Website audit — what we actually check, what it costs and what you get out

Most audit offers describe the same thing as "comprehensive analysis". This text says what we check, in what order, what it costs and what you get at the end — so it can be compared against another offer rather than merely believed.

Start with the distinction that determines the value of everything else.

An audit is not a scan from a tool

Free tools — Lighthouse, SEO crawlers, browser extensions — do one thing well: they show symptoms. They report that the site loads slowly, that image descriptions are missing, that headings are out of order, and turn it into a PDF in thirty seconds.

What they will not do: say why. A tool states the site is slow. It does not state that it loads a three-megabyte script from a plugin nobody has used in two years, and that removing it clears half the list at once.

Hence a simple rule when choosing a supplier: if somebody offers an audit for free, they have probably dropped your address into a machine, generated a report and want to sell a retainer. We price our time, because we check what a machine cannot see — a person's work, not a script's.

The tools are not worthless. They are a starting point, not a result.

Three layers, and an order that is not a matter of taste

The order follows from dependency, not preference. Work in a higher layer cannot take effect if the layer below is broken. Rewriting the content of a site the search engine does not show is the most expensive way of not solving the problem — and the most common.

Layer 1: technical

This is the foundation and the first thing we check:

• Whether every URL responds with what it should. A page that exists answers 200, a deleted one 404, a moved one 301 or 308. A mismatch between code and content is a defect with no symptom at all.
• Speed on a phone, not on an office connection. Measured in the conditions the site is actually opened in.
• Whether a crawler can fetch the site — robots.txt, the sitemap, accidental blocks left over from a deployment.
• Certificate and redirects — whether all traffic lands on one address, over HTTPS, with no duplicate between the www and non-www versions.
• Whether the analytics works at all, and whether it counts your own traffic as customers.
• Whether every language version answers the same way. On a site running in two or three languages, each version has its own addresses, its own redirects and its own way of breaking.

The most common finding here is not spectacular: the site works, but loads scripts nobody added on purpose — leftovers from a previous supplier, analytics wired in twice, fonts from three sources. Each is harmless alone; together they add up to seconds of load time.

One thing on this layer is worth separating from the technical check, because it is not a technical question: what the measurement is allowed to collect. A Swiss company that serves customers in the EU is bound by the GDPR for that processing, and Swiss data protection has its own rules besides. We do not read the policy text and we do not give legal opinions — we check the mechanism: what loads before anyone clicks the banner, what loads only after, and whether the two match what the banner claims. Where they do not, that is a finding for a lawyer, and we write it as one.

Layer 2: visibility

Here we answer the question visit statistics cannot pose: does the search engine show this site at all?

• How many pages are indexed and how many skipped — and with what verdict, because "not discovered" and "crawled and rejected" are different problems with different solutions.
• Which queries the site appears for, and whether they are buyers' queries or accidental.
• Whether pages compete with each other for the same phrase — because then none wins.
• Whether the redirects put in after an address change actually work. Checked from the URLs the search engine knows, not from a list of rules somebody once wrote.
• Whether each language version is indexed on its own. One version carrying the whole site is a common pattern, and it looks healthy in a summary.

It is worth knowing the four verdicts a search engine uses to describe a page, because each leads somewhere different:

Checking this takes a quarter of an hour and settles whether the work goes into content, into linking, or somewhere else entirely. Without that distinction, "we have no traffic" is a diagnosis about as useful as "it hurts".

Layer 3: content and usability

Only here does a conversation about what the site says make sense:

• Whether it answers the questions that come up in sales meetings — price, timing, who this is for, what happens after signing.
• Whether it is possible to get in touch without seven form fields, and whether the form delivers the message.
• Whether the structure reflects the customer's journey, not the firm's org chart.
• Accessibility — keyboard operation and legible contrast. Whether it is also an obligation depends on the kind of organisation you are and who you serve; either way it is a share of visitors, not a formality.

