Medium / Article Titles Beginner

36 ChatGPT prompts for Medium writers to craft curiosity-gap article titles

Your Medium title is the only thing standing between your brilliant draft and the void. Almost everything else - the cover image, the subtitle, the claps, the read time - depends on the first four to ten words a reader sees. If those words fail to prick the brain, the rest of the article never gets a chance. That is the entire game.

A curiosity-gap title is one that opens a question the reader’s brain can’t help but try to close. It does not lie. It does not trick. It simply teases a piece of information the reader does not have yet, and the brain reads on to fill the gap. George Loewenstein’s 1994 information-gap theory frames this perfectly: when we sense a gap between what we know and what we want to know, we feel an unpleasant tension, and we click, scroll, or read to make the tension go away (Wikipedia – George Loewenstein).

This guide gives you 36 extremely detailed ChatGPT prompts for Medium curiosity-gap titles, grouped by the job they do. Some mine topics. Some turn topics into lists. Some turn lists into questions. Some force a story frame, a contrarian frame, or a “how to” frame. A final group helps you A/B test and measure so the title that ships is the one that earned it. Every prompt is multi-line, ready to paste, and comes with a worked example plus a pro tip. By the end you will have a complete workflow - from blank page to title that pulls reads.

Pull quote - the single most useful stat in this whole piece: “On average, five times as many people read the headline as read the body copy.” - David Ogilvy, cited in Buffer’s 30+ Headline Formulas guide (Aug 2014). On Medium, where algorithm and search both rank on click-through, that 5× gap is the difference between 200 reads and 1,000.

Quick answer / TL;DR box

  • A curiosity-gap Medium title teases information the reader does not have yet, using Loewenstein’s 1994 information-gap principle to create a mental itch that only a click can scratch.
  • The strongest gaps on Medium have four parts: a specific audience, a concrete promise, an open loop, and a length-and-format cue (number, list, or question).
  • A “good” Medium title typically lands in the 8–18 word range, contains a number or power word in the first three positions, and answers a “what’s in it for me” question in under one second.
  • ChatGPT is a multiplier on these structures, not a replacement for them. The 36 prompts below are organized by formula (lists, questions, stories, how-tos, contrarian, A/B), and each one is a multi-line recipe you can paste into ChatGPT in seconds.
  • If you only do one thing this week: pick 5 of your drafts, run them through prompts 1, 6, 11, 16, and 21, and A/B test the top three with a tool like Coschedule Headline Analyzer or Anyword.

Why most Medium titles die at 200 reads (with 2026 data)

Medium’s distribution is brutal. The platform reached 60 million monthly readers in 2016, and that number is the most recent official count cited in Wikipedia’s Medium entry (last edited 2026). It is also the number still doing the rounds in 2026 industry write-ups. A platform that big, with that much content, drowns most new stories before they reach 200 reads.

Two things make this worse in 2026:

  1. Read-time pays. Wikipedia’s Medium entry confirms the Partner Program “distributes a portion of each member’s subscription fee to the writers they read most each month.” A title that pulls a click but loses the reader in 30 seconds is a net loss. Your title has to do two jobs: get the click and earn the read.
  2. AI content is paywalled-out. Per the same Wikipedia article, starting May 1, 2024, Medium banned AI-generated content from enacting paywalls and receiving payouts via the Partner Program. Generic, formula-feeling titles - the kind a sloppy ChatGPT prompt spits out - are now algorithmically suspect. Your title has to feel human, or at least human-curated.

Buffer’s headline analysis, drawing on David Ogilvy, is the simplest way to see why titles matter: for every 5 people who read a title, only 1 reads the body (Buffer, 30+ Headline Formulas). If 200 people saw your last Medium post, only about 40 of them read the title all the way through. The other 160 scrolled past. A better title can move that 40 to 100 - a 2.5× lift without changing a single word of the article.

A curiosity gap is the cleanest, least spammy way to do it. It teases - it does not trick. Readers don’t feel cheated when the article delivers. They feel smart for clicking. That’s the contract you want.

The 4-part curiosity-gap title anatomy

Curiosity-gap title: a headline that opens a clear, specific, unanswered question the reader wants to close, while hinting that the article holds the answer. The four parts are:

  1. Specific audience - who this is for, in plain English. “for first-time SaaS founders” beats “for entrepreneurs.”
  2. Concrete promise - what the reader walks away with. A number, a transformation, a specific insight.
  3. Open loop - a deliberate gap between what the title says and what the body reveals. “The 3-word trick that 7-figure writers use” creates a loop: what is the 3-word trick?
  4. Format or length cue - a number, a question mark, “how to”, or a story frame. This sets expectations and lets the brain know how much to commit.

Stack the four in that order and you get something like: “For first-time Medium writers: the 3-word opening line 7-figure writers use (and you can steal it tonight).” The audience is the first half. The promise is the second half. The open loop is the parenthetical tease. The format cue is the “3-word” and the implied “how to.”

The 36 prompts below are organized by which of these four parts they emphasize. Use them in any order, but always make sure all four show up in the final title.

SECTION 1 - Topic-mining prompts (prompts 1-5)

These five prompts help you find angles worth writing about in the first place. A great title on a boring topic is a waste. Mine the topic pool first, then feed your best ideas into the formula prompts later in the guide.

Prompt 1 - The Reader-Worry Miner

Purpose: Generate 10 curiosity-gap titles built on the actual worries your Medium readers have. This prompt is great when you’re staring at a draft and don’t know which angle to lead with.

Multi-line prompt (paste into ChatGPT):

You are a Medium editor in 2026 who specializes in viral curiosity-gap titles.

Topic: [INSERT YOUR TOPIC, e.g. "freelance writing rates in 2026"]
Audience: [INSERT AUDIENCE, e.g. "first-time Medium writers in the Partner Program"]

Step 1: List 10 specific worries this audience has about the topic.
Step 2: For each worry, write a curiosity-gap Medium title that:
  - names the audience in the first 4 words
  - hints at an answer without giving it away
  - uses a number, question, or "how to" as the format cue
  - sits between 8 and 18 words
Step 3: Score each title 1-10 on curiosity-gap strength, specificity, and click-feel.
Step 4: Return the top 3 with a one-sentence rationale for each.

Do not use clickbait tricks like "You won't believe" or "Shocking truth."

Example output (for topic = freelance writing rates in 2026, audience = first-time Medium writers):

Worry 3: “Am I undercharging?” Title: For First-Time Medium Writers: 7 Signals You’re Undercharging by 50% (and the 1 Question to Ask Clients First) - Score: 9/10 Rationale: Audience-specific, numeric specificity (50%, 7, 1), and a closed loop (the “1 question”) force a click to close the gap.

Pro tip: The “worries” step is the secret. ChatGPT is much better at titles when it’s working from concrete human fears, not abstract topics. If the worries list feels generic, ask it to “go deeper - what’s the unspoken version of each worry?”

Prompt 2 - The Search-Intent Rewriter

Purpose: Take a real Google “People Also Ask” question and turn it into a curiosity-gap Medium title. This is gold for SEO traffic because the question is already proven to be searched.

Multi-line prompt (paste into ChatGPT):

I'm writing a Medium article on [TOPIC]. The article answers this exact
People Also Ask question from Google:

"[PASTE THE PAA QUESTION]"

Generate 8 curiosity-gap Medium titles that:
  - turn the question into a clickable headline (not a question mark headline)
  - hide the answer behind a specific promise or number
  - lead with the audience in the first 3 words
  - land between 10 and 18 words
  - follow David Ogilvy's rule: front-load value, withhold the answer

For each title, add a 1-line note on which AEO (Answer Engine Optimization)
search query it would also match.

