How to audit a website properly

Most website audits are a tour of someone's opinions. The colours feel dated, the hero could be stronger, the copy is a bit long. All possibly true, none of it a reason to spend money. A proper audit is not a list of things a stranger would change. It is a clear picture of where the site is losing people, and what to do about it first. The method matters, and it runs in a particular order.

Where should a website audit start?

It starts with the data, not the design. Before forming a single opinion, look at what is actually happening. Where does traffic come from, where does it land, and where does it leave. Session recordings and heatmaps show you the moments people hesitate. Funnel analysis shows you the step where they give up. The design conversation is far more useful once you know which page is quietly bleeding.

How do you tell what's broken from what's taste?

From there, the discipline is separating what is broken from what is merely not to taste. A form that fails on mobile is broken. A button colour you would have chosen differently is taste. Only one of those is worth a line in the report. Most audits blur the two, which is how a client ends up repainting the walls while the roof leaks.

How does the site read to a buyer?

Then read the site the way a buyer does. Walk the actual journey. Arrive from a cold search, with the question a real buyer has, and see whether the site answers it in the order they need. Most sites are organised around how the company is structured, not how the buyer thinks, and that mismatch is usually the biggest single problem. It never shows up in a checklist of components.

Can the site be found by search and AI?

It is worth checking whether the site can be found at all. A site that converts brilliantly and cannot be found is a shop with the lights off. Look at the basics of search and answer-engine visibility. Is the structure legible to a crawler, is there structured data, does the site turn up when an AI is asked the question it should own. Visibility problems are cheap to miss and expensive to ignore.

What should an audit end with?

Above all, a real audit ends with a decision, not a document. The test is not how long it is. It is whether, at the end, you know the one or two things to change first and why. If the report leaves you with forty tickets and no priority, it has described the problem without helping you solve it. Data before opinion, broken before taste, buyer before org chart, found before beautiful, decision before document. It is not complicated. It is just rarely done in that order.