Stop Killing Your Masterpieces With Amateur Headlines
Chuck instinct. Let insights from 1000+ titles guide you.
I was fed up with seeing my best ideas die quiet deaths beneath weak headlines.
Something had to be done. I knew I needed to understand what really makes readers click. Not just gut feelings or best practices — but real, data-backed insights.
So I did something a bit obsessive: I analyzed 1,016 Medium article titles and subtitles. The results surprised me.
Most beginner writers craft the titles last, treating it like an afterthought.
Huge mistake.
The title isn’t just a label. It’s the roadmap — the spin we decide to give an article and the emotions we want to evoke in the reader.
It’s the difference between an article that sinks to the bottom of the content pit and one that goes viral.
In this article, you’ll discover:
Which content categories perform best (and which popular ones are secretly failing)
The emotional triggers that drive engagement
A custom AI prompt to generate title ideas
Click here to jump straight to the key findings.
For the data enthusiasts who want to understand the research methodology (and why you should trust these results), read on to understand how I approached this analysis.
Creating a “Success Score” To Compare Articles
The Data Collection
First, let me share how I gathered the data. I scraped 1,016 Medium titles and subtitles from articles published between 1-6 days ago.
Why this specific timeframe?
One day minimum: This gives articles enough time to gather initial engagement. Articles need at least 24 hours to reach their core audience, receive shares, and collect meaningful feedback.
Six day maximum: I capped it at 6 days because:
Medium's homepage and topic pages primarily feature content from the past week
Most articles receive the bulk of their engagement in the first week
Longer timeframes would introduce too many variables (like changes in Medium's algorithm or trending topics)
By focusing on this 1-6 day window, I could compare articles that:
Had enough time to prove their worth
Were competing for attention under similar conditions
Faced the same Medium algorithm and distribution patterns
The Comparison Challenge
Next, to compare headlines, I needed a score against each. But that wasn’t simple.
How do you fairly compare
A day-old piece with 800 claps and 20 comments
A 5-day-old piece with 2500 claps and 40 comments
Simply comparing the number of claps or comments each generated would be unfair on the newer article.
Comparing the engagement they get per day or per hour is also flawed since visibility decreases exponentially over time — newer articles get more eyeballs.
Also, not all engagement is equal. Someone taking time to write a comment shows deeper connection than a quick clap.
The Solution: Composite Performance Score (CPS)
To solve this, I developed a Composite Performance Score (CPS). Think of it like a “viral potential detector” that:
Weighs comments more heavily than claps (60% vs 40%), as comments represent deeper engagement.
Adjusts for an article’s declining visibility with age
Combines these factors to predict an article’s true performance. It provides a fair way to compare a 5-day-old viral hit with a 16-hour-old rising star.
Understanding Declining Visibility
When an article is published on Medium, it initially gets a lot of attention. It’s featured on homepages, sent to subscribers, and generally has a spotlight shining on it.
However, as time passes, newer articles take that spotlight, and this article naturally gets fewer views. This drop in visibility isn’t a steady decline; it happens quickly at first and then slows down over time.
This pattern is known as exponential decay.
Think of exponential decay like this: Imagine you have a hot cup of coffee. It cools down rapidly when it’s piping hot, but as it gets cooler, the rate at which it loses heat slows down.
Similarly, an article’s visibility drops sharply soon after it’s published and then decreases more gradually.
The Concept of Half-Life
To measure how quickly something decreases, scientists use the idea of a half-life. It’s the time it takes for something to reduce to half its initial amount.
In our case, the half-life is the time it takes for an article's visibility to drop to half of what it was when first published. For my research, I considered half life to be 24 hours.
Calculating the Composite Performance Score (CPS)
Intitially I arrived at a combined score of comment counts and claps using the formula:
Then, to arrive at the CPS, I used the idea:
If an article has been published long enough that it has already received 80% of its lifetime exposure, and got a combined score of 100, the composite score would be 100/0.8 = 125.
If a newer article has received only 20% of its share of exposure and got a combined score of 50, its composite score would be: 50/0.2 = 250.
This means we increase the CPS of newer articles to account for the fact that they haven’t had as much time to gather claps and comments.
To find out how much exposure an article has left at a certain time, I used the half-life concept:
Exposure Remaining = 50% after 1 half-life, 75% after 2 half-lives, etc.
But I needed a formula to calculate this at any time, not just at half-life points.
This is the formula I used to calculate the exposure already used:
Where:
Final equation:
Okay, enough of the dizzying formulae. Let’s see CPS in action:
Article A (16 hours old): 800 claps, 20 comments → CPS = 1,605
Article B (5 days old): 2500 claps, 40 comments → CPS = 1,244
Despite having fewer raw numbers, Article A scores higher because it shows stronger engagement in a shorter time.
Categorizing the Article Titles
To understand the type of titles that perform better, I mapped each to one or more of the following categories:
How-To: Instructional guides that promise to teach.
Listicle: Numbered lists.
Personal Story: Narratives that share personal experiences.
Question: Titles that pique curiosity with a query.