Avoid: "How to" if the article isn't a true tutorial. Avoid question marks
unless Betteridge's law (the question ends in "no") is acceptable.

Example output (PAA = “how much should I charge for a 1000-word article in 2026”):

For Freelance Writers: The 3 Numbers Behind Every 2026 Rate Sheet (and Why $0.10/word Is the New Floor) - Matches: “what to charge freelance writers 2026”, “freelance writing rate calculator”

Pro tip: Betteridge’s law of headlines (Wikipedia) tells us a question mark in a headline often signals a “no” answer, but A 2013 study by Lai & Farbrot found questions addressing the reader in second person get higher click-through rates than statements. So: rewrite the question as a statement that contains a hidden question. That’s the move.

Prompt 3 - The Reverse-Engineer From a Top Performer

Purpose: You see a Medium post crushing it. You want a similar (not copycat) curiosity-gap title for your own post. This prompt teaches the structure without copying the words.

Multi-line prompt (paste into ChatGPT):

Here is a curiosity-gap Medium title I admire:

"[PASTE TITLE]"

Step 1: Break the title into its 4 parts:
  - Specific audience
  - Concrete promise
  - Open loop
  - Format / length cue
Step 2: List the 3 structural moves that make it work.
Step 3: Now write 6 new titles for my own article using the same
  structural moves but on this new topic and audience:
  - Topic: [YOUR TOPIC]
  - Audience: [YOUR AUDIENCE]
  - Unique angle: [ONE-SENTENCE ANGLE]

Each new title must:
  - Be original, not a paraphrase
  - Keep the 4-part anatomy
  - Land between 10 and 18 words
  - Use a different power verb than the original

Example output (if admiring title = “I Wrote 100 Medium Articles in 12 Months. Here’s the Brutal Math Behind $4,200.”):

Structural moves identified: confession frame, exact number, ”$” sign, brutal honesty tone. New titles on topic = “cold email outreach for SaaS founders”: I Sent 1,200 Cold Emails in 90 Days. The 11-Word Reply That Closed 2 SaaS Deals - Score: 8/10

Pro tip: Resist the urge to copy the words. Copy the structure. “Brutal math behind $4,200” works because of confession + exact number + dollar figure. Swap in your own confession, your own number, your own dollar figure. ChatGPT often over-paraphrases; force it to keep the structure and change the substance.

Prompt 4 - The Comment-Section Gold Digger

Purpose: Mine real reader pain from comment sections, Reddit, Quora, and Amazon reviews. Real comments = real curiosity gaps.

Multi-line prompt (paste into ChatGPT):

I'm writing a Medium article on [TOPIC]. Here are 8 real comments /
questions / complaints I pulled from [Reddit / Quora / Amazon / a
competitor's Medium post]:

[PASTE 8 COMMENTS, ONE PER LINE]

Step 1: Cluster these into 4 distinct reader pains.
Step 2: For each pain, write a curiosity-gap Medium title that:
  - quotes or paraphrases the actual reader language
  - hides the answer behind a specific number or promise
  - uses first-, second-, or third-person story framing
  - lands between 8 and 18 words
Step 3: Pick the single best title and explain why the comment language
  makes it irresistible.

Example output (topic = “beginner investing” and 8 Reddit comments about being scared to start):

Beginner Investors: The 3-Second Question That Made Me Invest $500 the Day I Was Scared - uses real reader language (“scared”) plus a specific number and a confession frame.

Pro tip: Comment language is the only “voice of customer” research you can do in 30 minutes. Save every meaningful comment you see in your niche for one month. When you have 50+, run this prompt. You will get titles that sound like the reader’s own thought.

Prompt 5 - The 4-Us Stress Test

Purpose: Score your existing draft titles against the “4 Us” headline framework (Useful, Urgent, Unique, Ultra-specific) popularized by Copyblogger and Joseph Sugarman.

Multi-line prompt (paste into ChatGPT):

You are a senior copy editor trained in the "4 Us" of headlines
(Useful, Urgent, Unique, Ultra-specific).

Score these 5 Medium draft titles on each of the 4 Us, 1-10:

1. [TITLE 1]
2. [TITLE 2]
3. [TITLE 3]
4. [TITLE 4]
5. [TITLE 5]

For each title, output:
  - Useful score and what would make it more useful
  - Urgent score and what would make it more urgent
  - Unique score and what would make it more unique
  - Ultra-specific score and what would make it more specific
  - Total score out of 40
  - A revised curiosity-gap version that scores 35+

Return the table in markdown.

Example output (draft = “Tips for better Medium titles”):

Useful: 4, Urgent: 2, Unique: 1, Ultra-specific: 1. Total: 8/40. Revised: For New Medium Writers: 11 Title Formulas That 7-Figure Writers Steal From Each Other (and the 1 to Avoid) - Estimated 38/40.

Pro tip: The 4 Us work because they map directly to reader brain states. Useful = “what’s in it for me?” Urgent = “do I need this now?” Unique = “have I seen this before?” Ultra-specific = “do I trust the claim?” If a title is weak in one, the brain ignores it. The 36 prompts in this guide fix different U combinations, so mix and match.

SECTION 2 - Number & list prompts (prompts 6-10)

Numbers in titles are a cheat code. Buffer’s headline research, citing KISSmetrics, found readers “absorb the first three words of a headline and the last three words,” and a number in either position increases skim-comprehension dramatically (Buffer, 30+ Headline Formulas). These five prompts are dedicated to numeric, list-style curiosity-gap titles.

Prompt 6 - The Odd-Number Oddity

Purpose: Generate curiosity-gap titles that use odd, specific numbers (not round 5s or 10s) to feel more credible and clickable.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]

Odd, specific numbers consistently outperform round ones in curiosity-gap
headlines (e.g., "7" beats "5" or "10", "31" beats "30").

Generate 10 curiosity-gap Medium titles that:
  - use an odd number between 3 and 49
  - pair it with a specific promise or transformation
  - keep the open loop tight (the article must answer the count)
  - land between 8 and 18 words
  - front-load the audience or the number in the first 3 words

For each title, also write a 1-sentence subhead that would go on
Medium as the deck. Total output: 10 title/subhead pairs.

Example output (topic = “morning routines for writers”):

7 Morning Routines of Writers Who Publish Every Week (and the 1 That Actually Matters) - Subhead: Spoiler: it’s not the cold shower or the 5am alarm.

Pro tip: Pair your odd number with “and the 1” at the end. The “1” creates a smaller, closable loop inside the bigger list. Readers click to find out which one is the secret.

Prompt 7 - The Listicles-With-A-Twist

Purpose: Standard listicles are dead. Twisted listicles (lists where the items are surprising, contradictory, or personal) are alive. This prompt generates them.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]

Write 8 curiosity-gap Medium titles that are "twisted listicles":
  - the list is the surprise (e.g., "9 Things I Stopped Doing in 2026")
  - the list items contradict what most readers expect
  - at least one item in the list is shockingly small (e.g., 1) or large
  - the title promises a story behind the list

Constraints:
  - 10-18 words
  - use the structure "[NUMBER] [THINGS] [AUDIENCE] [STOPPED /
    STARTED / LEARNED / WISH THEY KNEW]"
  - hide the items in the title (don't list them)
  - avoid generic verbs (use steal, ban, drop, swap, replace, untag)

Example output (topic = “LinkedIn for solopreneurs”):

11 LinkedIn Habits I Banned in 2026 (Solopreneurs: Number 7 Tripled My Replies Overnight)

Pro tip: Copyblogger’s “Headline + a little something extra” pattern fits twisted listicles perfectly. The listicle is the headline. The “something extra” is the one item that performed 10× better. Readers have to find out which one.

