The Best Headlines for Advertising: How to Grab Attention in 1 Second, With Formulas, Length Limits per Traffic Source and a Testing Method

An ad headline has one job: make the right person stop for a second and want the next line. Not everyone, the right person. Most bad headlines fail because they try to be clever for everybody, and most good ones look almost boring on paper: one clear idea, said to one kind of reader, in words they'd use themselves, with a promise the next page actually keeps.

I started online as a freelance copywriter back in 2010, so headlines are the part of advertising I've been fiddling with the longest. The biggest lesson from all those years isn't a magic formula. It's that nobody, including me, can reliably guess which headline will win. What you can do is write better candidates, fit them to each traffic source, test them in a way that doesn't lie to you, and stay on the right side of the rules while you do it. That's what this guide covers.

It's written for affiliates and media buyers, so everything here is about headlines that get paid traffic to click and then convert. If you want the copywriting career side, I covered that in how to get paid for copywriting, and the page the headline sends people to has its own guide on building a high conversion rate landing page.

What makes an ad headline grab attention in one second?

A headline grabs attention when it does four things at a glance: it stops the scroll, it tells the reader "this is for you", it promises something specific, and it matches what happens after the click. Miss any one of the four and you either get ignored or get clicks that don't pay.

• Stop. Something in the first three or four words has to interrupt the pattern: a number, a sharp question, an unexpected word, a named problem. Feeds and search results are a wall of similar text, and similar text is invisible.
• Select. The reader decides in a blink whether the ad is about them. "Freelancers", "dog owners", "first-time landlords" does that work. A headline that speaks to everyone is heard by nobody in particular.
• Promise. A concrete outcome, a number, a time frame, a price. "Save money on energy" is wallpaper. "The thermostat setting that trims winter bills" makes a promise someone can picture.
• Match. The landing page has to continue the exact thought. If the headline says "3 mistakes", the page should open with three mistakes, not a hero banner and a "Shop now" button.

Headline length limits by traffic source

Every traffic source gives you a different amount of space, and a headline that gets cut off loses its punchline first. Here are the limits and recommendations from each platform's own help pages; they change from time to time, so glance at the field counter in the ad editor before you launch.

Google and Microsoft: write a set, not a single headline

On search you don't write "the" headline anymore. A responsive search ad takes up to 15 headlines and 4 descriptions, and Google mixes them, showing up to three headlines at a time. So you're really writing a kit of parts that have to read well in any order.

The practical consequence: every headline has to make sense standing alone, and no two should say the same thing in different words. Google's own guidance asks for many unique headlines rather than small variations, at least one with the keyword in it, and suggests pinning sparingly, because every pin cuts the number of combinations the system can try. Google also says advertisers who raised Ad Strength from "Poor" to "Excellent" saw about 15% more clicks and conversions on average. That's Google grading its own homework, but the logic holds: three near-identical headlines give the system nothing to choose from.

A kit for a hypothetical password manager offer might look like this (every line fits in 30 characters):

• Password Manager for Teams (keyword)
• Stop Reusing Passwords Today (problem)
• Share Logins Without Sharing (benefit with a twist)
• Compare 5 Password Managers (comparison, great for a review page)
• Set Up in 10 Minutes (objection killer, only if it's true)
• Passwords Synced on Any Device (feature turned into outcome)

Two things will get a search headline disapproved before anyone sees it. Google's editorial rules ban capitalization and punctuation "that aren't used correctly or for their intended purpose" and gimmicky repetition, so "FREE!!! FREE TRIAL" is out. And if a keyword you need is longer than 30 characters, put it in a 90-character description instead of mangling it. I cover keyword choice and bidding in how to make Google Ads profitable.

Meta and TikTok: the first line is the headline

On social, the image or the first seconds of video stop the thumb, and the first line of text decides whether the person reads on. The bold "headline" under a Meta image is closer to a button label: a few words that say what they get, like "Free 7-day meal plan" or "See this week's prices".

So write the hook for the first line of primary text and keep the most important words in the first 125 characters, since the rest sits behind "See more", which very few people tap. On TikTok, the on-screen text and the first spoken sentence are your headline, and the caption is support. I go deeper into TikTok creatives in TikTok advertising.

Native: the headline is the whole ad

In a native widget, the title is basically all you have. The thumbnail helps, but the reader came for the article they're already reading, and your title has to be more interesting than the next paragraph. Realize's own guidance suggests front-loading the main point, opening a knowledge gap without misleading, and starting with 4 to 6 title variations per URL. That's about the right number for a first test, by the way, whatever the network.

