The headline examples that win A/B tests in our database share one trait: they tell a visitor exactly what they get, in the visitor's words, within the first few words.
Movexa added the single word "supplement" to a landing page headline and purchases rose 89.97%; Infinite Age swapped wellness slogans for an outcome-led headline and completed purchases rose 35%.
Clever, branded or vague headlines are the ones that lose.
This page covers how to write a headline, the formulas that keep showing up in winning tests, and how to test headlines without fooling yourself.
For the full list of headline and copy results, see the headlines and copy element hub.
Headline examples from real A/B tests
Every row below is a test from LiftArchive, with the control and variant headline as the source described them. Some are translations; we note where.
| Control headline | Winning headline | Result | Ran by |
|---|---|---|---|
| "Dominate the battlefield with speed & precision" | "The best damn VR gunstock on the planet" | +$19,441/month revenue | Convertibles |
| "A step-by-step plan for opening your internet store" (translated from Russian) | "Open your internet store in one evening" | +9.52% leads | Imsider.ru |
| "Could You Use a Free Plan For Promoting With Social Media?" | "Yours is the only newsletter that I actually read." | +24.31% sign-ups | Laura Roeder Social Media |
| "Men's Clothing" | Roughly "Order men's clothing easily for bargain prices" (translated from Finnish), plus a short intro text | +127% conversion | Tulos |
| Product title as a label, e.g. "LED Face Mask" | "Improve Your Skin from Home with the LED Face Mask" | +12.36% revenue per visitor | Dave Diederen |
| Movexa headline that didn't say what the product was | Same headline with "supplement" added | +89.97% purchases | Source Patrol |
Look at what the winners did. Not one of them got cleverer.
They named the product, set a time frame, used a customer's own sentence, or stated the offer.
The VR gunstock control sounds more like an ad, and it lost to a line that simply says what the thing is.
@oliverkenyon
Vasta
Convertibles
SplitBase
@kasper_dolkHow to write a headline
The headline carries more weight than any other line on the page.
David Ogilvy put it this way in Confessions of an Advertising Man: "On the average, five times as many people read the headline as read the body copy" (excerpt via Campaign).
That was print advertising in 1963, but the web makes the gap wider, not smaller.
Online, people scan. Jakob Nielsen's 2009 eye-tracking analysis found that users typically see about the first 2 words of list items and links before deciding whether to keep reading.
Put the words that carry the meaning first.
Here is the process we see behind most winning headline tests.
- Start from research, not a blank page. SplitBase found that BestSelf's visitors were not scrolling past or understanding the Self Journal's above-the-fold copy. The new headline spoke to "professionals" and named the core benefit of reaching goals, and Self Journal sales rose 27%.
- Say what the product is. If a visitor from an ad can't tell what you sell within a few seconds, nothing else on the page matters. That is the whole lesson of the Movexa and VR gunstock tests.
- Lead with the outcome the customer wants. Kasper Dolk changed one product hero headline to focus on the ideal customer's desired outcome and reported +$823,115.54 revenue per year. This is the headline version of leading with benefits over features.
- Use their words. Pull phrases from reviews, support tickets and survey answers. Laura Roeder's team took a reader's comment and made it the headline.
- Match the traffic source. A headline that repeats the promise of the ad or email that sent the visitor removes doubt that they landed in the right place. See ad message match.
- Cut the brand voice if it hides the offer. Oliver Kenyon replaced a branded hero headline with a bold one about the offer and saw conversion rate go from 0.9% to 3.6%. That was a before-and-after comparison, not a split test, so treat the size of the jump with care.
The headline also needs a job on the page.
A headline in the hero section states the value proposition.
A headline above a guarantee or review block summarizes that block.
Dave Diederen rewrote a money-back guarantee section with a bold headline and benefit-driven copy and reported +12.71% revenue per visitor.
Headline formulas that keep winning
A formula is a starting shape, not a guarantee.
These seven come from patterns that repeat across our tests, with an example of each.
