How Much Text Should an Ad Have

Everyone in advertising repeats that less is more. The counter-argument is obvious: if the ad never says what you sell, nobody can want it. Short ads and long ads both win, in different places, and the number that decides which one wins is measurable.

The number that decides how many words an ad can carry is how long somebody actually looks at it. Looking time has been measured with eye trackers on real roads and real screens, and it sets a hard limit on how many words an ad gets.

A driver looks at a billboard for a third of a second

The US Federal Highway Administration put eye trackers in cars and drove 55 people past real billboards in Reading, Pennsylvania and Richmond, Virginia.

• The average look at a digital billboard lasted 0.379 of a second, and at a printed billboard 0.335 of a second. A glance at a billboard lasted about as long as the average glance drivers gave the road ahead.
• The longest single look at any billboard in the Federal Highway Administration study was 1.335 seconds.
• Adding up every glance a driver gave one billboard, the total came to between 0.674 and 1.386 seconds. Drivers looked at each billboard in two or three quick glances rather than one long look.
• A second field study using eye trackers in traffic found drivers gave advertising just 1.2% of all their glances. Each glance at an ad in that study lasted 0.5 to 0.75 of a second.

A driver has time to read only three to five words

A 2019 review of 190 studies covering 18,573 people put adult silent reading at 238 words a minute, which is just under four words a second.

• A single 0.335 second glance at a printed billboard is long enough for one or two words.
• The full 0.674 to 1.386 seconds a driver spends on one billboard is long enough to read three to five words. Three to five words is not enough for a sentence, or for a tagline plus a phone number plus a web address.
• A display banner that somebody does look at gets 1.3 seconds on average, which is long enough for about five words. The 1.3 second figure for banners was measured by the eye-tracking firm Lumen Research.

Very simple billboards only work for brands people know

If the brand name takes two of the three to five words a driver can read on a billboard, there is almost nothing left for the message.

• A logo the reader already knows is taken in at a glance, so it uses up none of the three to five words a driver can read. A well known brand can spend every word on the message instead of on its own name.
• An unfamiliar brand name has nothing already in the reader's head to match it against, so it has to be read like any other words. A two word brand name nobody knows leaves a billboard only two or three words for everything else.
• A brand name with no category beside it leaves the reader with no idea what the business does. "Vlix" tells a driver nothing, and "Vlix, language apps" tells them what it is in three words.

A new brand cannot introduce itself and persuade in five words

A very short billboard line relies on the logo to say who the ad is from, and a brand nobody knows yet cannot rely on that.

• A billboard has three jobs and room for five words, and a logo people know says who the ad is from without using up any of those words. The three jobs are saying who you are, saying what you sell, and giving one reason to care.
• An unfamiliar brand name plus its category costs two to three of a billboard's five words, which leaves room for one fact at most. Adding "15 minutes a day" to "Vlix, language apps" makes seven words, well past what a driver can read.
• A new brand has to choose between introducing itself and making an argument, because five words will not carry both.
• A new brand should choose being understood over persuading, because a driver who cannot tell what you sell has nothing to remember you by.

The seven word billboard rule rests on a misread study

Almost every billboard guide says seven words or fewer, and credits George Miller, an American psychologist, and his 1956 paper on the magical number seven.

• The psychologist George Miller said his own magical number seven had nothing to do with reading. Miller's seven was a limit for telling apart simple signals and for immediate recall, "neither of which has anything to do with a person's capacity to comprehend printed text".
• The real working memory limit is now put at three or four items, not seven.
• The seven word billboard rule lands near the right number of words by luck, and the short looking time is the real reason billboards need few words.
• The widely quoted claim that six to eight words gives 65% better recall has no traceable source. The 65% recall claim appears on billboard company blogs with no study attached.

Every ad medium ranked by how many words fit the time

Each place an ad can run gets the same arithmetic. The number of words is the measured looking time multiplied by four words a second.

When the reader controls how long they look, as in a feed or a search result, the time limit disappears and the advice reverses.

Use Repeat Recorder to time your billboard line out loud

Saying a billboard line out loud tests its length better than reading it in 200 point type. The 2019 review of 190 reading studies measured reading aloud at 183 words a minute, so five words takes about 1.6 seconds to say. Repeat Recorder, one of my own apps, replays what you said and then records your next attempt.

• Repeat Recorder plays your line back about a second after you say it, so you hear the length rather than guess at it. A line you cannot finish in one breath is too long for a billboard.
• Repeat Recorder can loop one take as many times as you set, or forever until you stop it. Hearing a five word billboard line over and over shows the one word doing no work.
• Repeat Recorder has a fully hands-free mode where speaking starts the recording and stopping replays it.

Longer ads beat shorter ads whenever the reader can click

I used a public archive of randomised advertising experiments run by the news publisher Upworthy between 2013 and 2015, where each test changed only the wording. I kept the 10,592 tests where the picture was identical too, then compared the shortest and longest wording inside each test.

