1 "tell" - a habit that, if removed, would make the writing
   sound generic

Output as a short table I can paste at the top of every ChatGPT session
when I'm drafting a new essay. Keep total length under 350 words.

PASTE YOUR 5–10 PARAGRAPHS BELOW THIS LINE:
[ ]

Example output (excerpt):

#	Pattern	Example from your writing
1	Short declarative → long causal sentence	”I quit. The reason was simpler than the LinkedIn post I’d been drafting for three weeks.”
2	Openers starting with “Look,” or “Here’s the thing"	"Here’s the thing about going viral…“
3	Default emotion: dry, slightly tired affection	-
4	Almost never: earnest inspirational tone	-

Pro tip: Don’t pick your 5 best paragraphs. Pick 5 random ones, including ones you’re slightly embarrassed by. The “tell” only shows up in the ones you didn’t curate.

Prompt 2 - Anti-voice sample (what you are NOT)

Purpose: Telling ChatGPT what to avoid is sometimes more powerful than telling it what to do.

Below are 3 short samples of writing I actively want to AVOID sounding
like. They are: (a) generic LinkedIn thought leadership, (b) corporate
"we're excited to announce" copy, (c) AI slop in the style of a
ChatGPT default response.

For each one, give me:
- 5 specific tells that mark it as that style
- A 2-sentence summary of the voice
- A "do not" rule I can add to my future prompts

Keep total length under 250 words. Do not rewrite the samples.

SAMPLE A (LinkedIn-style):
[ ]

SAMPLE B (corporate copy):
[ ]

SAMPLE C (default AI):
[ ]

Pro tip: Save the three “do not” rules as a single line - NEVER: hedge every claim, NEVER: open with a question, NEVER: end with "let me know in the comments" - and paste it at the bottom of every prompt you use. It does more than you’d think.

Prompt 3 - Steal a real writer’s structure, not their voice

Purpose: Borrow how a writer organizes an essay, not how they sound.

I want to draft a Substack essay that uses the same structural
moves as one specific published piece, but with my own voice.

Here's the published piece:
[PASTE 600–1,200 WORDS]

In 250 words or less, give me:
1. The skeleton of the piece (5–8 sections, each labeled with a
   3–7 word description of its job)
2. The exact "turn" the writer makes between sections 2 and 3
3. The sentence that does the heaviest lifting (the thesis in disguise)
4. The closing move (return to opening image? call to action? question
   to the reader?)

Then stop. Do not start writing my essay.

Pro tip: Steal from writers outside your niche. A software engineer can borrow the structure of a Joan Didion essay from Slouching Towards Bethlehem. The structure is the free part; the voice is what you have to earn.

Prompt 4 - Voice snapshot card

Purpose: Generate a one-paragraph “voice card” you can paste at the top of every ChatGPT session.

Based on the 5–10 paragraphs I pasted in earlier, write a single
paragraph (60–90 words) that describes my voice so well that a smart
editor could hand it to a copywriter and the copywriter would write
something I'd actually sign.

The paragraph must include:
- My average sentence length (in words)
- My default emotional register
- One thing I do that almost no other writer does
- One thing I must never do

Do not use the words "distinct", "authentic", "unique", "compelling",
"nuanced", or "vibrant". Do not start with "Your voice is…".

Example voice card (you’ll replace the content):

Mid-30s, slightly tired, very online, allergic to fake enthusiasm. Average sentence is 14 words; the long ones are always the second sentence of a paragraph. I open with a scene and close with a question, never with a call to subscribe. I swear, but only in service of a point, never for flavor. I never use the words “delve”, “landscape”, or “in today’s world”.

Pro tip: Pin that voice card in a notes app. It’s the single most useful artifact in your whole writing process.

First-line & open prompts (Prompts 5–9)

A “first line” is the one sentence that decides whether the reader scrolls or stops. The platforms all know this - Substack’s own product notes keep adding drop caps, templates, and design hooks. None of which matter if the first line is a thesis statement.