Success Story: Celebrations of achievements and victories.
News Update: Timely insights on recent events.
Opinion: Personal viewpoints and critiques.
Secret Reveal: Insider knowledge or truths.
Call to Action: Urgent prompts encouraging immediate action.
The Emotional X-Factor
Beyond just counting numbers, I tracked how each title/subtitle combination made the reader feel. I looked for six core emotions: anger, disgust, fear, joy, sadness, and surprise.
I tagged articles that didn’t strongly trigger any of these emotions as ‘neutral’ — though spoiler alert: the neutral ones rarely won the race.
This emotional fingerprinting revealed interesting patterns about what makes readers click and engage.
The Winners Circle: Top-Performing Article Categories
News Updates (560 Average CPS)
Only 37 out of 1,016 articles (3.6%) used news as their primary angle. But these rare news-focused pieces were disproportionately successful.
They had the highest average engagement score among all categories
A stunning 59% of them reached the top 30% of all content
26% broke into the elite top 10%
Well-timed articles riding current events or trends have significantly higher chance of going viral.
To put this in perspective: Out of 100 articles published, only about 4 might be news-focused.
But those 4 articles have better odds of going viral than any other type of content. Almost 2 of those 4 would likely hit the top 30%, and 1 would have a solid shot at the top 10%.
Best performing article in this category:
How Donald Trump Won the 2024 Election
And analyzing how Kamala Harris lost
Score: 1515.48
Claps: 3500
Comments: 153
Hours since published: 144
Personal Stories (375 Average CPS)
Personal Stories are the second-most popular among all categories, behind opinion pieces, making up 23.5% of all articles (239 out of 1,016).
About one-third of personal stories (34.8%) made it to the top 30%, and those that do perform exceptionally well.
The average score for a Personal Story in the top 30% is 820.63 — higher than many other categories.
When they make it to the top 10%, they average an impressive 1,345.82 points.
In fact, the highest-scoring article in the entire dataset was a Personal Story titled “I Got A Proposal” which scored 2,612.05 points — nearly double the average top 10% score.
This suggests that personal stories are a high-risk, high-reward format. The majority perform modestly, but those that resonate with readers can achieve unprecedented engagement levels that other formats rarely reach.
Best performing article in this category:
I Got A Proposal
I'm Getting Married 😊
Score: 2612.05,
Claps: 5900,
Comments: 148,
Hours since published: 96
Surprising Underperformers
Some traditionally “safe” content types performed worse than expected:
How-to Articles (156 Average CPS)
Less than 12% reached the top 30%
Only 2% made it to the top 10%
Lower average engagement despite being common
Generic how-to content is oversaturated. Unless you have a truly unique angle, consider other formats.
Listicles (147 Average CPS)
Lowest average performance among major categories
Only 8% reached the top 30%
None made it to the top 10%
The classic “5 Ways to...” is losing its effectiveness. Readers want deeper, more nuanced content.
The Full Breakdown
Here’s the full breakdown of the performance of all the categories:
The chart shows how different types of articles are distributed across Medium. For each category (like Opinion, Personal Story, etc.), there are two bars:
The green bar shows the percentage of articles where this category was the main focus
The purple bar shows the total percentage including articles where the category appeared either as main or secondary focus
For example, looking at Opinion pieces:
23.4% of articles were primarily opinion pieces (green bar)
This rises to 35.6% when including articles that used opinion as a secondary element (purple bar)
These bar charts show how different categories perform in terms of reaching top engagement levels on Medium. Each category has two measurements:
Green bars show performance when the category is the primary focus
Purple bars show performance when the category appears as either primary or secondary
The first graph shows a broad view of success (top 30%), where News Updates dominate — 59.3% of primarily news-focused articles achieve this benchmark. Success Stories and Personal Stories follow, with around 35% of articles reaching this tier.
The second graph tracks articles reaching the elite top 10%. News Updates maintain their lead — 25.9% of news-focused pieces hit this elite level. Personal Stories and Questions follow at around 13% each.
Notably, traditional formats like How-to guides and Listicles consistently rank at the bottom in both metrics, with Listicles showing 0% success in reaching the top 10%.
The Emotional Factor: What Makes Readers Engage
The data revealed a surprising truth about emotional engagement: Strong negative emotions often drive more engagement than positive ones.
Here’s the breakdown:
Disgust leads the pack
A stunning 50% of articles expressing disgust reach the top 30%
This is twice the success rate of joyful content (25%)
However, only 8.3% make it to the top 10%, suggesting it's good for broad appeal but rarely creates viral hits
Sadness and Surprise show strong performance
37% of sad articles reach the top 30%
34% of surprising content breaks into top 30%
Sadness maintains its power in the top 10% (12.9%)
Joy surprisingly underperforms
Only 24.9% of joyful content reaches the top 30%
A mere 7.1% makes it to the top 10%
This is the lowest performance among all emotions
Readers engage more deeply with content that challenges them emotionally than content that simply makes them feel good. The most successful articles often tap into complex emotions like disgust at societal issues or sadness about shared human experiences.