Prompt 8 - The Steal-From-X List

Purpose: Reframe a “things [AUDIENCE] should know” list into a “things [SUCCESSFUL GROUP] do that [YOUR AUDIENCE] don’t” list. The “outgroup” creates curiosity.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
My audience: [YOUR AUDIENCE]
Outgroup I admire: [SUCCESSFUL GROUP, e.g. "7-figure writers", "YC founders",
"professionals earning $200K+"]

Write 10 curiosity-gap Medium titles using the "Outgroup does X, you don't"
frame:
  - number between 5 and 21
  - explicit reference to the outgroup in the first 8 words
  - explicit (or implied) reference to my audience in the title
  - 10-18 words
  - hide the actual list (don't list items in the title)

For each, add 1 sentence on why an outgroup reference boosts CTR.

Example output (topic = “morning routines”, audience = “freelancers”, outgroup = “7-figure copywriters”):

12 Things 7-Figure Copywriters Do Before 9am (Most Freelancers Skip 8 of Them)

Pro tip: The phrase “7-figure” or “6-figure” or “Y Combinator” or “FAANG” is itself a curiosity trigger because it signals an outgroup with a secret. Use it sparingly and only if you can deliver in the body - readers will bounce hard if the outgroup is fake.

Prompt 9 - The Countdown Curiosity Title

Purpose: Numbered countdown titles (“from #21 to #1 the list gets better”) are an old Outbrain playbook that still works. Use this when you have a list with a clear ranking.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
List I want to rank: [LIST OF 5-15 ITEMS, ONE PER LINE]
Audience: [AUDIENCE]

Write 8 curiosity-gap Medium titles in the countdown style:
  - mention the total count (e.g., "12 Things...")
  - tease that the BEST item is at #1 (or the worst is at #1)
  - hide what the items actually are
  - 10-18 words
  - the open loop is "what's at the top / bottom"

Also write 2 alternative titles that do NOT use countdown but use the
same list. Total: 10 titles.

Example output (topic = “AI tools for solo writers”):

21 AI Tools Ranked Worst to Best for Solo Writers in 2026 (You Already Use #14)

Pro tip: “You already use #14” is a personalization trick. It assumes a behavior the reader probably has, then makes them click to confirm or deny. It’s the lowest-friction curiosity gap you can build.

Prompt 10 - The X-Item Reader-Promise Title

Purpose: Force the title to make a specific, falsifiable promise. No fluff, no “ultimate guide,” no “everything you need to know.”

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]

Write 10 curiosity-gap Medium titles that:
  - use a number between 3 and 21
  - end with a specific, falsifiable promise (a dollar amount, a time
    saved, a count, a transformation, a specific outcome)
  - front-load the audience in the first 4 words
  - 10-18 words total
  - the open loop is "how will the article deliver on the promise"

For each title, also write 1 sentence explaining exactly what data,
story, or framework the article would need to deliver the promise
without feeling clickbait-y.

Example output (topic = “productivity”, audience = “writers”):

For Writers: 9 Time-Blocks That Replaced My 60-Hour Weeks (I Now Ship 40 Posts a Month in 28 Hours)

Pro tip: Pair a specific promise with a falsifiable mechanism. “I now ship 40 posts a month in 28 hours” is so specific the reader knows exactly what they’re clicking for. If the body delivers, the title compounds trust. If the body doesn’t, the title burns trust forever. Pick promises you can 100% back up.

SECTION 3 - Question prompts (prompts 11-15)

Question titles are the most overused - and most underused - format in Medium. Overused because most writers just paste a question. Underused because most writers don’t know that questions addressing the reader in second person dramatically outperform statements. A 2013 study published in Social Influence and cited in Wikipedia’s Betteridge’s law entry found that “questions that address or reference the reader have statistically significant higher click-through rates than rhetorical or general questions.” Use these five prompts to weaponize that.

Prompt 11 - The “Are You” Self-Audit Title

Purpose: A question that puts the reader in the headline as a character, then hides the answer.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]

Write 10 curiosity-gap Medium titles that start with "Are you" and:
  - name a specific behavior, mistake, or symptom in the question
  - hide the answer behind a number or framework
  - land between 10 and 18 words
  - use second person ("you", "your") to address the reader directly
  - pair the "Are you" with a promise that the body gives a fix

For each title, list 3 follow-up question titles the article itself
could use as H2 subheads.

Example output (topic = “freelance rates”, audience = “writers”):

Are You Charging 40% Less Than You Should in 2026? (3 Rate-Reset Questions to Ask Yourself Today)

Pro tip: “Are You [VERB]-ing” is the only question frame that consistently beats a statement title. That’s because the reader’s brain automatically answers “yes” or “no” - and to resolve the answer, they click. Betteridge’s law is right for general questions. For “are you specifically doing X” questions, the law is wrong, and CTR goes up.

Prompt 12 - The “What If” Possibility Title

Purpose: Open a hypothetical future the reader can picture themselves in, then dangle the cost of getting it wrong.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]

Write 10 curiosity-gap Medium titles using the "What if" structure:
  - open with a vivid, specific future scenario
  - pair it with a hidden cost or hidden opportunity
  - the body must answer the "what if" (you need to actually deliver
    the scenario)
  - 10-18 words
  - no question mark at the end (convert the question into a
    statement that contains a question)

For each title, add 1 line describing the worst-case scenario the
article prevents and the best-case scenario it enables.

Example output (topic = “career pivots”):

What If You Quit Your 9-to-5 in 2026 With $4,200 Saved? (The 7-Month Plan I Wish I Had on Day 1)

Pro tip: Convert every “what if” into a statement. “What If You Quit Your Job” is a question. “What Happens When You Quit Your Job With $4,200” is a statement that contains a question - and statements avoid the Betteridge’s-law “the answer is no” problem. Big CTR swing, no cost.

Prompt 13 - The “Why” Mystery Title

Purpose: Use “why” to imply there’s a hidden reason behind something the reader already does. They click to find the reason.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]

Write 10 curiosity-gap Medium titles that:
  - start with "Why" or contain "the real reason" / "the hidden reason"
  - point at a behavior the audience does daily or weekly
  - hide the actual reason behind a number or a story
  - 10-18 words
  - imply that the mainstream explanation is wrong or incomplete

For each title, also write 1 sentence naming the mainstream
explanation the article will overturn.

Example output (topic = “productivity”, audience = “writers”):

Why 90% of Writers Hit a Wall at 800 Words (the Real Reason Isn’t Writer’s Block)

Pro tip: “Why” is the most dangerous headline word. It works because readers believe there is always a hidden reason, even when there isn’t. Use it when you actually have a contrarian or non-obvious reason. If your “why” is a cliché (“consistency is the real reason”), skip this format and use Prompt 14 instead.