Push: plain beats clever

Push is the most cramped format: 30 characters on mobile on many networks, and the device trims the rest. Clever wordplay dies here. Short, concrete, benefit-first lines win: "Track every package in 1 app" says more in 28 characters than any pun. More on the formats themselves in the best paid traffic sources compared.

Headline formulas that keep working

A formula is a structure, not a sentence to copy. The ones below have worked for decades because each one answers a question the reader is silently asking: "what's in it for me?", "is this for me?", "why now?". Use them to produce candidates fast, then let the test pick.

The formula is the easy part. Three habits make it work.

Specific beats impressive. "10 minutes a day" is more believable than "fluent fast", and believable is what gets the click from people who have been burned before, which is everyone. Odd, precise numbers often read as more honest than round ones, as long as they're true.

Steal the reader's words, not yours. People don't search for "credential management solutions". They type "how to share Netflix password safely". The closer your headline sits to the words in the reader's head, the less work it takes them to recognize it as "for me".

Curiosity needs a payoff. The curiosity gap is the strongest native formula and the easiest to abuse. If the page doesn't answer the question in the first screen, the visitor feels tricked and leaves. You paid for that click.

Pick the angle before you polish the words

The angle is what you say; the formula is how you say it. Most beginners test five wordings of the same angle and conclude "headlines don't matter". Five different angles in a plain formula will usually teach you far more than one angle in five clever wordings.

Here's a simple way to build a first batch: take 4 or 5 angles, write each in 3 or 4 formulas, and you've got 15 to 20 headlines in half an hour. Then cut ruthlessly to the 4 to 6 that are most different from each other. Different is the point; you're testing ideas, not commas.

If you get stuck producing angles, that's a creativity problem rather than a copywriting one, and the habits in the 7 keys to your creative genius (generating lots of ideas before judging any, borrowing from unrelated fields) apply directly.

Do negative headlines get more clicks?

Often, yes, at least in news. One large study looked at more than 22,000 headline A/B tests run by the news site Upworthy, covering over 370 million impressions, and found that each extra negative word raised click-through rate by about 2.3%, while each extra positive word lowered it by about 1%. Sad words helped; joyful ones hurt; and, interestingly, fear words slightly lowered clicks.

Before you rewrite every ad as a horror story, notice two things. The effects are per word and small, so negativity is seasoning, not the meal. And the study measured clicks, not sales. In affiliate work, that's the gap that eats budgets.

Here's an example (made-up numbers, run through a calculator, not a real campaign). Two native headlines each get 50,000 impressions. The offer pays $30 per sale, and clicks cost $0.20.

Headline A "wins" on CTR by a mile and loses on money. The scary curiosity headline pulled in people who wanted to see what the fuss was about; the calm specific one pulled in people who actually had the problem. (And yes, 4 versus 5 sales proves nothing statistically; the table shows the mechanism, not a verdict. The next section is about getting a real verdict.)

So judge headlines on revenue per thousand impressions, or at least on EPC, as soon as you have conversions. CTR is the early signal, not the scoreboard. You can run your own numbers in the campaign ROI calculator.

The fear that stops people: "my headline isn't good enough yet"

The trap I see most often isn't writing bad headlines. It's spending three evenings polishing one "perfect" headline and then being too attached to it to test it properly. When it loses, it feels personal, so the test gets extended "a bit longer", or the losing data gets explained away.

Psychologists have a name for part of this. The so-called IKEA effect describes how we value things more when we've put effort into them, which is lovely for bookshelves and terrible for ad copy. The headline you sweated over feels better than it is, and the one you dashed off feels worse. Your audience doesn't know or care which was which.

The practical answer isn't "be braver". It's to change the process so no single headline carries your ego:

1. Write in volume, judge later. Twenty quick headlines in one sitting, no editing. The good ones are usually somewhere around number 12, after the obvious ones are out of your system.
2. Decide the rules before launch. How many clicks, what metric, what result means "kill". Written down while you're calm, not at midnight when the numbers look bad.
3. Treat losers as data. A losing angle tells you what the audience doesn't care about. That's worth money; it just doesn't feel like it.
4. Keep the test cheap. If a headline test costs $40, losing it is a lesson. If it costs $400, it's a crisis. The next section shows how to size it.

The Upworthy numbers are oddly reassuring here. A site that lived on headline testing ran tens of thousands of tests precisely because its editors couldn't call the winners in advance. You're allowed not to either.

How to test ad headlines without fooling yourself

Test headlines by sending the same traffic to the same page with only the headline different, and wait until each version has enough clicks to tell a real difference from luck. The number that matters is clicks per version, not impressions or days.

