Hand your phone to a neighbor who has never seen your website. Let her look at the first screen for five seconds, take the phone back and ask what your company does.
If she names the job and your town, move on. If she says “something with houses,” that's the first thing to fix.
What is the five-second test?
It's a quick first-impression check: show a page for five seconds, hide it and ask what the person took in. A 2016 paper in the Journal of Usability Studies describes the same method. It shows a fixed page for exactly five seconds, then asks a few open questions.[1]
Five seconds is a habit of the trade, not a measured limit. That paper says the method “has never been scientifically proven.”[1] In a 2023 study, people saw website pages for two, five or ten seconds. A visually complex page shown too briefly was linked to trouble saying what the page was for.[2]
The 50-millisecond figure you may have seen comes from a 2006 study in which people rated how appealing home pages looked.[3] That's a snap rating of looks, not a test of understanding. So I'd use five seconds as a repeatable look, not a law: same page, same questions, new people each time.
How do you run it on your own website?
Show your phone's first screen to three people, one at a time. That's what appears before anyone scrolls.
- Load the page on cell data, scroll to the top and tap nothing. If it crawls or jumps while it loads, fix the speed problem first.
- Pick homeowners who've never seen your site and don't work for you. Google's own content checklist suggests an honest assessment from people you trust who aren't affiliated with your site.[4] A US General Services Administration (GSA) list of usability-test indicators flags friends, family and close colleagues of the team as sampling bias, so I'd pick a neighbor you barely know over anyone close.[5]
- See them one at a time, not as a group, the way the Department of Homeland Security (DHS) usability kit does.[6]
- Tell them a page will show for five seconds and to remember what they can.[1]
- Hand over the phone, count five seconds, take it back and lock the screen.
Three is a start, but the DHS kit says to try for at least four people who might really use what you're testing.[6] It adds that you don't need a large sample. Add a fourth if you can find one.
What do you ask when the five seconds are up?
Ask three open questions, one at a time: what does this company do, where does it work, and what would you do next? Don't hint. A GSA sample script asks “What are your first impressions? What is this site for?” before anyone clicks.[5]
| Ask | A yes sounds like | A no sounds like |
|---|---|---|
| What does this company do? | “Siding.” | “Home improvement.” |
| Where does it work? | “Raleigh.” | “North Carolina somewhere.” |
| What would you do next? | “Call the number.” | “Scroll and look around.” |
Mark a yes only when the answer would send the right kind of job to your phone. Swap in your own trade and town, and write each answer down before you score it.
What do the results tell you to fix first?
My rule: if two or more people can't say what you sell or where, fix that screen before paying for SEO. Ranking higher just sends more people to the same screen.
- They miss the job: rewrite the headline and the line under it in the words a customer uses.
- They miss the town: put it in the first line. A footer address doesn't count.
- They don't know what to do: show one tap-to-call number or a short form without scrolling.
The 2023 study's advice is to fit the display time to the page and the person.[2] Mine is simpler: if someone misses at five seconds, let them look for ten and ask again, but keep the five-second score.
If they get it then, the screen may be too busy, or that person may just need longer; if a second person does the same, trust the screen. If they still miss it, the words probably aren't there. Either way, fix one thing and test again with three new people, because the first three have seen the old screen.
What does a first screen need for a big job?
Four things: the job and the town in plain words, one proof, one action, and nothing in the way.
- The job and the town, in the first line. Google's content checklist asks if a site has a primary purpose or focus, and if the main heading gives a descriptive, helpful summary of the content.[4] Say it the way a customer asks: “Siding replacement in Raleigh,” not “Quality you can trust.”
- One proof. Something a stranger can check. Pick one: a real photo of a finished job, the star rating from your Google Business Profile, or a license number you can show. How to photograph your jobs covers the photo.
- One action. A phone number that dials when tapped, or a short form with a few fields.
- Nothing in the way. A pop-up offer can sit on top of your lines. Google's advice is to use banners that take up only a small fraction of the screen, not full-page overlays.[7] A slider shows one slide at a time, so a five-second look may catch the wrong one.
Open your home page on your phone, read only what shows before you scroll, and tick off all four.
What does a fix look like?
Fewer, plainer words. Here's a made-up siding company in Raleigh, before and after.
A siding company's first screen in Raleigh
- Before: a slider with a stock photo of a smiling crew. Headline: “Quality you can trust.” Under it: “Family owned. Excellence in every project.” One button: “Learn more.” The phone number hides in a closed menu.
- After, headline: “Siding replacement in Raleigh, Cary and Apex,” with “We replace vinyl and fiber cement siding on homes in these towns” under it.
- After, proof: one photo of a finished house, captioned “Fiber cement siding on a two-story house in Raleigh,” next to the star rating from the company's Google profile.
- After, action: a “Call the office” button that dials when tapped, and a short form for name, phone, address and the job.
Illustrative example
Write the lines from your own jobs. If a chatbot drafts them, run the checks for AI-written pages. How to write a service page covers everything below the first screen.
What should you skip?
Skip anything that coaches the answer or turns three opinions into a score.
- Helping. If someone stalls, write “stalled” and move on. The DHS kit says not to help participants.[6]
- Leading questions. “Did you see the license?” suggests the answer. The same GSA page lists leading questions as something to avoid.[5]
- A percentage. Three people can't support one, so count the no's. DHS says not to worry about statistical significance. The goal is to learn enough to make decisions.[6]
- A paid testing service. Try it by hand first.
What can't five seconds tell you?
Anything past whether the first screen is clear. Here's what it leaves out:
- Trust and price. A clear screen isn't a convincing one. Before a big job, a visitor usually checks reviews, photos and prices. What homeowners check before they call a remodeler covers that.
- Whether Google shows the page. Start with why your website doesn't show up on Google.
- Whether anyone calls. A lead log shows that. Three neighbors can't.
- Broken buttons. The 2016 paper's author doubts the test is a good way to find usability problems, since people probably notice those only after the first five seconds.[1] Tap your own number and send your own form instead; the website audit checklist has both checks.
I'd run this one myself. Pay for a second look when your three people pass and the phone is still quiet, because then the cause is somewhere else.
Our website clarity audit looks at your site the way a first-time visitor does on a phone. It checks what you do and where, then your main action, your proof, and how clearly you handle price or quotes.
The free AI visibility check is a review by hand of the website and Google profile, with the first fixes, usually within two business days. The Expert Action Plan is $149, the Complete Visibility Report plus a 30-minute consultation.
Every page and abstract cited here was read on October 15, 2026.