Prompt 5 - 10 opening lines in 10 different modes

Purpose: Beat the blank page by giving yourself variety on purpose.

I'm drafting a Substack essay on the topic below. I want 10 candidate
opening lines, each in a different mode. For each line, give me:
- The line itself (max 22 words)
- The mode label (e.g., "scene", "contradiction", "confession",
  "question", "vivid image", "second-person address", "in medias res",
  "blunt claim", "small observation", "callback to a past post")

Constraints:
- None of the 10 lines can start with the same word.
- At least 3 lines must be under 10 words.
- At least 1 line must be a question, but it cannot be a yes/no
  question.
- No line may contain the words "imagine", "picture", "delve",
  "tapestry", "in today's world", or "let's dive in".

Essay topic (1–2 sentences):
[ ]

Example output (excerpt):

Scene (8 words): “The bakery on 4th was out of bread again.”
Contradiction (11 words): “I was the most qualified person in the room, and it didn’t help.”
Confession (9 words): “Here’s what I haven’t told my newsletter yet.”

Pro tip: Read all 10 out loud. The line that makes you slightly uncomfortable is almost always the right one. Comfortable openings are skippable.

Prompt 6 - Open with a sensory detail, not a thesis

Purpose: The fastest way to sound conversational is to open with what you saw, heard, smelled, tasted, or touched.

I want to open my essay with a sensory detail, not an abstraction.

Topic of the essay:
[ ]

My constraints:
- The detail must be something I could only have noticed in person
  (no "the smell of coffee" - too generic)
- It must include at least one of: a number, a color, a sound, a
  texture, a temperature
- It must happen at a specific time ("Tuesday at 7:14 a.m." is
  better than "one morning")
- It must take between 1 and 4 sentences total

Give me 3 versions. For each, give the detail and a short
explanation (1–2 sentences) of how it sets up the essay's actual
argument. Do not write the rest of the essay.

Pro tip: Sensory openings age well. A thesis opening ages like milk - by next year, the news cycle has moved on, but the moment in the bakery still works.

Prompt 7 - Kill your first sentence

Purpose: Sometimes the fastest path to a great opening is deleting the bad one you already wrote.

Here's the first sentence of my draft:
"[PASTE YOUR FIRST SENTENCE]"

Now do the following:
1. Diagnose what's wrong with it in 2 sentences (be specific, not
   polite - I'm a grown-up)
2. Give me 3 replacement first sentences, each from a different
   angle:
   (a) go smaller and more specific
   (b) cut the throat-clear and start one beat later
   (c) invert the implied meaning
3. For each replacement, write 1 sentence explaining the trade-off
   I'm making by choosing it.

Pro tip: If the model can’t find anything wrong with your first sentence, paste the second sentence in too. The real problem often lives one beat later.

Prompt 8 - Drop the reader into a conversation already in progress

Purpose: Open like the reader just walked into your kitchen mid-sentence. This is the conversational essay’s secret weapon.

I want my opening paragraph to feel like the reader walked into a
conversation that was already happening. Do not start with "Imagine"
or "Picture this."

Topic of the essay:
[ ]

Give me 3 openings, each 1–2 sentences, that use one of these
moves:
(a) start with "Look," or "Here's the thing," or "Okay, so-"
(b) start with the second half of a sentence
    (e.g., "-and that's why I deleted the app on a Tuesday.")
(c) start with a half-finished thought the reader has to finish
    themselves (e.g., "The worst part isn't the rejection. It's-")

For each, give a 1-sentence explanation of the conversational
implication.

Pro tip: This style is harder to sustain than it looks. Save it for essays where you have a strong point of view. Neutral essays need neutral openings.

Prompt 9 - Promise the reader a payoff in the first 30 words

Purpose: A conversational essay isn’t a thriller, but it still has a contract with the reader.