Prompt 14 - The “How” Promise Title

Purpose: A “How to” or “How I” title that hides the actual mechanism. The most-clicked Medium format, and the most-misused.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]

Write 10 curiosity-gap Medium titles that use the "How to" or
"How I" frame:
  - name a specific outcome (a number, a transformation, a status)
  - hide the mechanism (the actual "how")
  - 10-18 words
  - include the audience in the first 4 words
  - avoid "How to be a better X" - too vague

For each title, list 3 alternative openings (instead of "How to")
that still create the same curiosity gap. Examples: "The 3-step
process I used to", "I figured out how to", "The $0 trick that
got me".

Example output (topic = “email list growth”, audience = “writers”):

How I Grew My Email List From 47 to 4,200 in 6 Months (Without a Single Paid Ad)

Pro tip: The “How to” frame works best when paired with a specific number that seems impossible or implausible. “How to grow your list” is weak. “How I grew from 47 to 4,200” is strong because the brain wants to know the math. The article must deliver the math.

Prompt 15 - The “Do You Make These Mistakes” Trap

Purpose: A classic from John Caples’s Tested Advertising Methods (cited in Buffer’s headline guide). Questions that address the reader’s mistakes pull enormous clicks because nobody wants to be the fool who keeps making the same mistake.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]

Write 10 curiosity-gap Medium titles in the "Do you make these
mistakes" / "Are you still doing X wrong" frame:
  - name 3-7 specific, common mistakes in the title
  - front-load "Do you" or "Are you" in the first 3 words
  - hide whether the reader actually makes them
  - 12-22 words
  - end with a promise (a fix, a checklist, a new framework)

For each title, list the 5 actual mistakes the article would need
to cover to deliver on the promise.

Example output (topic = “LinkedIn outreach”, audience = “SaaS founders”):

Do You Still Make These 7 LinkedIn Outreach Mistakes in 2026? (Most Founders Get 1 and 4 Wrong)

Pro tip: A 2013 study of 26,000+ news headlines, cited in Wikipedia’s Betteridge’s law entry, found that “yes/no” question headlines can be answered “yes” or “no” - and Betteridge’s law breaks down for mistake questions because the reader’s reflex answer is “I don’t know.” That uncertainty is the click. Use it.

SECTION 4 - Story prompts (prompts 16-20)

Story titles work on Medium because the platform is built for serialized, personal, and narrative writing. A confession, a transformation, or a “here’s what happened” frame gets the click because the reader wants the resolution.

Prompt 16 - The Confession Title

Purpose: Open with a failure, mistake, or embarrassing moment. Readers click to find out what happened next.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]
True story I can tell: [ONE-SENTENCE STORY]

Write 10 curiosity-gap Medium titles in the confession frame:
  - open with a vivid, specific moment (a number, a date, a
    dollar amount, or a location)
  - use first person ("I", "My", "We")
  - end with the outcome or the lesson (without revealing the
    exact story)
  - 10-18 words
  - the open loop is "what happened and what did I learn"

For each title, write 1 sentence on what the article's first
paragraph must do to deliver on the confession promise.

Example output (topic = “Medium growth”, audience = “writers”):

I Published 47 Medium Articles in 6 Months and Made $214 (Here’s the 1 Mistake That Cost Me $4,000)

Pro tip: A confession title works only if the confession is real and specific. “$214” is real. “I made very little” is not. Real numbers and dates are the trust signal. Without them, the title reads as a humble-brag.

Prompt 17 - The Before/After Transformation

Purpose: Show the contrast between the old state and the new state. The gap between them is the curiosity.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]
Before state: [SHORT DESCRIPTION]
After state: [SHORT DESCRIPTION]
Timeframe: [HOW LONG THE TRANSFORMATION TOOK]

Write 10 curiosity-gap Medium titles that:
  - show the contrast (before X → after Y)
  - name the timeframe in the title
  - hide the actual mechanism that drove the change
  - 10-18 words
  - use a story frame ("I went from X to Y", "How I turned X into
    Y", "From 0 to $20K in 90 days")

For each title, name the single mechanism the article must
reveal to deliver on the title.

Example output (topic = “freelance writing”, audience = “writers”):

From $0 to $11,400/Month in 11 Months (The 1 Pricing Change That Drove 80% of the Growth)

Pro tip: The strongest transformation titles are specific in two ways: the timeframe (11 months) and the lever (the pricing change). If you can only be specific in one, be specific on the lever - readers want to know the one thing that drove the change, not the calendar.

Prompt 18 - The “I Tried X” Experiment Title

Purpose: Frame the article as a personal experiment with measurable results. The reader clicks to see the data.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]
Experiment: [WHAT YOU DID, e.g. "I published every day for 30 days"]
Result: [WHAT HAPPENED, e.g. "1,200 new followers and $640 in earnings"]

Write 10 curiosity-gap Medium titles that:
  - lead with the experiment
  - lead with the result in the first half
  - hide the specific tactics used in the experiment
  - 10-18 words
  - use first person
  - the open loop is "what were the tactics and which worked"

For each title, also list 1 alternative title that doesn't use
"experiment" language but uses the same data.

Example output (topic = “Twitter growth”):

I Tweeted 3 Times a Day for 30 Days (The 1 Type of Tweet That Drove 71% of My Growth)

Pro tip: The “I tried X” frame is now the single most-clicked format on Medium, according to multiple creator-economy writers in 2025-2026. The reason is asymmetric risk: the reader gets the data, the writer already paid the cost. If you’ve actually run the experiment, this is the prompt for you.

Prompt 19 - The Day-One Diary

Purpose: A specific day-in-the-life title that gives the reader a peek behind the curtain. Works for “what does a 7-figure writer actually do” posts.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Role: [YOUR ROLE OR THE ROLE YOU'RE PROFILING]
A typical day: [LIST 4-8 THINGS THAT HAPPEN IN THE DAY]

Write 10 curiosity-gap Medium titles that:
  - center on a single day (or a single morning, or a single hour)
  - include a specific number (the time, the count, the dollar
    amount)
  - hide which parts of the day are most important
  - 10-18 words
  - use first person or third person ("A Day in the Life of")

For each title, name the 3 parts of the day the article must
cover to make the title pay off.

Example output (topic = “writing a book”):

A Day in the Life of a Writer Who Hits 2,000 Words by 8am (6 Small Habits That Add Up)

Pro tip: Diary titles work because they promise the reader the texture of a successful life, not just the outcome. The trick is the smallest specific time unit that still feels real. “A Day” is okay. “A Morning” is sharper. “The 47 minutes between my alarm and my first cup of coffee” is irresistible.

Prompt 20 - The Hero-Customer Title

Purpose: Center the title on a real customer, reader, or person you helped. The hero is not you - it’s them. This is StoryBrand 101, and it pulls trust + curiosity at the same time.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Customer / reader / case study: [NAME, ROLE, ONE-LINE OUTCOME]
Their problem before: [THEIR STARTING POINT]
Their result after: [THEIR END POINT]

Write 10 curiosity-gap Medium titles that:
  - name the customer or anonymize them as "a 7-figure X" / "one
    Medium writer"
  - hide the specific mechanism that helped them
  - 10-18 words
  - third person ("She went from X to Y", "How [NAME] turned X
    into Y")
  - the open loop is "what did I do that worked"

For each title, list the 1 piece of data or proof (a metric, a
quote, a before/after) the article must include to make the
title credible.