• The shorter ad wording won only 43.9% of 10,372 randomised Upworthy tests. On average the shorter wording got 6.4% fewer clicks than the longer wording.
• In the Upworthy tests, the bigger the length gap between two wordings of the same ad, the worse the shorter wording did. Where one wording was at least ten words longer, the shorter one won just 34.9% of the time and got 18.0% fewer clicks.
• Adding one word to an Upworthy headline made no measurable difference, while adding five words won about 6% more clicks.
• Longer wording beat shorter wording separately in 2013, 2014 and 2015, and also when length was measured in characters instead of words.

Extra words only win when they carry a fact

I split the Upworthy headline tests by what the extra words in the longer wording actually added.

• Where the longer wording also added a name, a number, a price or a percentage, it won 57.4% of 7,137 tests and gained 8.1% more clicks.
• Where the longer wording added only mood and story words with no new facts, the gain fell to 2.4% and was too small to tell apart from chance.
• In the Upworthy tests the shorter wording lost because it said less, not because readers prefer brevity.
• Every word you add has to tell the reader something they did not already know.

Adding words stops helping at about 18 words

Splitting the Upworthy tests by how long the shorter wording already was shows where the gain from adding words runs out.

• When the shorter wording was under ten words, the longer version gained 16.6% more clicks, the biggest gain in the Upworthy tests.
• When the shorter wording was 13 to 19 words long, the gain from going longer had shrunk to about 6.5% more clicks.
• When the shorter wording was past 19 words, the gain from going longer was too small to tell apart from chance. About 18 to 20 words is as long as a clickable headline is worth making.
• The click rate by headline length across every Upworthy test also peaks at 18 words. The click rate climbed from 0.95% at six words to 1.23% at eighteen, then eased off.

Making every word specific backfires in half of tests

Two Cornell researchers, Marianne Aubin Le Quéré and J. Nathan Matias, scored how concrete each wording was across 8,977 experiments in the Upworthy archive, published in Scientific Reports in 2025.

• Where an Upworthy headline was already vague, the vaguer of two wordings got 5.5% fewer clicks than the more concrete one.
• Where an Upworthy headline was already concrete, making it more concrete cut clicks by 9.9%. A headline made of nothing but specifics reads like a dry list of product details.
• In the Cornell study, 50.9% of the Upworthy tests had headlines that were already too specific, and only 8.7% had headlines that were too vague.
• Adding a few facts to a long headline wins, and making every single word a specification loses.

Ads that ask a question and hold back the answer won 56% of click tests

Two ten word versions of the same article ran against each other in the Upworthy archive. "Ever Wonder What Happens To Your Body When You Fast?" took 1.22% of the clicks, and "Slow Down. Here's Everything You Need To Know About Fasting." took 0.33%.

• The Upworthy headline that left its question unanswered won 56.3% of 1,667 randomised tests where both wordings were exactly equal in length. The unanswered question also won when it was the shorter of the two wordings.
• Leaving one question unanswered is not the same as a headline that is vague from start to finish. In the Cornell study, making an already vague Upworthy headline vaguer cut clicks by 5.5%. A winning curiosity headline names the person, the place and the problem, and leaves out only the answer.
• Naming the specific person or product in a headline made no measurable difference to clicks in either direction. Across 671 Upworthy tests the headline that named its subject won 51.6% of the time, the same as tossing a coin.

Leaving a question unanswered only works where people can click

Every winning curiosity headline in the Upworthy tests was seen by somebody who could get the answer with one tap.

• A reader only acts on curiosity when the answer is one tap away, and a driver going past at speed cannot tap anything. A question on a billboard stays unanswered, because the driver has no way to find out the answer.
• The Upworthy archive holds only two usable tests at billboard length, so it cannot show whether a curiosity line works on a billboard. Because of that gap, the billboard advice rests on eye-tracking studies of looking time and recall instead.

Use Camera to Clipboard to collect billboards you drive past

The billboards you already pass every week are the best training for writing short ad lines, but photos of those billboards usually stay on your phone. Camera to Clipboard, another of my own apps, puts a photo from your phone straight onto your computer's clipboard.

• Camera to Clipboard sends the photo to your computer clipboard, so you paste it into a document instead of emailing it to yourself.
• Camera to Clipboard pairs your phone to your desktop once and then works over your network. Camera to Clipboard needs no cable and no upload step.
• Pasting billboard photos side by side is how you count the words in a billboard that worked on you. Most billboards that worked on you will turn out to use three to five words.

Dense detail hurts attention but elaborate design helps

A 2010 study in the Journal of Marketing eye-tracked 249 real ads and found that two things both called complexity pull in opposite directions.

• Cramming more detail, more objects and more small elements into an ad reduced attention to the brand and made people like the ad less.
• An elaborate creative idea in an ad raised attention to its picture, to the ad as a whole, and to how well people understood the ad.
• Cut the number of things competing for attention in an ad, not the ambition of the ad's idea. You cannot buy more looking time, but you can decide whether your idea is interesting.