Here's my current opening paragraph (up to 120 words):
[PASTE IT]

I want to rewrite the first 30 words so they answer 3 questions
the reader is silently asking in the first 5 seconds:

1. "Is this for me?" (signal who the post is for)
2. "What am I going to get if I keep reading?" (the payoff)
3. "Why now?" (the timeliness or the personal stake)

Constraints:
- Must still match my voice card (paste it below)
- Must be a single paragraph, not multiple
- Must not contain the words "you", "we", or "I" more than twice
  total
- Must not use the phrase "in this post" or "in this essay"

Voice card:
[PASTE YOUR VOICE CARD FROM PROMPT 4]

Pro tip: Test the rewrite by reading only the first 30 words to a friend who’s not in your field. If they can finish the sentence “this post is going to argue that…” in their own words, your opening is doing its job.

Storytelling prompts (Prompts 10–14)

The “story” in a conversational essay isn’t a hero’s journey. It’s a scene with one specific moment in it, told in enough sensory detail that the reader feels like they were there. Stephen King’s On Writing (2000) and Joan Didion’s The White Album (1979) are still the templates; in 2026, the Substack essayists who pull this off are Anne Helen Petersen on Culture Study and Emily Sundberg on Feed Me.

Prompt 10 - Build a scene, not a summary

Purpose: Convert a vague memory into a 3-sentence scene.

Here's a memory I want to use in my essay. Right now it's a
summary, not a scene. Help me turn it into a scene.

Summary version:
[PASTE 1–3 SENTENCES]

Now rewrite it as a scene with these properties:
- A specific time and place (down to the hour if I can)
- At least one sensory detail I can see in my head
- A line of actual or near-actual dialogue (made up is fine, but
  mark it as made up)
- A shift - the moment something changed, even if the change is
  tiny

Maximum 4 sentences. Maximum 80 words.

Do not moralize. Do not explain the symbolism. Just the scene.

Example output:

Before: “I had a hard conversation with my dad about money that year.” After (80 words): “It was 11:40 on a Wednesday in November. He stood at the sink scraping burnt rice into the disposal. I was still in my work coat. ‘So how much do you actually have,’ he said, not looking up. The disposal screamed for two seconds, then went quiet. I told him. He turned the water off and didn’t say anything for a long time.”

Pro tip: The best scenes have one line that does the emotional work. Don’t underline it. Let it sit.

Prompt 11 - Find the small moment, not the big one

Purpose: The big moments are obvious and unwritable. The small ones are gold.

I'm trying to write about [BIG EVENT: e.g., "leaving my marriage",
"getting laid off", "my father dying"].

I'm not ready to write about the event itself. I want to write
about the 30 minutes AROUND the event, not the event.

Give me 5 prompts - each 1 sentence, written in second person - that
I could answer with a small moment from those 30 minutes.

Examples of the kind of question I'm looking for:
- "What did the room smell like?"
- "Which object were you holding?"
- "What did you do in the 4 minutes before the phone call?"

Do not write the essay. Do not write the scene. Just give me 5
questions.

Pro tip: The scene that makes a reader cry is almost never the central event. It’s the thing the writer was doing with their hands during the event. Didion understood this; it is the entire trick of The Year of Magical Thinking (2005).

Prompt 12 - Use a “but” and a “therefore”

Purpose: Story scenes need connective tissue. The cheapest, cleanest version is the famous Pixar “but / therefore” rule - every sentence should connect to the next with “but” (change) or “therefore” (causality), not “and then” (tiredness).

Here's a draft scene of mine. It's 6 sentences:
[PASTE IT]

Rewrite it so that every sentence transition uses "but" (a change,
a reversal, a contradiction) or "therefore" (a consequence), never
"and then" or "also" or "next."

Keep the scene at the same length. Keep all my sensory details.
Do not add new characters or new information. Just fix the
connective tissue.

Pro tip: If you can’t connect two sentences with “but” or “therefore,” one of them doesn’t belong. Cut it.

Prompt 13 - Name the thing the protagonist wanted

Purpose: A scene without desire is just weather. Give the protagonist (almost always you) a clear want.