Example output (topic = “career coaching”, customer = “Sarah, 31, product manager”):

Sarah Was Laid Off in March 2026. By July, She’d Landed a $174K Remote Role. Here’s the 1 Email That Did It.

Pro tip: The “hero-customer” title is the most overused and under-delivered format in B2B and creator content. It only works if the body actually shows the work, not just the result. Lead with the customer, not with yourself. Always.

SECTION 5 - “How to” prompts (prompts 21-25)

The classic Medium format. These five prompts focus on making the “how to” feel fresh, specific, and impossible to skip.

Prompt 21 - The Reverse-How-To

Purpose: Take a “how to” and reverse it: instead of telling people how to do X, tell them how to stop doing X, or how to do the opposite of X. Reverse titles are weirdly magnetic.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Standard "how to" headline: [e.g. "How to write a Medium title"]
Audience: [AUDIENCE]

Write 10 reverse-curveball curiosity-gap Medium titles that:
  - invert or contradict the standard "how to"
  - lead with the contradiction in the first 5 words
  - hide the alternative in the title
  - 10-18 words
  - use "why most", "what to do instead", "the opposite of",
    "stop doing", "kill your", or "the anti-[X] guide"

For each title, write 1 sentence on the contrarian insight the
article must lead with to make the title work.

Example output (topic = “writing headlines”):

Stop Writing “How To” Medium Titles (The 3 Formats That Beat Them Every Time in 2026)

Pro tip: Reverse titles work because they violate the pattern the reader expects. Your article doesn’t have to actually reverse the standard advice - it can just say “the standard advice is broken, here’s what’s working now.” The title sets the expectation. The body delivers the surprise.

Prompt 22 - The Step-Ladder How-To

Purpose: A “how to” that breaks the process into 3, 5, or 7 explicit steps. Steps feel actionable, and actionability drives clicks from readers who are ready to do the work.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]
End state the reader wants: [END STATE]

Write 10 step-ladder curiosity-gap Medium titles that:
  - name the number of steps in the title
  - name the end state in the title
  - hide the actual steps (don't list them)
  - 10-18 words
  - include the audience in the first 4 words
  - the open loop is "what are the steps in order"

For each title, list the 5-7 actual steps the article must cover
in order, with the hardest step highlighted.

Example output (topic = “freelance pitching”, audience = “writers”):

For New Writers: The 7-Step Cold Pitch That Books 3 Calls a Week (Step 4 Is Where Most People Quit)

Pro tip: “Step X is where most people quit” is a powerful inline loop. It teases a specific step and implies the rest of the article tells you how not to quit. The body should reward the click by actually showing the step in detail.

Prompt 23 - The Time-Boxed How-To

Purpose: A “how to” with a specific time budget. “How to write 1,000 words in 30 minutes” beats “How to write faster” every time.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE]
Time budget: [HOW LONG, e.g. "in 11 minutes", "in 7 days", "before
your next meeting"]

Write 10 time-boxed curiosity-gap Medium titles that:
  - name the time budget in the title
  - pair it with a specific outcome (a number, a transformation)
  - 10-18 words
  - include the audience in the first 4 words
  - the open loop is "what's the exact sequence to hit this in
    the time given"

For each title, also write 1 sentence on the one shortcut or
compromise the article must reveal to deliver in the time
budget honestly.

Example output (topic = “writing”, audience = “busy professionals”):

For Busy Writers: How to Ship a 1,500-Word Medium Post in 90 Minutes (The 4-Paragraph Skeleton I Use)

Pro tip: The shorter the time budget, the more the reader needs to trust you. “In 90 minutes” is credible. “In 90 seconds” is clickbait. Stay on the side of credibility or you burn trust in the first paragraph.

Prompt 24 - The Tool-Stack How-To

Purpose: A “how to” that names specific tools (ChatGPT, Surfer, Notion, Beehiiv, Substack, etc.) and stacks them into a workflow. Tool-stack titles pull readers who are actively building.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Tools I want to feature: [LIST 2-5 TOOLS]
Audience: [AUDIENCE]
Workflow result: [WHAT THE READER ENDS UP WITH]

Write 10 tool-stack curiosity-gap Medium titles that:
  - name 2-4 specific tools in the title (real, current tools)
  - name the result in the title
  - hide the exact sequence in which the tools are used
  - 10-18 words
  - the open loop is "in what order, and what does each tool
    actually do"

For each title, list the 4-6 step sequence the article must
walk through, with one tool per step.

Example output (topic = “newsletter creation”):

How I Built a 4,200-Sub Newsletter Stack With ChatGPT, Beehiiv, and Notion (In 11 Days)

Pro tip: Tool-stack titles work when the named tools are recognizable to your audience. If your audience doesn’t know what Surfer is, naming it is noise. Name the tools they already use or want to use. ChatGPT is universal. A niche SEO tool may not be.

Prompt 25 - The Beginner-To-Pro How-To

Purpose: A “how to” framed as a journey from beginner to pro. The “beginner” lowers the bar, the “pro” raises the aspiration, and the gap is the click.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Audience: [AUDIENCE, beginner]
Pro state: [WHAT PRO LOOKS LIKE]

Write 10 "beginner-to-pro" curiosity-gap Medium titles that:
  - name the starting point (beginner, 0 to 1, day 1, never
    written before)
  - name the end point (pro, 6-figure, $10K months, etc.)
  - hide the actual journey (the milestones, the tools, the
    framework)
  - 10-18 words
  - the open loop is "what are the milestones in order"

For each title, list the 5-7 milestones the article must
cover, with the milestone most beginners fail at highlighted.

Example output (topic = “writing online”):

From 0 Followers to 11,000: The 6 Milestones Every New Medium Writer Hits (Most Stall at #4)

Pro tip: “Most stall at #X” is the inline loop. Use it only if the body actually addresses the stall point in detail. If the article doesn’t, the title feels manipulative. Always deliver on the promise of the inline loop.

SECTION 6 - Contrarian & AEO prompts (prompts 26-30)

Contrarian titles exist to challenge a widely-held belief in your niche. AEO (Answer Engine Optimization) titles exist to be quoted by Google, ChatGPT, Perplexity, and Bing Copilot. These five prompts combine both - they challenge a belief and make a clean, citable claim.

Prompt 26 - The “Everyone Is Wrong” Title

Purpose: A direct challenge to a popular belief. Pulls readers who secretly suspect the belief is wrong but want confirmation.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Popular belief in this niche: [THE BELIEF, e.g. "write every day",
"long-form wins", "post on social 5x a day"]
Audience: [AUDIENCE]

Write 10 contrarian curiosity-gap Medium titles that:
  - directly contradict the popular belief in the first 6 words
  - hide the alternative viewpoint behind a number or a story
  - 10-18 words
  - the open loop is "if X is wrong, what's actually right"
  - use "stop", "kill", "is a lie", "doesn't work", "is dead",
    or "is the wrong move"

For each title, write 1 sentence on the data, study, or
first-hand experience the article must cite to make the
contrarian claim credible.

Example output (topic = “content marketing”, belief = “post daily”):

Posting Daily on Medium in 2026 Is a Waste of Time (I Made 4× More Posting Once a Week)

Pro tip: Contrarian titles are the single easiest way to get quoted in AI search engines. The cleanest ones are also the most falsifiable. A 2025-2026 HubSpot-style study on posting frequency can make a contrarian title citable for years.

