Most websites that underperform don’t look broken. The design is tidy and everyone in the office can find what they need. Meanwhile, visitors who don’t know the business hunt for a price that isn’t there, tap headings they think are links, or abandon a form that rejects their phone number. A website UX audit finds those moments before you spend money redesigning around guesses.
UX is where strategy, design, content and development meet, so a polished site can still fail its visitors. Of the ingredients of what makes a good website, it is the hardest to judge from the inside.
The short answer
A website UX audit is a structured review of how well real visitors complete the key tasks on your site, such as finding a service, comparing options or sending an enquiry, and of where and why they fail. A credible audit combines three kinds of evidence:
- Expert review against usability heuristics, usually Jakob Nielsen’s ten.
- Behavioural data from analytics, site search and session recordings.
- Usability tests, watching around five real people per round attempt those tasks.
The output is a prioritised list of problems, each with evidence, a severity rating and a recommended fix.
What a website UX audit checks
A good audit is organised around tasks, not pages. “Is the homepage good?” can’t be answered. “Can a buyer on a phone find out whether we make this part and ask for a quote?” can be tested. Along each task, the auditor examines:
| Area | What the audit asks |
|---|---|
| Findability | Can visitors reach what they came for from wherever they land? |
| Clarity | Do pages answer real questions, price and process included, in customers’ words? |
| Trust | Is there enough evidence, such as past work and named contacts, to choose you? |
| Forms and contact | Can people get in touch without needless fields or confusing errors? |
| Mobile and feedback | Do menus, tap targets and buttons work and respond clearly under a thumb? |
| Accessibility basics | Can keyboard, zoom and screen reader users finish the task? |
| Languages | Does each language version, forms included, work as well as the main one? |
It isn’t a critique of taste: “the blue feels cold” is only a finding if visitors struggle because of it. Nor is it a full WCAG 2.2 review, which our guide to accessibility audits covers, or an SEO audit.
The three methods behind a credible UX review
Each method has a blind spot the others cover.
| Method | What it tells you | Blind spot |
|---|---|---|
| Heuristic evaluation | Where the design breaks established usability principles | Predicts problems; doesn’t prove visitors hit them |
| Analytics and recordings | Where and how often visitors drop out | Shows what happened, rarely why |
| Usability testing | Why people struggle, in their own words | Small samples say little about frequency |
Heuristic evaluation
A reviewer walks each key task, noting every point where the site breaks a recognised usability principle. The best-known set is Jakob Nielsen’s ten usability heuristics, published by the Nielsen Norman Group (NN/g). On a business website, they look like this:
| Heuristic | Check whether… |
|---|---|
| Visibility of system status | Forms confirm they were sent; buttons react when tapped |
| Match between the system and the real world | Labels use customers’ words, not internal codes |
| User control and freedom | Pop-ups close easily; people can go back and edit |
| Consistency and standards | Links and buttons look the same everywhere |
| Error prevention | Fields accept international phone numbers and long names |
| Recognition rather than recall | Options are visible, not hidden behind cryptic icons |
| Flexibility and efficiency of use | Returning visitors can jump straight to what they need |
| Aesthetic and minimalist design | Pages lead with what matters, not competing banners |
| Help users recognise, diagnose and recover from errors | Error messages explain the fix, next to the field |
| Help and documentation | FAQs and contact options appear where people get stuck |
NN/g recommends three to five independent evaluators, because any one reviewer misses problems.
Analytics and session recordings
Your data shows where the experience leaks. In Google Analytics 4, start with:
- A funnel exploration from a key service page to the contact page to a submitted form, split by device. A funnel that holds on desktop and collapses on mobile is a finding in itself.
- Form starts versus submissions. A large gap usually points at the form, not the offer. GA4’s automatic form events miss some forms, so check they fire (tracking enquiries in GA4).
- Site search terms, if your site has search: what visitors couldn’t find through the menu, in their words.
Session-recording tools such as Microsoft Clarity or Hotjar then show individual visits. Clarity, for example, flags rage clicks (rapid clicks on one spot), dead clicks (clicks that produce no response) and excessive scrolling. Watch twenty recordings of visitors who reached the contact page without sending the form; patterns usually emerge. Mask form inputs and respect consent rules.
Task-based usability testing
Usability testing means watching real people attempt real tasks while they think aloud. It is the most direct way to see why something fails: the visitor who reads “Solutions” as a software product, or scrolls past the price because it looks like an advert.
You don’t need a lab. NN/g’s long-standing advice is that testing with about five participants per round uncovers most of the problems a qualitative test will find, and that several small rounds with fixes in between beat one big study. Distinct audiences, such as buyers and distributors, each need their own participants, and so does each language your customers use: a translated label can be accurate and still not be the word people look for.
How to run a basic UX audit yourself
A useful in-house version needs a spreadsheet, a video call tool and a few hours of other people’s time.