Here's a scene from my draft:
[PASTE 4–8 SENTENCES]

Now do two things:
1. In 1 sentence, name the specific thing I (the protagonist)
   wanted in that scene. The want must be small enough to fit
   in 10 words, and concrete enough to be visible.
   (Bad: "I wanted to be happy." Good: "I wanted the waiter to
   not recognize me.")
2. In 1 sentence, name the smallest thing that got in the way
   of that want.
3. Rewrite the scene's first sentence to make the want visible
   without stating it.

Pro tip: If you can’t name the want in 10 words, the scene is about more than one thing. Split it into two scenes.

Prompt 14 - Use a callback to close the loop

Purpose: A conversational essay that returns to its opening image is twice as memorable. This prompt forces the callback.

My essay's opening image is:
[PASTE 1–3 SENTENCES FROM THE OPEN]

My essay's current ending is:
[PASTE 1–3 SENTENCES FROM THE CLOSE]

Write 3 alternative closing paragraphs, each 2–4 sentences, that
do one of the following:
(a) return to the same physical object but with a new meaning
(b) return to the same person but with a different look on their
    face
(c) invert the opening image (e.g., the bakery had bread this
    time, but the line was longer)

Constraints:
- Do not use the phrase "and just like that" or "in the end"
- Do not introduce any new information
- The reader should be able to see the connection to the opening
  without me flagging it explicitly

Pro tip: The Austin Kleon move in Show Your Work (2014) is exactly this - open and close on the same small object. The reader’s brain does the work; you don’t have to.

Opinion & point-of-view prompts (Prompts 15–19)

An “opinion” in a conversational essay is not a hot take. It’s a defensible position the writer is willing to be wrong about. The Substack writers who do this well - Matthew Yglesias, Slow Boring; Glendon Swarthout, Yglesias-adjacent; Tressie McMillan Cottom - all share the same move: they tell you what they think, then tell you what would have to be true for them to be wrong.

Prompt 15 - Find the claim hiding in your story

Purpose: Most “opinion” essays are really story-plus-claim. Find the claim.

Here's a draft of mine. It's mostly story, no clear opinion yet:
[PASTE 400–800 WORDS]

In 200 words or less, give me:
1. The 3 implicit claims already hiding in the draft (i.e., the
   things the story assumes are true)
2. The single claim the draft is most actually trying to make,
   even if the writer doesn't realize it
3. The version of that claim that would make the reader's
   smartest friend nod but not yawn (i.e., specific, surprising,
   defensible)

Do not rewrite the draft. Do not add to the draft. Just identify.

Pro tip: If the model gives you a claim that scares you a little, that’s your claim. Polite claims get polite opens; scary claims get replies.

Prompt 16 - Steelman the opposite, then take a side

Purpose: Strong opinions are aware of the strongest version of the other side.

My current claim for the essay is:
[PASTE IT]

In 200 words, write the strongest possible version of the
opposite view. The view must:
- Be something a smart, fair-minded person could actually hold
- Use the best available evidence, not strawmen
- End with a sentence that, if I just quoted it, would make me
  look dishonest if I ignored it

Then, in 50 words, tell me: what would have to be true for them
to be right AND for me to be right? (Often there's a "both/and"
reading that the essay can earn.)

Pro tip: Pasted into the essay itself, the steelman is the single best trust-building move. Skip it, and the comment section will write the steelman for you, badly.

Prompt 17 - Decide what you actually believe

Purpose: Most essay drafts hedge. This prompt forces a position.

Here's the section of my draft where I'm hedging:
[PASTE 2–4 SENTENCES]

I'm hedging because I'm scared of being wrong. Help me.

1. Tell me, in 1 sentence, the most specific thing I'm actually
   afraid of.
2. Give me 3 versions of a clear claim that takes a position.
   Label them: SAFE (most defensible), STRETCH (some readers will
   disagree), BOLD (some readers will unsubscribe).
3. Pick the one I should use, and explain why in 1 sentence
   grounded in my stated audience:
   [DESCRIBE YOUR AUDIENCE IN 1–2 SENTENCES]

Pro tip: The BOLD version is almost always the right one if your audience is small and paying. Substack is a subscription business - your 1,000 true fans will reward honesty far more than the 100,000 lurkers will punish it. (Kevin Kelly’s 1,000 True Fans essay from 2008 is the original math on this.)

