Free resource · no email required
The UX audit checklist I actually run
26 questions, grouped into the four lenses every audit goes through. Each one says what a failure costs you and how to test it yourself — no analytics, no session recordings, no testing platform. A browser, a private window, and someone who has not seen the page before.
Work through the lenses in order. It matters: a page that fails the first one rarely gets a fair reading of the fourth, because nobody stays long enough to reach it. Mark each question pass, fail or not applicable, and write one line on what you saw — that line is what turns a checklist into something you can still act on a week later.
When you are done, sort the failures by what they cost, not by how easy they are to fix. Then take the three most expensive and the five cheapest. That is the shape of every report I send, and the Uxerfy Score explains how it all turns into one number. New to this? What a UX audit is covers the ground before the checklist.
First impression
Everything here is decided in about five seconds, before anyone reads a feature list. It is also the cheapest place to find points: most pages are one sentence away from being understood.
-
Can a first-time visitor say what this does, in their own words, after five seconds?
If they cannot, nothing further on the page gets read. This single question predicts more drop-off than anything else on this list.
Show the top of the page to someone who does not know the product, take it away after five seconds, and ask them what it was. Their words, not yours.
-
Is there one obvious next step, or several competing for the same click?
Two equally-weighted calls to action do not double the conversions; they split the decision and lower both.
Squint at the first screen until the text blurs. Count how many things still pull the eye. More than one primary is one too many.
-
Does the first screen show the thing the visitor came for?
A tall hero that pushes the pricing table, the product shot or the signup below the fold makes the page feel emptier and slower than it is.
Load it on a laptop at 1366×768 and on a phone. Is the thing you are selling visible, or is it the word "Pricing" on a coloured background?
-
Does the page hold still while it loads?
Layout shift moves the target after the visitor has committed to the click. It is experienced as the page being broken, not as the page being slow.
Reload with a throttled connection and watch the first two seconds. Anything that jumps — a banner, a font swap, an image without dimensions — is costing you.
-
Is it obvious what kind of thing this is — a product, a service, a tool, an agency?
Visitors classify before they evaluate. A page that skips the category forces them to do the work, and some will not.
Read only the headline and subhead. Could they belong to a consultancy, a SaaS product and a newsletter equally? Then they are not saying enough.
-
Does anything on the first screen exist only because a stakeholder asked for it?
Every extra element taxes the same attention the primary action needs. Cost is paid in conversions, not in pixels.
For each element, name the visitor question it answers. Anything that answers an internal question rather than a visitor one is a candidate for removal.
The commonest failure is a headline that names a category — "the modern platform for teams" — where a visitor needed a benefit. It reads as confident and says nothing.
The path to value
This is where signups leak quietly. Nothing here looks broken — each step is reasonable on its own, and the loss only shows in the aggregate.
-
How many steps stand between arriving and the core action?
Every step is a chance to leave. Steps that could be deferred until after commitment are pure loss before it.
Walk it yourself in a private window and count. Include email verification, plan selection, and anything that sends you to another tab.
-
Is every form field necessary at that moment?
A field asked too early is a reason to close the tab; the same field after the visitor is invested is answered without complaint.
For each field ask: what breaks if I collect this later? If the answer is "nothing, it is just useful", it is being asked too early.
-
Does each step say what happens when it is completed?
People hesitate at buttons that could plausibly charge them, email them, or commit them to something. Naming the outcome removes the hesitation.
Read each button label alone, without the surrounding page. Does it say what follows, or does it say "Continue"?
-
Can a visitor see progress, and see the end?
An unbounded flow feels longer than a bounded one of the same length. Uncertainty about how much is left is itself friction.
Count the screens. If there are more than three and no indication of how many remain, the visitor is being asked to trust an unknown quantity.
-
What happens if someone stops halfway and comes back tomorrow?
Losing progress converts an interested visitor into a lost one, and they rarely start again.
Start the flow, close the tab, return. Is the state there? Was an email sent that lets them resume?
-
Is the fastest path the one the page recommends?
Where a shorter route exists — social sign-in, a demo without an account, a trial without a card — but the page pushes the longer one, the design is working against itself.
Compare the promoted route to the quickest possible route. If they differ, ask what the longer one buys, and who it buys it for.
-
Does the visitor reach something that feels like value, or only like setup?
A flow that ends in an empty dashboard has moved someone through the door and left them standing in a corridor.
At the end of onboarding, ask: has this person seen the thing they came for, or have they only configured the ability to see it later?
The commonest failure is asking for company size, role and phone number before the visitor has seen anything work. Each field is defensible; together they are a wall.