This doesn’t mean you should be negative — but rather that addressing difficult truths and complex emotions tends to create stronger reader connections than purely uplifting content.
Practical Applications: How to Use This Data
Now that we understand what drives engagement, let’s put these insights to work.
We’ll take a topic and map it to the top performing categories and evoke a range of emotions using ChatGPT/Claude.
This prompt turns even the driest subject into 9 potentially viral headlines.
The idea here is to generate different ways of thinking about the same overall content we want to produce.
Here’s the prompt:
You are tasked with generating nine (9) compelling title and subtitle combinations for an article based on a given topic. The goal is to create titles that maximize click-through rates by aligning with specific categories and evoking designated emotions.
Instructions:
Input: You will receive an article topic from the user.
Output: Generate nine (9) unique title and subtitle pairs according to the following category and emotion combinations:
Category 1: News Update
Emotion A: Anger
Emotion B: Fear
Emotion C: Sadness
Category 2: Personal Story
Emotion A: Anger
Emotion B: Fear
Emotion C: Sadness
Category 3: Question
Emotion A: Anger
Emotion B: Fear
Emotion C: Sadness
Guidelines for Each Category:
News Update:
Style: Present the topic as a recent or breaking news story.
Tone: Objective yet engaging, suitable for a news headline.
Emotion Evocation: Use language that elicits the specified emotion in the reader regarding the news.
Personal Story:
Style: Frame the topic as a first-person narrative or personal experience.
Tone: Intimate and reflective, drawing the reader into the personal journey.
Emotion Evocation: Convey the specified emotion through the personal account.
Question:
Style: Pose a provocative or thought-provoking question related to the topic.
Tone: Inquisitive and engaging, encouraging the reader to seek answers.
Emotion Evocation: Use the question to stir the specified emotion in the reader.
Emotions to Evoke:
Anger: Use strong, assertive language that highlights injustice, wrongdoing, or frustration related to the topic.
Fear: Employ cautious or alarming language that points out threats, risks, or uncertainties.
Sadness: Utilize empathetic and sorrowful language that reflects loss, regret, or misfortune.
Title and Subtitle Requirements:
Title:
Should be attention-grabbing and concise (ideally under 70 characters).
Must clearly reflect both the assigned category and the emotion.
Should be directly related to the given topic.
Subtitle:
Provides additional context or a teaser that complements the title.
Should further evoke the assigned emotion.
Encourages the reader to click and read more.
Formatting:
Number each title/subtitle pair from 1 to 9.
Clearly indicate the category and emotion before each pair.
Present the title in bold and the subtitle in italics for clarity.
Are you ready?
Once you type this into the language model, it would say it’s ready and would ask you to input the topic you want to talk about. Go ahead and provide the input and wait for magic to happen.
Let’s see this framework in action. I tested it with an intentionally dry topic: the “Door-in-the-Face” compliance technique in psychology. Here’s what it generated:
Category 1: News update
News Update + Anger: “Manipulative Sales Tactics Exposed: Door-in-the-Face Scheme Preys on Consumers” Investigation reveals widespread abuse of psychological compliance techniques by major retailers, leaving customers feeling duped and exploited.
News Update + Fear: “Hidden Mind Games: Door-in-the-Face Tactics Could Be Controlling Your Decisions” Alarming study shows how this psychological technique silently influences millions of consumers daily, experts warn of growing manipulation crisis.
News Update + Sadness: “Lost Trust: Door-in-the-Face Technique Destroys Consumer-Business Relationships” Recent findings show lasting damage to customer faith as beloved brands admit to systematic manipulation strategies.
Category 2: Personal story
Personal Story + Anger: “I Was Trained to Trick You: My Life as a Door-in-the-Face Manipulator” Former sales executive reveals the gut-wrenching truth about how they were forced to exploit customer psychology.
Personal Story + Fear": “The Day I Realized Every 'No' Was Part of Their Plan” My chilling discovery of how salespeople used the Door-in-the-Face technique to control my purchasing decisions for years.
Personal Story + Sadness: “How I Lost $50,000 Falling for the Door-in-the-Face Trap” A heartbreaking journey through manipulation, regret, and the painful lesson that changed my life forever.
Category 3: Question
Question + Anger: “Is Your 'No' Just Being Used Against You?” Why the Door-in-the-Face technique is making a mockery of your right to refuse - and what you can do to fight back.
Question + Fear: “Could Saying 'No' Make You More Vulnerable to Manipulation?” The terrifying psychology behind why rejecting the first offer might be exactly what they want you to do.
Question + Sadness: “When Did We Stop Trusting Each Other?” How the Door-in-the-Face technique became a symbol of our lost faith in genuine human connection.
Each of these provides a unique angle to the same topic. Once I find something I like, I can go ahead and research to craft the final piece.
You can experiment with the prompt to try different categoriey/emotion combinations.
There’s no such thing as a “boring” topic — only boring presentations.
Even a psychological compliance technique can become a compelling story when you understand the elements that drive engagement.
Want personalized help implementing AI in your writing process?
I offer 90-minute sessions where we build your custom AI content system together.