Prompt 27 - The “Best/Worst X” Definitive Title

Purpose: “Best X” and “Worst X” titles work because they promise a ranked opinion. The reader clicks to see the ranking.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Items to rank: [LIST 5-15 ITEMS, ONE PER LINE]
Audience: [AUDIENCE]

Write 10 "best of" or "worst of" curiosity-gap Medium titles that:
  - name the year (2026)
  - name the audience
  - use a number between 5 and 30
  - 10-18 words
  - tease that the ranking is subjective, opinionated, or based
    on first-hand use
  - the open loop is "what's at #1 and #last"

For each title, also write 1 sentence on the criteria the
article uses to rank the items.

Example output (topic = “AI writing tools”):

The 9 Best AI Writing Tools for Medium Writers in 2026 (Ranked After 90 Days of Real Use)

Pro tip: “After 90 days of real use” is the trust signal. A 2026 best-of list without a usage window feels like a clickbait roundup. A 2026 best-of list with a clear usage window feels like research. The body must reference specific tests, prompts, or outputs from that window.

Prompt 28 - The Data-Backed “Study of X” Title

Purpose: Title the article as a study, mini-experiment, or data analysis. Pulls readers who trust data more than opinions.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Data source: [WHAT DATA YOU COLLECTED, e.g. "100 of my own Medium
posts", "a 1,000-post scrape", "60 days of A/B tests"]
Audience: [AUDIENCE]

Write 10 "data study" curiosity-gap Medium titles that:
  - mention the data source in the title
  - mention a specific number (post count, day count, sample size)
  - hide the actual finding
  - 10-18 words
  - the open loop is "what did the data show"

For each title, list the 1 chart or data point the article must
include to make the title credible.

Example output (topic = “Medium titles”):

I Analyzed the 100 Top Medium Articles of 2026 - 7 Title Patterns Showed Up in 83% of Them

Pro tip: Even a tiny dataset can be turned into a study. You don’t need 10,000 rows. You need 1 clean finding. “Of the 30 titles I A/B tested in May 2026, those with odd numbers had 2.1× higher CTR” is a study. The body shows the data. The title promises it.

Prompt 29 - The AI-Answer-Engine Title

Purpose: Write a title that Google’s AI Overviews, ChatGPT, Perplexity, and Bing Copilot will quote directly. AEO titles use plain factual language and front-load the answer.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
Specific factual claim: [THE CLAIM, e.g. "the average Medium read
time in 2026 is 4 minutes 12 seconds"]
Audience: [AUDIENCE]

Write 10 AEO-optimized curiosity-gap Medium titles that:
  - open with the most search-friendly phrase ("The best [X] in
    2026", "What is the [X]", "How to [X] in [TIME]")
  - include the year (2026) if relevant
  - include a number or specific claim
  - 10-18 words
  - the open loop is "what is the specific answer the article
    delivers"

For each title, also list the 3-5 "People Also Ask" questions
this title is most likely to be cited for in AI search.

Example output (topic = “Medium headline length”):

What Is the Best Medium Title Length in 2026? (Data From 4,200 Top Posts)

Pro tip: AI search engines quote titles that contain the specific phrase a user searched for, in the same order. The “What is the…” opener matches a common search query exactly. Pair it with a year and a number, and you have an AEO-friendly curiosity-gap title.

Prompt 30 - The “I Was Wrong” Mea Culpa

Purpose: A title where the author reverses their own previous advice. Pulls huge trust + curiosity. Readers click to find out what changed.

Multi-line prompt (paste into ChatGPT):

Topic: [TOPIC]
My previous advice: [WHAT YOU USED TO RECOMMEND]
What I now believe: [WHAT YOU BELIEVE NOW]
What changed: [THE TRIGGER, e.g. "data", "a reader's email",
"a personal experiment"]

Write 10 "I was wrong" mea culpa curiosity-gap Medium titles that:
  - own the previous advice explicitly ("I was wrong about X",
    "I take it back", "I was wrong in 2024, here's what I
    believe in 2026")
  - name the year
  - hide the new advice
  - 10-18 words
  - the open loop is "what changed and what do I believe now"

For each title, also list the 1 piece of evidence (a chart, a
quote, a study) the article must lead with to make the
admission credible.

Example output (topic = “writing frequency”):

I Was Wrong About Posting Every Day on Medium (The 30-Day Test That Changed My Mind in 2026)

Pro tip: Mea culpa titles convert like crazy because the reader doesn’t expect them. They also burn risk: if the admission feels fake, the reader bounces. Only use this prompt if you can show the data or experience that changed your mind.

SECTION 7 - A/B test & measure prompts (prompts 31-36)

The last six prompts are for the after part of title writing. Most writers pick one title, hit publish, and never look back. That’s a huge mistake. A/B testing is how titles go from “pretty good” to “data-backed best.” These prompts help you design, run, and learn from title tests.

Prompt 31 - The 5-Title Generator

Purpose: Generate 5 strong candidate titles for the same article. Pick the best 2 and A/B test them. This is the simplest, most-used prompt in the entire library.

Multi-line prompt (paste into ChatGPT):

Article summary: [2-3 SENTENCES]
Audience: [AUDIENCE]
Primary emotion you want the reader to feel: [EMOTION, e.g. "curious",
"inspired", "validated", "slightly scared"]
Word I want to avoid: [WORD]

Generate 5 curiosity-gap Medium titles for this article:
  - each title uses a different formula (one list, one question,
    one story, one contrarian, one how-to)
  - each title is between 10 and 18 words
  - each title includes the audience in the first 4 words
  - each title has a different open-loop structure

For each title, output:
  - the formula used
  - the open loop
  - the predicted CTR strength (low/medium/high) and why

Example output (article = “How I rewrote my Medium title 11 times to get 6,200 reads”):

  1. For New Writers: I Rewrote My Medium Title 11 Times (The Final Version Did 6,200 Reads in 7 Days) - Story - open loop: which version won - predicted: high

Pro tip: Use the 5-title generator before you write the article, not after. The title you pick changes the article you write. If you pick a story title, you write a story. If you pick a listicle title, you write a list. Title first, draft second.

Prompt 32 - The Headline Analyzer Scorer

Purpose: Score your draft title against the major headline analyzers (Coschedule, Anyword, Sharethrough). Use this prompt when you don’t have access to the tools.

Multi-line prompt (paste into ChatGPT):

Score this Medium title against the major headline analyzer rubrics:

"[PASTE TITLE]"

Score 1-100 on each:
  - Coschedule style: balance of common/uncommon/Emotional words,
    length under 70 chars
  - Anyword style: predicted engagement, clarity, emotion
  - Sharethrough style: curiosity, relevance, value
  - Outbrain style: specificity, intrigue, actionability

For each score, list 2 specific changes that would raise the
score. Then output a revised title that scores 80+ on all four.

Example output (title = “How to Write Better Medium Titles”):

Coschedule: 42, Anyword: 51, Sharethrough: 39, Outbrain: 44. Revised: For Medium Writers: 9 Title Formulas That Doubled My Reads in 30 Days (Coschedule Scores Each One) - Estimated 84/82/79/81.

Pro tip: Coschedule, Anyword, and Headlime all use slightly different rubrics, but they all reward: specific numbers, emotional words, and clarity about the reader’s payoff. The closest single thing to a “title grader in ChatGPT” is this prompt.

Prompt 33 - The A/B Test Designer

Purpose: Design a 2-arm A/B test for your Medium title. The prompt outputs hypothesis, audience, success metric, and decision rule.