Only 35% of display ads are ever actually seen

Lumen Research measures how often anybody looks at a banner with eye trackers rather than with delivery reports.

• In Lumen Research's eye tracking, only 35% of digital display ads got looked at at all. The other 65% of display ads were shown on a page and paid for without anyone looking at them.
• Of the display ads shown on a page, just 9% were looked at for more than one second and 4% for more than two seconds.
• Only about 30% of display ads that counted as "viewable" were actually seen. "Viewable" is an industry definition that only checks whether the ad's pixels were on screen.
• Readers avoid looking at web banners on purpose, which an eye-tracking study showed back in 2003. In that 2003 study, only 46.9% of people could remember seeing any banner at all afterwards.

Billboards now beat display ads for attention

Phones were expected to make billboards obsolete, but attention measurements show billboards now hold attention better than display ads.

• Roughly 2.5 seconds of attention is the point where an ad starts entering memory, and 85% of digital ads get less attention than that. The 2.5 second threshold comes from the attention research of Karen Nelson-Field, an Australian media researcher.
• About 89% of roadside ads hold attention for the 2.5 seconds an ad needs to enter memory, against 15% of digital ads.
• A distinctive brand asset can get an ad into memory in 1.5 seconds, which is less time than a billboard already gets. A distinctive brand asset is a colour, a shape or a character the audience already links to you.
• The strongest numbers favouring billboards come from companies that sell billboards, so treat the size of the gap as unsettled. A five-year Kantar study reporting 13.3% higher ad awareness for roadside advertising was commissioned by Clear Channel Outdoor, which owns billboards.

Display ads look cheaper until you count only the ones seen

On money the ranking starts the other way round, because a display impression sells for less than a billboard impression.

• A billboard costs about half a cent per person reached and a display banner about a third of a cent. A $5.21 rate per thousand impressions is 0.52 cents a head, and a $3.12 rate is 0.31 cents.
• Counting only the impressions that held someone for 2.5 seconds, a static billboard costs about 72% less than an open exchange banner. The open exchange banner starts 40% cheaper per impression but ends up at $20.80 per 1,000 attentive impressions, against $5.85 for the billboard.
• A billboard impression and a display impression are not the same unit, which is the biggest weakness in this comparison. A billboard impression is estimated from traffic passing the site, while a display impression is an ad a server sent.

There is no solid proof that billboards change what people do

Whether a billboard changes what people actually do has been tested properly only once, by Donald Green, a Columbia University political scientist, and colleagues during the 2020 American election.

• In Donald Green's experiment, 298 billboards were randomly assigned across 155 areas, and every result the authors had committed to in advance was too small to tell apart from no effect. Turnout in that experiment came from public records for 18.4 million people. The billboards in that experiment used short, clear wording.
• Donald Green's billboard experiment counted only whether a person turned up to vote at all, not who they voted for or whether they bought anything.
• Smaller studies have measured billboards moving sales, so what is missing is the strong proof rather than the effect itself. The marketing researchers Bhargava and Donthu recorded an immediate sales response to outdoor advertising in 1999.

Treat a billboard as a way to make a name familiar rather than a way to make somebody act this week.

Use Repeat Recorder to rehearse an ad line while driving

The drive itself is usable time for writing billboard lines. Repeat Recorder, one of my own apps, has an automatic mode where speaking starts the recording and going quiet replays it, so you never touch the phone.

• Repeat Recorder's automatic mode needs no taps at all, which is what makes it usable with your hands on the wheel. You say a candidate billboard line, Repeat Recorder plays it back, and you say the next one.
• Repeat Recorder saves the takes worth keeping into folders you name yourself. Ten spoken versions of one line, kept together, is a better shortlist than ten typed ones.

Ten ads rewritten to fit the time they actually get

Fitting an ad to its looking time goes in both directions. Roadside and banner ads come down to the number of words there is time for, while ads the reader can click go up.

The three feed ads go the other way, up from four to six words to about sixteen, and every word added carries a number or a name.

Conclusion: how much text an ad should have

The rule "less is more" is about time, not about advertising, and it only applies where the looking time is short and fixed.

• Count how long people look at the ad, then allow four words a second.
• A billboard gets three to five words, and a display banner about five.
• An ad the reader can click should run to about 18 words. Shorter wording got 6.4% fewer clicks in the Upworthy tests.
• Add words only when they carry a name, a number or a price. Added facts won 8.1% more clicks, and added mood words won nothing measurable.
• In an ad people can click, name the person and the problem, then leave the answer out. Making an already concrete headline more concrete cut clicks by 9.9%, and making an already vague headline vaguer cut clicks by 5.5%.
• A very simple billboard only works if people already know the brand.
• If nobody knows your brand yet, spend the billboard on being understood rather than on persuading.
• Do not expect a billboard to make people look you up.
• Ignore the seven word billboard rule and George Miller's 1956 memory paper behind it. George Miller said his magical number seven had nothing to do with reading.