Prompt 18 - Use the Jobs to be Done frame to sharpen the take

Purpose: JTBD (Jobs to be Done) is the Clayton Christensen / Strategyn framework that says people don’t buy products, they “hire” them to do a job. The same frame is wildly useful for opinions: the reader isn’t “interested in your topic” - they’re hiring your essay to do a job.

My essay topic:
[ ]

My reader is: [DESCRIBE THEM, 1–2 SENTENCES]

In 150 words or less:
1. Finish this sentence: "When a reader opens this essay, the job
   they are hiring it to do is ________."
2. Finish this sentence: "If the essay does that job well, the
   reader will ________. If it fails, they will ________."
3. Rewrite my opening claim so it explicitly promises the job in
   job 1. Maximum 18 words. Must end with a noun, not an
   adjective.

Pro tip: This prompt is the cheapest way I’ve found to turn a vague essay into one that gets forwarded. People forward essays that did a specific job for them, not essays that “explored a topic.”

Prompt 19 - Find the contrarian angle nobody else has

Purpose: A conversational essay on Substack lives or dies by whether the reader feels like they’re getting the actual take, not the consensus take. This prompt hunts for the angle that isn’t already a thousand other posts.

The topic of my essay:
[ ]

The conventional wisdom on this topic (1–2 sentences, as if I'm
explaining it to a friend at a bar):
[ ]

Now give me:
1. Three "contrarian-adjacent" angles - i.e., angles that aren't
   contrarian for sport, but that quietly disagree with the
   conventional wisdom in a way a thoughtful person could
   respect
2. For each, name the piece of evidence I would need to have to
   defend the angle honestly
3. For each, name the risk of writing it (who will I lose, who
   will I gain)

Pro tip: If you can’t find evidence for the contrarian angle, it’s not contrarian - it’s just contrarian-shaped. Skip it.

Revision prompts (Prompts 20–25)

Revision is the part writers skip. It’s also the part readers can’t fake. Substack’s Once and Future Media Forum (May 2026) and the April 2026 product post on post templates both emphasize the design layer; the writing layer is still on you. These six prompts handle the boring-but-load-bearing work.

Prompt 20 - Cut 20% without losing the soul

Purpose: Almost every draft is too long. Cut without gutting.

Here's my full essay draft (target: 1,200 words; current: 1,500–1,800):
[PASTE IT]

Your job: cut it to exactly 1,200 words without losing the soul.

Rules:
- You may cut sentences, phrases, or whole paragraphs
- You may NOT add new ideas, examples, or transitions
- You may NOT change the opening or closing sentence
- Every remaining sentence must still match my voice card:
  [PASTE VOICE CARD]

Output:
- The revised essay (1,200 words)
- A 100-word "what I cut and why" note at the end

Pro tip: Run this prompt twice. The first cut is mechanical; the second cut is where the essay actually tightens. Stephen King in On Writing calls this “the second draft = the first draft minus 10%.” I think he’s underselling it.

Prompt 21 - Find the paragraphs doing nothing

Purpose: Most drafts have 2–3 “filler” paragraphs that exist to bridge two ideas. Cut them.

Here's my draft:
[PASTE IT]

For each paragraph, write one of these labels:
- LOAD-BEARING: removing it would collapse the argument
- TRANSITIONAL: it's bridging two load-bearing paragraphs
- ATMOSPHERIC: it's setting a mood but not making a claim
- DEAD WEIGHT: removing it would change nothing

Then list every paragraph labeled ATMOSPHERIC or DEAD WEIGHT, in
the order they appear, and propose for each:
(a) cut it entirely
(b) merge it into the nearest load-bearing paragraph
(c) compress it to one sentence

Do not rewrite the essay. Just diagnose and propose.

Pro tip: If more than 30% of your paragraphs are atmospheric, the essay is probably a personal essay trying to be a think piece. Pick one and commit.