Multi-line prompt (paste into ChatGPT):

Article: [BRIEF DESCRIPTION]

I want to A/B test 2 Medium titles:
  - Title A: [PASTE]
  - Title B: [PASTE]

Output:
  - Hypothesis (one sentence: which will win and why)
  - Primary success metric (CTR, read-through, read time, claps)
  - Minimum sample size (assume [YOUR TYPICAL ARTICLE TRAFFIC]
    views per week)
  - Test duration needed to reach 95% confidence
  - Decision rule (when do I pick the winner)
  - 1 follow-up test to run with the loser
  - 2 things to keep constant (image, subtitle, tags, publication
    time)

Example output:

Hypothesis: Title B will win because the specific number (5×) creates a stronger curiosity gap than the general claim in Title A. Primary metric: CTR. Min sample size: 1,400 impressions per arm. Test duration: 7 days. Decision rule: declare a winner only after 95% confidence. Follow-up: test the winning title with a different image.

Pro tip: A/B testing on Medium is harder than on Substack or WordPress because Medium doesn’t have native A/B testing for titles. Workaround: publish, wait 24 hours, then change the title and wait 7 more days. Compare CTR from each window. Yes, it’s noisy. Yes, it’s still better than guessing.

Prompt 34 - The 25-Headline Upworthy Sprint

Purpose: Inspired by Upworthy’s famous “write 25 headlines per story” rule (cited in Buffer’s headline guide). Generate 25 curiosity-gap titles for the same article, then have ChatGPT pick the top 3.

Multi-line prompt (paste into ChatGPT):

Article summary: [2-3 SENTENCES]
Audience: [AUDIENCE]
Article's biggest value: [THE ONE-SENTENCE TAKEAWAY]

Generate 25 curiosity-gap Medium titles for this article. Constraints:
  - 10-18 words each
  - each one must use a different formula, opener, or angle
  - no two titles should be paraphrases of each other
  - include at least 3 list, 3 question, 3 story, 3 how-to, 3
    contrarian, and the rest wildcards
  - the audience appears in the first 4 words of at least 15 of
    them

After the 25, rank your top 5 and explain in 1 sentence each why
they out-crowd the rest.

Example output:

Top 3 picks: #4 (story frame, specific number), #11 (curiosity-gap hidden number), #19 (contrarian + dollar). All three pass the 4 Us test (Useful, Urgent, Unique, Ultra-specific).

Pro tip: The Upworthy 25-headline rule works because volume forces variety. ChatGPT often defaults to the same formula 5 times in a row. Force it to spread across 5+ formulas. You’ll end up with 2-3 titles you never would have written by hand.

Prompt 35 - The Tagline Pair Maker

Purpose: Pair your title with a killer Medium subtitle / deck. The deck shows up in feeds and previews; many readers skim the deck more than the title.

Multi-line prompt (paste into ChatGPT):

Title: [PASTE TITLE]
Article summary: [2-3 SENTENCES]

Generate 10 Medium decks (subtitles) that:
  - sit between 8 and 18 words
  - add new information the title doesn't have
  - use a different formula than the title
  - the deck + title together read like a single 2-line thought
  - the deck is the "what's in it for me" line the title
    couldn't be
  - the deck can be a question, a promise, a number, or a
    mini-confession

Rank the top 3 and explain in 1 sentence why they pair best.

Example output (title = “I Posted 47 Medium Articles in 6 Months”):

Top deck: Here’s the $214 check, the 12 things that didn’t work, and the 1 thing I wish I’d done on day 1. - Pairs because the title promises volume, the deck promises honesty.

Pro tip: A great title + great deck is the most underused lever on Medium. The deck is your second hook. If the title pulled the click, the deck pulls the scroll. Use them as a pair, not as duplicates.

Prompt 36 - The Post-Publish Learn Logger

Purpose: After you publish, log the title, the result, and the lesson. Over 6 months this log becomes your private database of what works in your voice.

Multi-line prompt (paste into ChatGPT):

I just published a Medium article.

Title: [PASTE TITLE]
Date: [DATE]
URL: [PASTE URL]
Impressions at 7 days: [NUMBER]
Reads at 7 days: [NUMBER]
Read ratio: [PERCENT]
Claps: [NUMBER]
Best-performing tag: [TAG]
Publication (if any): [NAME]

Step 1: Diagnose what the title did well.
Step 2: Diagnose what the title did poorly.
Step 3: Predict which 2 of these 4 changes would most improve
  the next title:
    a) Add a specific number
    b) Add a contradiction or contrarian angle
    c) Shorten by 3-5 words
    d) Add a story / first-person frame
Step 4: Rewrite the title using those 2 changes.
Step 5: Save this as a row in my "title performance log" with
  these columns: Date, Title, Reads, Read Ratio, Top Change,
  Next Title.

Example output:

Top changes: (a) and (b). Revised title: For New Medium Writers: 47 Articles, $214, and the 1 Mistake That Cost Me $4,000 (I’ll Never Make It Again).

Pro tip: The post-publish log is the single biggest unlock for title writers. Most writers publish and forget. The 1% who log 30+ titles over 6 months end up with a personal playbook no one else has. That’s the edge.

Comparison table: prompt categories vs. title type vs. expected output

Prompt sectionWhat it does bestTitle type producedBest Medium contextPair it with
1 - Topic mining (1-5)Finds anglesIdea → title seedStuck-on-draft situationsCoschedule Headline Analyzer
2 - Numbers & lists (6-10)Builds trust via specificityListicles with a twistTech, AI, productivity, businessAnyword
3 - Questions (11-15)Pulls CTR via second-person framingQuestion → statement titlesSelf-improvement, career, moneyA/B test 2 versions
4 - Story (16-20)Builds parasocial trustConfession, transformation, experimentPersonal essay, creator economy, freelancingSubstack cross-post
5 - How-to (21-25)Converts intent to actionStep-ladder, time-boxed, tool-stackTutorials, beginner contentNotion checklist in body
6 - Contrarian & AEO (26-30)Citable in AI search”Everyone is wrong”, best-of, studySEO traffic, evergreen postsAEO schema markup
7 - A/B & measure (31-36)Picks the winnerThe pick + the dataAny post, any nicheGoogle Search Console

This is the matrix I keep open in a second tab. Pick the row that matches your post’s intent. Run 1-2 prompts from that row. A/B test. Log. Repeat.

People Also Ask - FAQ on Medium curiosity-gap titles

1. What is a curiosity-gap headline, and does it still work in 2026?

A curiosity-gap headline opens a specific question the reader’s brain wants to close, without giving away the answer. It works because of Loewenstein’s information-gap theory (1994), which says the gap between what we know and what we want to know creates a kind of mental itch (Wikipedia - George Loewenstein). In 2026, it works better than ever because AI search engines (Google AI Overviews, ChatGPT, Perplexity) literally quote headlines that have clear factual gaps in them. If you frame the gap well, both human readers and AI engines pull your title.

2. How long should a Medium title be in 2026?

The sweet spot is 10-18 words, with a hard ceiling around 70 characters to keep it from being truncated in feeds. Buffer’s headline research, drawing on KISSmetrics, found that readers “absorb the first three words and the last three words” of a headline (Buffer, 30+ Headline Formulas). So front-load value, finish with a flourish, and keep the middle tight.

