A headline analyzer that scores your headline against real A/B tests and shows what helps it and what hurts it.
Paste your headline, or several, 1 per line. It's scored by a model trained on real A/B tests, and you see which parts helped or hurt in those tests.
Where it's shown
Your headline
1 headline per line. Several lines are ranked, best first, so you can paste in a list from your AI and pick the winner.
The score
Type or paste a headline to score it.
What you type stays in this browser. The scoring runs on your device.
This free headline analyzer scores your headline with a model trained on real A/B tests.
It learned from 1,913 tests where the same story ran with different headlines. It tells you which parts of yours helped or hurt in those tests, and how often it's right.
Here's what a headline analyzer is, what the tests found, and how to write headlines that win.
What is a headline analyzer?
A headline analyzer scores a headline before you publish it, and shows what to change.
Most score against rules someone wrote. This one scores against what won: real headlines that ran against each other, with the clicks counted.
How the score works
The analyzer measures 19 parts of your headline: its length, numbers, questions, "I" and "you", punctuation and more.
A model weighs each part by how much it moved clicks in the tests. Your score says where your headline lands among every headline in them: 70 means it beats 70% of them.
How accurate it is
The model trained on 1,530 tests and was then graded on 383 tests it never saw.
In those, the headline it scored higher won 64% of 859 head-to-heads that had a clear winner. A coin flip gets 50%.
That's a strong hint, not a guarantee. Your own numbers are the real test.
What the tests found
In head-to-heads with a clear winner, this is how often a headline with each part beat the same headline without it:
| Part | Won against the same headline without it | Head-to-heads |
|---|---|---|
| A number | 64% | 987 |
| "This", "these" or "here's" | 61% | 1,236 |
| "I" or "my" | 60% | 673 |
| Starting with a number | 58% | 289 |
| A superlative like "best" or "ever" | 57% | 486 |
| An emotional word like "amazing" | 57% | 357 |
| A negative word like "never" or "stop" | 51% | 1,111 |
| "You" or "your" | 51% | 1,105 |
| A word in capitals | 47% | 258 |
| "We" or "our" | 43% | 529 |
| A colon, dash or bar splitting it in 2 | 38% | 357 |
| Opening with how, why or what | 35% | 570 |
| A question | 35% | 923 |
| An exclamation mark | 34% | 123 |
The surprise is questions. They're common advice, and they lost almost 2 out of 3 times.
5 patterns that won
The analyzer suggests the rewrites your headline is missing, ordered by what the tests back:
- Add a specific number. "7 ways" beats "ways".
- Tell it as your story. "How I…" beat the same line without "I".
- Point at the answer. "This is how…" and "Here's…" won more than they lost.
- Answer instead of asking. Turn the question into the statement it's hinting at.
- Say it in 1 part. A colon or dash splitting the headline in 2 lost most of the time.
Headlines for each format
The analyzer checks your headline against where it's shown:
| Format | What it checks | Source |
|---|---|---|
 YouTube title | The 100-character limit, and a score from real channels' videos | YouTube Help, and 25,385 videos from 172 channels |
 LinkedIn first line | The cut before "… more": about 160 characters on a phone, 230 on a laptop | Measured in a browser |
 Articles | The parts that won or lost, and the score | The Upworthy tests |
 Landing pages | The parts that won or lost | The Upworthy tests |
 Emails | A score from 56,262 real emails, and the parts that won or lost | Public newsletters' emails |
Where the data comes from
The model is trained on the Upworthy Research Archive, published in Scientific Data in 2021 by J. Nathan Matias, Kevin Munger, Marianne Aubin Le Quere and Charles Ebersole. It holds the A/B tests Upworthy.com ran on its headlines from 2013 to 2015, with the clicks on every version.
I used the open exploratory set, kept only tests where the image and excerpt were the same, and left out the months the archive flags for a randomization problem.
That's 1,913 tests. The model trained on 1,530 of them and was graded on the other 383.
It's news headlines from 1 site, 10 years ago. So treat the score as a strong hint, and your own numbers as the real test.
How to test your headlines
A score picks the likely winner. A test proves it:
- YouTube: Test & compare in YouTube Studio tries up to 3 titles and thumbnails, and keeps the one with the most watch time.
- Email: send 2 subject lines to 2 halves of your list, and keep the one with more clicks.
- Landing pages: run 2 headlines to 2 halves of your visitors, and keep the one that signs up more people.
Headline Analyzer FAQs
Questions about the headline analyzer? Here's what to know.
A headline analyzer scores a headline before you publish it and shows what to change. This headline analyzer scores against real A/B tests, not rules someone wrote.
In real A/B tests, good headlines more often had a specific number, "I" or "my", or a pointer like "this" or "here's". Questions, exclamation marks and colons lost more often than they won.
Questions mostly lost in the tests the headline analyzer is trained on: a headline with a question won only 35% of 923 head-to-heads against the same headline without one.
Numbers worked in the tests: a headline with a number won 64% of 987 head-to-heads against the same headline without one.
The headline analyzer picked the winner in 64% of 859 head-to-heads with a clear winner, in 383 real A/B tests it never saw. A coin flip gets 50%.
The headline analyzer is trained on the Upworthy Research Archive, 1,913 headline A/B tests Upworthy.com ran from 2013 to 2015, published in Scientific Data in 2021.
The headline analyzer reads English headlines. In other languages it still measures length, numbers and punctuation, but not the word parts.
Yes, the headline analyzer is free, with no sign-up. The scoring runs on your device.
What you type in the headline analyzer stays in your browser, so it's there when you come back. Nothing is sent to a server.
The headline analyzer is for creators, writers and marketers who want to know which headline to publish before they publish it.
For writing, the X character counter counts your post the way X does, and the LinkedIn post preview shows where LinkedIn cuts your first line. See all my free tools.
Navid Moazzez
AI business strategist & AI OS builder
Navid Moazzez helps creators and founders master AI and build their own AI Operating System (AI OS) to automate their business and life.
Navid.me is reader-supported. When you buy through links on this site, I may earn an affiliate commission. Learn more.
More free tools
Bluesky
Apple Books
Amazon
Free AI newsletterThe most actionable AI newsletter for founders
Every week, get proven AI strategies, curated tools, and step-by-step systems to grow your audience, create better content, and build a profitable creator business.
No fluff, no filler, no BS. Just five minutes each week that might level up your online business and life.
P.S. Sign up now to get free access to my ultimate AI tools guide for creators.