Prompt 22 - Replace the adverbs with the action

Purpose: Mark Twain’s rule: “Substitute ‘damn’ every time you’re inclined to write ‘very’; your editor will delete it and the writing will be as it should be.” Same idea, less colorful.

Here's my draft:
[PASTE IT]

Find every adverb (words ending in -ly, plus "very", "really",
"quite", "just", "actually", "literally") and for each one:
1. Quote the sentence
2. Explain what the adverb is doing (intensifying, hedging,
   softening, etc.)
3. Propose one of:
   (a) cut the adverb entirely
   (b) replace the verb with a stronger verb
   (c) rewrite the sentence to make the adverb unnecessary

Do not rewrite the full essay. Just list each instance, in order.

Pro tip: I aim for zero adverbs in the final version. It’s a useful constraint because it forces you to do the work in the verb, which is where it belongs.

Prompt 23 - Read it as a stranger

Purpose: You’ve read your draft 12 times. Find out how it reads on the first read.

Pretend you are a smart reader who has never met me, has never
read my Substack, and is reading this essay for the first time on
their phone at 7 a.m.

[PASTE FULL ESSAY]

Now answer, in numbered list, exactly these questions:
1. In the first 30 words, do I tell you what the essay is for and
   who it's for? Yes / no / kind of. Quote the 30 words.
2. Is there a moment in the first half that made you want to keep
   reading? Quote it.
3. Is there a moment in the second half that almost made you stop?
   Quote it.
4. Did you finish the essay knowing what to do next (reply, share,
   sit with it, etc.)? What action did the essay imply, if any?
5. Of the 5,000+ words published on Substack this week, would you
   have read this one to the end? One-sentence honest answer.

Do not rewrite the essay. Do not be polite. I will be more
disappointed by a polite answer than by a mean one.

Pro tip: This prompt catches the draft’s biggest single problem, which is almost always that the opening doesn’t deliver the essay’s actual promise. If Q1 is “kind of,” the opening is the first thing to fix.

Prompt 24 - Title and subtitle lab

Purpose: A Substack post lives or dies by the subject line. Generate options, then test.

Here's the finished essay (1,200 words):
[PASTE IT]

I want 12 subject-line options, grouped into 4 styles (3 each):
- CURIOSITY: opens a loop the reader wants closed
- SPECIFIC: includes a number, a name, a date, or a place
- STANCE: announces the opinion in the subject line itself
- SCENE: opens with a small concrete image

Constraints for every subject line:
- Maximum 50 characters (so it doesn't get cut off on mobile)
- Must be readable as a sentence fragment on its own
- Must NOT use the words "you", "your", "we", "our", or "I"
- Must NOT end with a question mark (questions underperform on
  Substack open rates; this is a soft claim, not a hard one -
  tell me if you have evidence to the contrary)
- Must NOT contain an emoji

For each, in 1 sentence, predict the click-through rate compared
to my other recent posts, and why.

Pro tip: Use the specific lines for posts that go to free readers, the stance lines for posts that go to paid. Your free list is colder; your paid list is hotter; the line has to do different work.

Prompt 25 - Final pre-publish checklist

Purpose: A 90-second gut check before you hit send.

Here's the final draft of my Substack essay:
[PASTE IT]

Voice card:
[PASTE FROM PROMPT 4]

Answer the following 12 yes/no questions, with a 1-sentence
explanation for any "no":

1. Does the opening land in the first 30 words?
2. Is the opening a scene, a confession, a contradiction, or a
   question - not a thesis?
3. Is there a clear "I" who wants something in the first 300
   words?
4. Is the main claim explicit by the 600-word mark?
5. Have I included at least one sensory detail per 200 words?
6. Have I cut all "I think" / "in my opinion" / "I feel that"
   hedges?
7. Have I avoided the words: delve, tapestry, landscape, in
   today's world, let's dive in, at the end of the day, in this
   post, in this essay, imagine, picture this?
8. Are there 0 adverbs in the last 200 words?
9. Does the closing paragraph return to the opening image or
   state the takeaway in plain language?
10. Is there a clear next action for the reader (reply, share,
    comment, subscribe)?
11. Would the writer I most admire (paste a name) finish this
    essay without rolling their eyes?
12. Am I, right now, slightly nervous to publish this? (If the
    answer is "no", the essay is probably too safe.)