This part surprises most people. I calculated the table below with the standard two-proportion sample size formula (95% confidence, 80% power). The impressions you need swing wildly with your CTR, but the clicks the weaker headline needs to collect stay almost the same:

So here's a rule of thumb you can keep in your head: to spot a 50% CTR difference you need roughly 80 clicks on the weaker headline; to spot a 20% difference, roughly 400. That's why small tweaks ("Get" vs "Grab") are a waste of a small budget, and different angles are not. A new angle can easily move CTR by half or more; a synonym almost never will.

And remember the money table above: CTR is the cheap first filter. To compare headlines on conversions, the same logic applies to sales, which means you'll need many more clicks. In practice most affiliates do it in two stages.

1. Stage 1, the CTR screen. Launch 4 to 6 headlines with very different angles, same image or thumbnail, same page. Kill anything clearly behind once the leaders have 80 to 100 clicks each.
2. Stage 2, the money round. Keep the top 2 or 3 and let them run until the leaders have a handful of conversions each, then compare EPC or revenue per thousand impressions, not CTR.
3. Stage 3, mutate the winner. Take the winning angle and write 3 or 4 variations of it: a different number, a different opening word, a question instead of a statement. Now small tweaks make sense, because you're working on a proven idea.
4. Repeat every few weeks. Headlines wear out as the same audience sees them again. When CTR on a winner drifts down for a week or two at stable bids, that's fatigue; bring in the next challenger.

Some practical details that keep a test honest. Rotate headlines evenly rather than letting the platform pick a favorite after 50 impressions (on Meta, test in separate ads in the same ad set or use its A/B test tool; on Google, read the asset report but don't over-read it, since Google mixes headlines). Run every test across a full week, because weekend and weekday audiences behave differently. Tag each variant so you can see it in your analytics; the free UTM generator builds the links in seconds. On native, push and pop, a tracker such as AdsBridge records which headline, placement and country brought each conversion; I compare trackers in smartlinks and TDS trackers.

How much does a headline test cost?

A first headline test usually costs less than people fear, because you only need enough clicks to see big differences. Here's an example with made-up but typical-looking numbers: a CPC of $0.20 and the goal of spotting a 50% CTR difference.

• About 80 clicks per headline × $0.20 = about $16 per headline.
• Five headlines in the CTR screen = about $80.
• The money round on the two survivors costs more, because it's counting sales; budget it from the offer's payout (letting each survivor spend two or three payouts is a common starting rule).

At $0.05 per click on push, the same screen costs a quarter of that. At $1.50 per click on search, it's around $600 for five, which is exactly why on search you lean on Google's own rotation and test angles across ad groups rather than running a strict five-way split.

Where the line is: headlines that get you banned (or sued)

An ad headline must be true, and you must be able to prove it before the ad runs. That's the core rule almost everywhere. In the US, the FTC's advertising guide for small business says ads must be truthful and non-deceptive and that you need a "reasonable basis" for claims before they run; health and safety claims need competent and reliable scientific evidence. The UK, EU, Canada and Australia have similar rules, so follow the platforms' rules and the laws of the countries your ads target.

The platforms are stricter than the law in places, and that's where most affiliates get caught. A few rules that directly shape headlines:

• Meta's personal-attributes rule. Meta bans ad text that asserts or implies personal attributes such as health conditions, financial status, religion or age. "New diabetes treatment available" is allowed; "Do you have diabetes?" is not. The fix is usually to talk about the product or a group, not to point at the reader.
• No shouting. Taboola's Realize bans all-caps titles and excessive punctuation (its own bad example is "10 WOMEN YOU WON'T BELIEVE EXIST! UNBELIEVABLE!"), and Google bans gimmicky capitalization, punctuation and repetition.
• Titles must match the page. Realize requires that titles "accurately reflect the subject of the landing page". A curiosity headline that leads to an unrelated sales page is exactly what gets rejected.
• No fake system messages. A push ad titled "Your parcel is moving" when there's no parcel, or a fake "1 new message" alert, imitates a real notification. It may get clicks for a week; networks and offer owners treat it as misleading, and it poisons the traffic.
• Earnings and results claims. "Make $500 a day from home" needs proof of typical results, and in make-money and finance verticals most networks will reject it outright. Testimonials must reflect typical experience, too.

The honest headline isn't only the safe one, by the way. It's usually the profitable one, for the reason the money table showed: a headline that tells the truth about the page pre-qualifies the click, and qualified clicks convert.

How to build a headline swipe file that actually helps

A swipe file is a saved collection of ads and headlines that worked, used as raw material for your own. The trick isn't collecting, everybody collects. The trick is collecting the right ads and filing them so you can use them on a Tuesday afternoon with a deadline.