Clarity of copy
Copy is the cheapest thing on a page to change and the most frequently mis-assigned to design. Most clarity problems are sentence problems.
-
Does every button say what happens when you press it?
"Submit", "Continue" and "Get started" describe the click, not the consequence. The consequence is what the visitor is deciding about.
Cover the page and read the button labels in a list. Would you know what each one does?
-
Does the headline name a benefit or a category?
A category headline requires the visitor to do the translation into their own problem. Most will not bother.
Ask "so what?" after the headline. If there is an obvious answer you did not print, print it.
-
When something goes wrong, does the page say how to recover?
An error that states a condition — "invalid input" — leaves the visitor to guess. Guessing at a form is where people quit.
Submit every form wrong on purpose. Empty, malformed, already-used email, wrong card. Read what comes back.
-
Is anything written in the vocabulary of the org chart rather than the visitor?
Internal names for features, plans and states are invisible to the people who wrote them and opaque to everyone else.
Mark every proper noun that is not your brand. Each one is a term the visitor must learn before the page is useful.
-
Does the page claim anything it does not immediately support?
An unsupported superlative costs more than it earns: it invites scepticism at the exact moment you need credit.
Find every "best", "leading", "seamless" and "effortless". For each, is there evidence within one scroll? If not, cut it or prove it.
-
Is the reading level appropriate to the moment, not to the industry?
Technical audiences do not want simplified copy, but nobody wants to parse a sentence twice while deciding whether to pay.
Read the decision-critical sentences aloud. Anywhere you have to restart a sentence, so does the reader.
The commonest failure is writing from the system outward — naming the feature rather than what it lets someone do, or an error that reports a state rather than a remedy.
Trust and friction
Trust problems announce themselves as hesitation, not as complaints. Nobody emails to say they were not sure whether they would be charged.
-
Can a visitor work out what this costs without leaving the page?
A price that requires a click, a calculator or a call is a price the visitor assumes is bad. Often it is not, and the assumption costs the sale.
Ask someone to find the monthly cost for a team of five. Time them. Anything over about twenty seconds is a problem.
-
Do reassurances appear where the doubt happens?
"No card required" three scrolls below the signup button is answering a question after the visitor has already left.
List the moments of hesitation — the button, the card field, the plan choice. Is the relevant reassurance within sight of each?
-
Is anything asked for before the visitor has a reason to give it?
Commitment requested ahead of demonstrated value reads as a bad trade, whatever the eventual value is.
For each thing you ask for — email, card, phone, calendar — name what the visitor has received by that point.
-
Is one option recommended, or is the choice left entirely open?
Undifferentiated options hand the sorting work to the visitor. A stated default speeds the decision for the average buyer and costs the others nothing.
Look at the plan table. If every column carries equal visual weight and none says who it is for, the page is not helping anyone choose.
-
Is it clear what happens after the commitment — and how to reverse it?
Visible exits increase entries. Someone who can see how to cancel is more likely to start.
Find the answers to: when am I charged, what happens at the end of the trial, how do I cancel. Count the clicks to each.
-
Does the proof match the claim it sits next to?
A generic testimonial beside a specific claim weakens both. Proof that does not address the doubt in front of it is decoration.
For each piece of social proof, name the objection it answers. If it answers none, move it to where it does.
-
Is anything on the page pressuring rather than informing?
Countdowns that reset, "only 2 left" that is never true, and pre-ticked boxes are read as evidence about the company, not about the offer.
Reload the page in a fresh session. Anything urgent that resets was never urgent.
The commonest failure is putting the reassurance in the footer and the doubt at the button — the two are three scrolls apart, and only one of them gets read.
Questions about the checklist
How long does it take to run this checklist myself?
About two hours for a single page if you are honest with yourself, and most of that is the testing rather than the reading. The path-to-value section takes the longest, because it is the only one you cannot do by looking — you have to walk the flow.
Do I need any tools to run it?
No. Every check is designed to run with a browser, a private window and someone who has not seen the page before. Nothing here needs analytics, session recording or a testing platform. Those help, but waiting for them is the reason most audits never happen.
What do I do with the answers?
Sort the failures by what they cost, not by how easy they are to fix — those are different lists, and mixing them is how harmless changes end up at the top. Then take the three most expensive and the five cheapest, and do those. That is the shape of every report I send.
Can I use this checklist for client work?
Yes, freely. It is published because a method nobody can check is not a method. If it is useful in your own practice, take it — a link back is welcome but not a condition.
I'll run the whole checklist on your page.
A PDF report within 24–48 hours — a score, 3 blockers, 5 quick wins. See what the audit covers.
Get a free audit →