Do not rewrite the essay. Just answer the 12.

Pro tip: If Q12 is “no,” the essay is too safe. If Q12 is “yes, and I can defend every claim,” publish it. If Q12 is “yes, and I can’t defend a single claim,” pull it back to draft.

Comparison table: prompt categories vs. essay section vs. expected output

Prompt #	Category	Essay section it powers	Expected output	When to run it
1	Voice discovery	Whole essay (audit)	5-pattern table	Once, then refresh quarterly
2	Voice discovery	Whole essay (anti-rules)	3 “do not” rules	Once, save as a pinned line
3	Voice discovery	Whole essay (skeleton)	5–8 section outline	Per essay
4	Voice discovery	Whole essay (voice card)	60–90 word voice card	Once, refresh every 6 months
5	First-line & open	Opening	10 opening lines	Per essay
6	First-line & open	Opening	3 sensory details	Per essay
7	First-line & open	Opening	3 replacement first sentences	When stuck on the open
8	First-line & open	Opening	3 mid-conversation opens	For opinion pieces
9	First-line & open	Opening	Rewritten first 30 words	After first full draft
10	Storytelling	The turn	4-sentence scene	Per essay
11	Storytelling	The turn	5 sensory prompts	When the scene is too big
12	Storytelling	The turn	But/therefore rewrite	After first draft of scene
13	Storytelling	The turn	Want-stated rewrite	When the scene feels static
14	Storytelling	The handoff	3 callback closers	At the end of the essay
15	Opinion & POV	The stake	1 hidden claim	When the draft is story-only
16	Opinion & POV	The stake	Steelman paragraph	Before publishing
17	Opinion & POV	The stake	SAFE / STRETCH / BOLD claim	When you’re hedging
18	Opinion & POV	The stake	JTBD-shaped claim	When the topic is vague
19	Opinion & POV	The stake	3 contrarian-adjacent angles	For differentiated takes
20	Revision	Whole essay	Trimmed 20%	After first full draft
21	Revision	Whole essay	Paragraph diagnosis	After first full draft
22	Revision	Whole essay	Adverb-by-adverb pass	After trim
23	Revision	Whole essay	Stranger-read feedback	Final pass
24	Revision	Open	12 subject lines	Right before publish
25	Revision	Whole essay	12-point pre-flight check	Right before publish

A 7-day “raw essay” workflow

This is the calendar I use when I want a real conversational essay out the door in a week. It assumes you’re writing one essay, ~1,500 words, in your own Substack voice.

Day 1 - Voice prep (45 minutes)

Run Prompt 1 on 5–10 past paragraphs.
Run Prompt 4 and save the voice card.
Pin both to a notes doc.

Day 2 - Open the question (30 minutes)

Pick the one thing you want to say this week.
Run Prompt 18 (Jobs to be Done) to name the job the essay is doing.
Run Prompt 19 to find the angle nobody else has.

Day 3 - Scene hunting (60 minutes)

Run Prompt 11 with the small-moment questions.
Pick 2–3 candidate scenes.
Run Prompt 10 on the best one.

Day 4 - Draft in one pass (90 minutes)

Write the full draft in one sitting. Don’t stop, don’t edit, don’t fact-check.
Paste it into Substack’s draft view.

Day 5 - Find the spine (60 minutes)

Run Prompt 15. Find the claim.
Run Prompt 17. Pick SAFE / STRETCH / BOLD.
Run Prompt 20. Cut 20%.
Run Prompt 21. Tag every paragraph.

Day 6 - Polish (60 minutes)

Run Prompt 22. Kill the adverbs.
Run Prompt 12. Make every transition “but” or “therefore.”
Run Prompt 14. Write the callback closer.
Read the whole thing out loud, once.