3. Are question-mark headlines bad on Medium?

Not always. The 2013 Lai & Farbrot study cited in Wikipedia’s Betteridge’s law entry found that questions addressing the reader in second person outperform statements. Betteridge’s law is right that “is X true?” headlines usually mean “no” - but “Are you doing X wrong?” is a different beast. That title gets a click because the reader wants to know if they’re the one doing it wrong. So: convert most questions to statements. Keep second-person “are you” questions.

4. Can I use ChatGPT to write Medium titles without losing my voice?

Yes, but you have to train the prompt. Feed ChatGPT 3-5 of your past titles that did well and ask it to “match this voice, structure, and rhythm.” The 36 prompts in this guide force the structure. The voice comes from your 3-5 example titles pasted in as style references. If you skip the style reference, ChatGPT defaults to generic blog SEO tone - which is the AI-tell most readers can smell a mile away.

5. What’s the best title analyzer tool for Medium writers?

Three stand out in 2026:

  • CoSchedule Headline Analyzer - best for structure, length, and emotional word balance.
  • Anyword - best for predicted CTR and audience matching.
  • Headlime - best for AI-suggested rewrites of your draft title. Pair any of them with prompt 32 above for a self-contained scoring workflow.

6. Does Medium’s Partner Program penalize “clickbait” titles?

It penalizes AI-generated content, not titles. Per Wikipedia’s Medium entry, since May 1, 2024, AI-generated content is blocked from paywalls and Partner Program payouts. A curiosity-gap title is fine, but the body must be human-written, original, and on-topic. The title sets the promise. The body keeps it. If the body is thin or off-topic, the title burns you in the algorithm.

7. How many times should I rewrite a Medium title before publishing?

Aim for 5-10 rewrites per article. The Upworthy 25-headline rule (cited in Buffer’s guide) is the gold standard: write 25, pick the top 3, A/B test 2. If you don’t have time for 25, do 5. If you do 25, you almost always land on a title you’d never have written on the first pass.

8. How do I know if my curiosity-gap title worked?

Track three numbers at 7 days:

  • CTR (reads ÷ impressions)
  • Read ratio (read-through ÷ reads)
  • Claps per read (claps ÷ reads) A “good” curiosity-gap title typically pulls a CTR 20-50% higher than your baseline, holds a read ratio above 60%, and gets more claps per read because readers feel the title promised something the body delivered. If CTR is high but read ratio is low, the title oversold. If CTR is low but read ratio is high, the title undersold. Log every test and tune from there.

9. Can I use these prompts for Substack, Beehiiv, WordPress, or Ghost?

Yes. The 36 prompts work for any platform with a title field. They work especially well for Substack and Beehiiv subject lines, which are functionally titles. For WordPress and Ghost, the same prompts drive both the H1 and the SEO title tag. The only platform-specific tweak is character count: Substack subject lines cap at around 60 characters, Beehiiv around 55, Medium around 70 visible before truncation. Keep your 10-18 word titles under those caps.

10. How do I avoid AI-tells in titles ChatGPT writes?

Three rules. First, never let ChatGPT start a title with “The” + abstract noun (“The Future of X”, “The Truth About Y”). Second, ban phrases like “in 2026 you need to” or “here’s why” from your final title. Third, paste in 3-5 of your own past titles and tell ChatGPT to match that voice. AI-tells live in the default. They vanish the moment you force a style reference.

A 14-day “30 titles” sprint

I built this sprint for one specific outcome: by the end of 14 days, you have 30 tested, logged, ranked Medium titles in your personal playbook. No fluff. No theory. Just reps.

Days 1-2: Setup

  • Pick 5 of your past Medium posts that under-performed.
  • Pull their titles, impressions, reads, and read ratios from Medium stats.
  • Run each through prompt 32 (the headline analyzer scorer). Log scores.

Days 3-5: Generate

  • For each of the 5 posts, run prompt 31 (the 5-title generator).
  • Pick the top 2 candidates per post. You now have 10 candidate titles.

Days 6-10: Test

  • Edit and re-publish each post with a new title. (Yes, you can edit a Medium title post-publish.)
  • Wait 48 hours. Log CTR, read ratio, claps.
  • Keep the winner. Restore the loser to the original or a new title.

Days 11-13: Mine

  • For each winning title, run prompt 34 (the 25-headline sprint) to find 3 more variations.
  • A/B test those variations across 3 new posts.

Day 14: Log

  • Compile your 30 tested titles into a single Notion or Google Sheet.
  • Columns: Date, Original Title, New Title, CTR Before, CTR After, Read Ratio, Top Lesson.
  • This sheet is now your private title playbook.

You will never need to write a Medium title from scratch again. You’ll start with 30 data points and a known-good formula stack.

Common mistakes to avoid

  • Mistake 1: Writing the title last. Write the title first, then write the article that delivers on it. If you write the article first, you end up with a summary title. If you write the title first, you end up with a focused article.
  • Mistake 2: Using “How to” for non-tutorials. “How to find your voice as a writer” is vague. “How I rewrote 47 Medium titles to get 6,200 reads” is specific. The word “how” only works when the body has a real mechanism to teach.
  • Mistake 3: Burying the audience. If your reader has to scan past 8 words to figure out who the post is for, they bounce. Put the audience in the first 4 words, every time.
  • Mistake 4: Trusting the first 3 ChatGPT outputs. Always ask for at least 5, ideally 10-25. The first 3 are ChatGPT’s “obvious” answers. The magic lives in options 7 through 15.
  • Mistake 5: Skipping the A/B test. A title that “feels” strong is often wrong. A title that wins an A/B test is always right. Even a noisy 7-day test beats a confident guess.
  • Mistake 6: Promising what the body can’t deliver. A title that says “5× your reads in 7 days” requires a body that explains exactly how, with data. If you can’t back the claim with a chart, a story, or a framework, soften the title. A deliverable title compounds trust. A non-deliverable one burns it.
  • Mistake 7: Using question marks when statements work. Betteridge’s law is real, and Wikipedia’s entry walks through the history. Convert most “Is X true?” headlines into “What happens when X” statements. You’ll get the same click without the “the answer is no” baggage.
  • Mistake 8: Forgetting AI search. In 2026, your title shows up in Google AI Overviews, ChatGPT search, Perplexity, and Bing Copilot. A title that contains a clear factual claim (“the best X in 2026 is Y”) gets cited directly. A title that’s pure poetry doesn’t. Optimize for both humans and machines.
  • Mistake 9: Copying the format, not the structure. When you admire a great title, copy the structure, not the words. “Brutal math behind $4,200” works because of confession + number + dollar. Steal the frame, swap the substance.
  • Mistake 10: Not logging. You will repeat every mistake you don’t log. The 5-minute post-publish log (prompt 36) is the single highest-ROI habit in title writing. Do it for 30 posts and you will never go back.

Final word

The title is the only sentence on Medium that competes for attention. Everything else - your hook, your structure, your claps, your read time - depends on whether that one sentence earned the click.

The 36 prompts above are a complete workflow. They mine topics. They build lists, questions, stories, how-tos, contrarians, AEO titles, and A/B tests. They give you ChatGPT a structure to fill in, not a blank page to freestyle. Structure + voice + data = titles that ship, scale, and compound.

Pick five drafts. Run them through prompts 1, 6, 11, 16, and 21. A/B test the top two. Log the result. Do it again next week. Within a month, you’ll have a personal title playbook no one else has. That’s the edge.

Now go write the title the article deserves.