Where to look, starting with the free stuff:

• Meta Ad Library. Free and public; search any brand or keyword and see the ads it's running on Facebook and Instagram.
• Google Ads Transparency Center. Google launched it in 2023; search an advertiser and see its search, display and YouTube ads.
• TikTok Creative Center. Shows top-performing ads by industry and region, with the hooks right there in the first seconds.
• Paid spy tools. Services like PowerAdSpy add filters across networks, engagement data and the date an ad was first seen. Worth it once you're spending regularly; not needed for your first test.
• Your own inbox. Big brands' email subject lines are headlines tested at scale. Save the ones you opened.

Now the tricks that make a swipe file useful rather than decorative.

Longevity is the signal. An ad that has been running for months is almost certainly making money, because nobody keeps paying for a loser. One ad live for 90 days beats fifty that appeared yesterday. Most ad libraries show the start date; sort by it.

File by angle and formula, not by niche. A folder called "fitness" is useless when you're writing for a VPN. A folder called "objection killers" or "audience callouts" works for every offer. Add a one-line note on why you saved it.

Steal structures across niches. The best headline ideas usually come from somewhere else. A meal-kit "No more 'what's for dinner?'" structure becomes "No more 'which password was it?'" for a password manager.

Mine the customers' words. Read one-star and three-star reviews of the product and its competitors, and Reddit threads where people complain about the problem. The exact phrases people use ("I'm sick of resetting passwords") are often better headlines than anything you'd write.

AI-assisted headline generation done right

AI is excellent at the volume part of headline writing and poor at the judgment part. Use it to produce many varied candidates from facts you give it, then do the choosing, the fact-checking and the length check yourself.

Where it goes wrong is predictable. Ask a chatbot "write me 10 headlines for a VPN" and you get ten versions of "Unlock Your Online Freedom", the same headline every other affiliate got from the same prompt. The fix is to give it what it can't know: the facts, the audience, the angles and the limits.

A prompt structure that works much better:

1. Facts only you can verify. What the product does, the real price or trial terms, the guarantee, one or two true specifics. Tell it explicitly: "use only these facts; don't invent numbers, ratings or results".
2. Who it's for. One sentence on the reader and the situation they're in, plus a few phrases you collected from real reviews.
3. Angles and formulas. "Write 5 headlines for each angle: pain, ease, price, social proof, curiosity. Use these formulas: number + outcome, audience callout, objection killer."
4. The source and its limits. "For native: under 60 characters, sweet spot 34 to 45. For Google: 30 characters max." Then check the lengths yourself, because language models are notoriously unreliable at counting characters. That spreadsheet =LEN() column earns its keep here.
5. A banned list. Words you never want ("unlock", "revolutionary", "game-changing"), plus the compliance rules from the section above, such as "never ask the reader about their health or finances".

Then use the AI a second time, after the test. Give it the winning headline and the losers and ask for 10 variations of the winner that keep its angle but change the number, the opening word or the structure. That "mutate the winner" step is where AI saves the most time, because the direction is already proven.

The built-in generators deserve the same caution. Google's responsive search ads can customize text automatically, Meta's Advantage+ creative can generate text variations, and tools like AdCreative write headlines along with the banners (I walk through its trial and credits in how to use AdCreative to advertise). They're handy, but you are the advertiser of record. If the generator writes "Rated #1" or "Save 40%" and you can't prove it, it's your account and your liability. Review every generated line before it goes live, or turn the feature off for offers in sensitive verticals.

Headline mistakes that quietly burn budgets

Most headline losses come from the same few habits, and the formulas above don't protect you from any of them.

• Calling a winner after a day. Twenty clicks against fifteen is noise. Wait for the click counts in the testing table, and a full week.
• Writing for yourself. Industry jargon and clever puns delight the writer. Readers want their own words and a clear benefit.
• Ignoring fatigue. A winner that ran for two months isn't broken, it's tired. Have the next challenger ready before CTR slides.
• Copying the market leader word for word. Even if it's allowed, the audience has already seen it. Borrow the structure, change the angle.

What to do next

Pick one offer you're already running or about to test. Write 20 headlines across 4 or 5 angles, using the formulas table, and check each against the length limit for your traffic source. Cut to the 4 to 6 most different ones, decide your kill rule (about 80 clicks per headline for a first screen), and launch with one image and one page. Then judge by EPC, not CTR, and spend the next round mutating whichever angle won.

If the clicks come but sales don't, the headline may not be the problem at all; look at the page with the landing page guide, and run the numbers in the campaign ROI calculator before you scale anything.