Day 7 - Publish (30 minutes)

Run Prompt 24. Pick the subject line.
Run Prompt 25. Pre-flight check.
Schedule for Tuesday or Thursday, 8 a.m. in your reader’s time zone.
Reply to every comment for the first 24 hours.

That’s the rhythm. It’s not magic. It’s just predictable.

Common mistakes to avoid

These are the 8 mistakes I see most often in Substack essays that are otherwise solid. Skip any one of them and your essay reads 2x better.

Opening with a thesis instead of a scene. If your first sentence could be the first sentence of any other essay on the same topic, it’s the wrong first sentence.
Hedging the claim. “Some people say X, but others say Y, and the truth is probably somewhere in the middle” is the most common essay shape on Substack, and it’s also the most forgettable. Prompts 15 and 17 fix this.
Writing “I think…” / “In my opinion…” / “It seems to me…” All three phrases add zero information and signal low confidence. Cut them. Your first-person framing already says it’s your opinion.
Letting ChatGPT pick the verbs. Adverbs are a tell; lazy verbs are a bigger tell. Run Prompt 22 every time.
Forgetting the closing CTA. “If this resonated, hit reply” outperforms “follow for more” by a wide margin. The reader’s job in the last paragraph is to know what to do, not to feel vaguely grateful.
Treating the subject line as an afterthought. The subject line is the most-edited sentence in any newsletter. Spend 20 minutes on it. Prompt 24 helps.
Imitating a writer’s voice instead of their structure. Steal like an artist, says Austin Kleon in Steal Like an Artist (2012). Steal the shape - the skeleton of where the turn happens, where the claim lands - not the slang, not the punctuation, not the verbal tics. Prompt 3 is built around this.
Shipping the first AI pass. Always rewrite in your own voice. Always. The June 2026 post on On Substack titled The independent writer’s advantage in the age of AI is explicit on this: the machines can summarize, but they cannot see, feel, or break news - and trust is about the messenger, not the sentence (source).

Final word

You don’t need more prompts. You need to use the ones you have. Take prompts 1, 4, 5, 10, 15, 20, and 25 - that’s 7 prompts - and run them on your next essay. That one essay will be better than the last five you’ve shipped.

Substack as a platform keeps adding tooling - the new Reply Rules (June 2026), the Substack Recording Studio (March 2026), the Substack TV app for Apple TV and Google TV (January 2026), and the long-rumored move into sponsored newsletter segments (December 2025). All of that is in service of the same bet Substack has been making since 2017: writers win when they own the relationship with the reader.

Your job is to make sure the writing on the inside of that relationship is actually yours. ChatGPT is fine. The voice is the moat.

Now close this tab, open a blank draft, and run Prompt 5.

Reader disclosure & educational-purpose notice

This page is published by SuperFreshAI for general informational and educational purposes only. By reading it, you agree to the points below.

Editorial independence. All reviews, guides, and recommendations are written by our editorial team based on hands-on use. Some links on this site are affiliate links, and some articles are produced as partner content — both are always clearly labeled. Our editorial conclusions are never shaped by partners or affiliates.
Not professional advice. Nothing on this page constitutes legal, financial, medical, tax, or other professional advice. AI tools, pricing, and capabilities change quickly — always verify current information with the tool's official documentation before making a decision.
Educational purpose only. The content here is intended to help you learn about AI tools and workflows. It is not a guarantee of results, performance, fitness for a particular purpose, or suitability for your specific situation. Your results may vary.
No warranties. The site and its content are provided on an "as is" and "as available" basis. We make no warranties, express or implied, about accuracy, completeness, reliability, or availability. See our Terms and Privacy for the full legal terms.
Your responsibility. You are responsible for how you use the information on this page, including any decisions you make based on it. Always do your own research and consult a qualified professional when appropriate.
Affiliate & partner disclosure. When you click certain outbound links, we may earn a commission at no extra cost to you. When a piece of content is produced as partner content, it is labeled at the top of the page. See our Editorial Policy for the full standards we follow.

By continuing to read, you acknowledge that you have read and understood this notice.