1. Pick three to five key tasks
Choose tasks tied to enquiries or revenue, phrased from the visitor’s side. For a manufacturer: check you make a kind of part, confirm certifications, request a quote. For a clinic: find a treatment, understand the cost, book. Search Console queries, your sales team’s common questions and a mapped customer journey show what visitors actually want.
2. Check the data, then walk each task yourself
Start with an hour in analytics: where each task leaks, and on which device, is your baseline for measuring fixes.
Then, on your phone and using mobile data, complete every task as a stranger would. Repeat on a laptop using only the keyboard. Log each problem with a screenshot and the heuristic it breaks. This UX audit checklist covers the most common failures:
For fixes, see common website design mistakes, or website navigation design if most problems involve menus and labels.
3. Write neutral task scenarios
Give participants a goal and a reason, never the route, and avoid words from your menu.
- Leading: “Go to Services, open Precision Machining and fill in the quote form.”
- Neutral: “Your company needs 2,000 custom aluminium brackets a month. Find out whether this supplier could make them, and ask for a price.”
Decide in advance what counts as success, so every session is judged the same way.
4. Recruit real customers
Recent customers, prospects who enquired but didn’t buy, or people who closely match your buyers. Not colleagues, and not friends who want to be kind. Aim for about five per round. With distinct groups, NN/g suggests three or four from each; treat each language as its own group. Remote video calls work well, even on a phone.
5. Watch without helping
Open with: “We’re testing the website, not you.” Ask people to think aloud, then stay quiet. When they ask “Should I click this?”, reply “What would you do if I weren’t here?” The moment you help, you stop learning.
Note every hesitation, backtrack and misreading, plus their exact words. Record with permission: a two-minute clip of a customer failing to find your prices settles internal debates faster than any slide.
6. Rate severity and decide what to fix
Nielsen suggests weighing how many people hit a problem, how badly it hurts and whether it keeps catching them out. He also warns that one person’s ratings are unreliable: have two or three people rate separately, then compare. A simple scale is enough:
| Severity | What it means | What to do |
|---|---|---|
| Critical | Stops people completing a key task | Fix before anything else |
| Serious | Causes long delays or wrong turns for several people | Fix in the next round of work |
| Minor | Irritates but doesn’t block | Batch with other changes |
| Cosmetic | Noticed, with no effect on the task | Fix if it’s cheap |
Then weigh severity against effort. A critical problem that takes an hour to fix goes first. Test again afterwards, because fixing one problem often exposes the next.
What a useful UX audit report contains
If you commission a UX review, the report is the product. A useful one contains:
- Scope and method: the tasks, audiences, devices and languages covered, and who took part.
- A one-page summary of the problems that matter most.
- Evidence for each finding: where it happens, what the visitor experiences, how many participants hit it, related analytics, severity and a specific fix.
- What already works, so a redesign doesn’t break it.
- A prioritised plan separating quick fixes from structural changes.
Be wary of generic advice (“make your CTAs more prominent”), a hundred unranked issues or automated-tool screenshots with no task evidence. Before you sign, ask how many real users will take part, and for a sample report.
From findings to fixes: repair or redesign?
Contained problems can be fixed on the site you have: a confusing label, unexplained pricing, a form with too many fields (see web form design). Fix, measure, repeat. That loop is the heart of conversion rate optimisation and continuous improvement.
Structural problems come from how the site was planned and built: navigation organised around the business rather than the buyer, templates with no room for what people need to know, a mobile layout that fails everywhere, or a CMS nobody can update. Patching these rarely lasts. When the audit keeps pointing at structure, a redesign is the honest fix, and the audit becomes its brief. Our website redesign process shows how that evidence shapes the design.
Frequently asked questions
What is the difference between a UX audit and usability testing?
Usability testing is one method: watching people attempt tasks. A UX audit combines it with expert review and analytics, so it can show where problems occur, how often and why.
Can an automated tool or AI run a UX audit?
Partly. Automated checkers catch some accessibility and performance issues, and at the time of writing (August 2026) AI tools can produce a plausible heuristic-style critique of a page. None can watch your customers try to get something done, so treat their output as a first pass.
Should you run it in-house or hire a specialist?
In-house teams know the business but struggle to see the site as a stranger does. The DIY process catches obvious problems; an outside reviewer adds pattern experience and the freedom to report awkward findings.
What to do next
Pick your key tasks this week, walk them on a phone, then watch five customers attempt them without helping. It is often humbling, and a better start than any opinion.
When the problems are structural, our website redesign service puts your findings to work: we audit first, keep what already works, redirect every old URL and launch without taking your site offline.
To clear the technical basics before testing with people, start with our free website audit. It isn’t a UX audit: a person reviews your speed, mobile experience, Google visibility, security and how easily visitors can get in touch. Fixing those first means your usability sessions surface design problems, not slow pages.