If you need a fuller system for the rest of the page, see copywriting frameworks.
| Formula | Shape | Test that used it |
|---|---|---|
| Outcome first | [Desired result] + [product as the way to get it] | Infinite Age, +35% purchases (Vasta) |
| Name the thing | [Product category] + [why it's the best choice] | VR gunstock, +$19,441/month (Convertibles) |
| Time-bound promise | [Result] in/within [short time] | "in one evening", +9.52% leads; "Get coverage within 24 hours" beat a 28.88% step-1 baseline (KlientBoost) |
| Audience call-out | [Benefit] for [specific group] | BestSelf, +27% sales (SplitBase) |
| Customer quote | A real sentence from a customer | +24.31% sign-ups |
| Offer up front | [What you get] + [price or deal] | Category page, +127% conversion (Tulos) |
| Outcome + objection tagline | Outcome headline, then a tagline that answers the #1 objection | ~+10% conversion across clients (Kasper Dolk) |
The outcome + objection pattern is worth a closer look.
Kasper Dolk's reasoning is that the visitor locks onto what they want before they see the price, so the product reads as the way to get it.
The objection tagline then handles the main reason they would leave.
Your headline should also carry your positioning. If you can't write the "name the thing" or "outcome first" version, the problem may be the unique selling proposition, not the wording.
What about negative or curiosity headlines?
Research on news headlines points one way: a 2023 study of Upworthy's randomized headline tests found that, for a headline of average length, each additional negative word raised click-through rate by 2.3%, while positive words lowered it. That is a real, pre-registered result.
But it measured clicks on news stories, not purchases or sign-ups.
A headline that wins the click can still lose the sale if the page doesn't pay it off.
Our store tests mostly reward clarity over drama, and curiosity headlines in our database tend to appear on advertorials and listicles, where the job is to get a cold visitor reading.
Headline testing: how to run it without fooling yourself
Headline testing means showing visitors different headlines at random and comparing a business metric, such as purchases or revenue per visitor.
It is one of the cheapest tests to build, which is why it's often the first test a team runs.
The process is the same as any A/B test.
Test angles, not word swaps
Dave Diederen's advice after seeing +18% conversion rate from a new listicle title: run your current headline against 2-3 variations that each take a different angle, such as benefit-led, curiosity-led and social proof-led.
Changing "great" to "amazing" rarely moves anything a test can detect.
Running several variants at once is an A/B/n test.
Each extra variant splits your traffic further and raises the chance of a false winner, so adjust for multiple comparisons.
Work out the sample size first
Headline changes usually produce small lifts, and small lifts need a lot of traffic.
The standard per-variant sample size for comparing two conversion rates is:
n = (z_alpha/2 + z_beta)^2 x [p1(1 - p1) + p2(1 - p2)] / (p2 - p1)^2
Worked example: your page converts at 3%, and you want to detect a 10% relative lift (3.0% to 3.3%) at 95% confidence (z = 1.96) and 80% power (z = 0.84).
- (1.96 + 0.84)^2 = 7.85
- 0.03 x 0.97 + 0.033 x 0.967 = 0.0610
- (0.003)^2 = 0.000009
- n = 7.85 x 0.0610 / 0.000009 = about 53,200 visitors per variant
Detecting a 15% lift (3.0% to 3.45%) needs about 24,200 per variant.
If your page gets 5,000 visitors a month, a headline test sized for a 10% lift would take most of two years.
That is the wrong choice: test bigger changes, or test on a higher-traffic page.
Use the sample size calculator to run your own numbers.
Pick the metric that pays
Kasper Dolk notes that an outcome headline may lower add-to-cart rate but lift revenue.
Judge headline tests on purchases or revenue per visitor, not clicks or scroll depth.
Expect flat results and some losses
Not every rewrite wins. Clean Commit tested short benefit statements in a homepage hero: the variant showed +7.7% conversion rate but only a 58.7% probability to beat control, so it counted as inconclusive.
On four product pages, the same agency's refined layout with rewritten value-focused copy lost 20.7% conversion rate, though that test had a small sample of 1,037 visitors and bundled several changes.
Also read headline case studies with some skepticism.
Several winners above were one of multiple variants, and the source reported only the best one.
Infinite Age tested two headlines and Tulos tested several, so part of each reported lift may be luck.
Where headline testing is the wrong tool
- Low-traffic pages. Use customer research and a five-second test instead, then ship the clearest version.
- When the offer is the problem. No headline fixes a product people don't want at that price.
- When you change the headline and everything else. Melbourne Innovation Centre changed the H1, supporting text and button together and saw +8.56% applications. That's a fine result, but you can't credit the headline alone.
Once the headline is settled, the same rules apply to the rest of the page.
Our guide to landing page copy covers the sections below the fold.
You can browse every result in the test database, or go back to the copywriting hub.