The check we always run and which always turns something up: we walk the customer's path from arrival to sending an enquiry, on a phone, not knowing where anything is. Somebody who knows the site cannot do this — and it is the only way to see where an outsider gets lost.

Language versions are separate sites, and they fail one at a time

On a market where a company site routinely runs in German and French, sometimes with Italian or English beside them, this is not a detail inside the three layers. It is the dimension that cuts across all of them, and the reason the layers above turn up clean while a third of the site is invisible.

The mechanism is simple and it is why the problem hides. Each language version has its own addresses, so each has its own indexing, its own redirects and its own broken form. Nothing forces them to fail together. A report averaged across the whole domain shows one healthy site; the French version can be entirely absent from it and the average barely moves.

What we check, and what a scan will not tell you:

• Whether the versions declare each other. The hreflang annotations tell a search engine that these are the same page in another language, rather than duplicates competing with one another. Missing or contradictory, and the engine picks one version and drops the rest.
• Whether every version is actually indexed — not the domain as a whole, but each language separately. This is the single most common finding on a multilingual site.
• Whether the switcher lands where it says. The classic defect is a language switch that takes you to the home page of the other version instead of the same page, from every page on the site. Nothing errors; people simply leave.
• Whether a version is complete. Half-translated sites are normal after a few years of additions. Worth knowing which half, and whether the untranslated pages are the ones customers arrive on.
• Whether the form works in each version — including the confirmation message, and including who receives an enquiry sent from the French page.

None of this is exotic work. It is the same three layers, run once per language, and it is the part most often skipped because the first pass looked fine.

A website audit and an SEO audit are not the same thing

Both names circulate interchangeably, and that is where most confusion in comparing offers comes from.

An SEO audit covers one layer: visibility. It answers why the search engine does not show the site high, and ends with recommendations on content, keywords and links.

A website audit covers all three layers, including what happens after arrival — whether the form works, whether the site works on a phone, whether the analytics measures the truth.

The distinction has a practical consequence: a firm with traffic but no enquiries does not need an SEO audit. The traffic is there — the problem is a floor above, and an SEO audit will not touch it. The reverse also holds: a firm the search engine does not show may have an excellent site and zero results.

If an offer does not say which of these you are buying, ask before signing.

A technical website audit — what exactly it covers

This name appears most often in offers and is used interchangeably with "website audit", though it means something narrower.

A technical audit is the first of the three layers above — whether the site works correctly as software and whether the search engine can read it. It does not judge the content, the offer, or whether the site persuades anyone to get in touch.

Specifically: server availability and response times, HTTP codes, redirects and redirect chains, canonical addresses, sitemap and robots file, page indexability, certificate and mixed content, mobile load speed, structured data, error handling.

It is the most measurable layer and therefore the easiest to sell — and the only one automated tools can partly perform. If an offer says "technical audit" at a full-audit price, ask outright about the other two layers. Sometimes they are simply not in scope.

Three findings a site owner will not spot alone

All three come from audits we ran — including one of our own site, where we found the most.

A URL that answers "everything is fine" and shows an error. Eleven addresses in our corpus answered 200 for sixteen months while serving an error page. The user saw the page was gone; the crawler was told everything worked, so it had no reason to report or replace it. Nobody noticed: statistics do not show addresses nobody visits, and the tools that check response codes saw nothing but 200s.

A hundred and twenty-four articles the search engine never showed once. Of 182 published texts on our blog, 124 had not a single impression over three months. Not "a poor position" — complete absence. Visit statistics showed this as "nobody reads it", which leads to the conclusion that the content is weak. The real problem was a floor below and needed entirely different work.

Analytics counting our own team as customers. One run of automated tests against production added 98 false hits before anyone noticed — on a site of our size, enough to change which pages look important and to send the next quarter's work somewhere useless. We wrote it up with three other measurement traps in four ways your numbers lie.

None of the three required a failure. All three came from default settings, or from a change after which nobody checked what happened to the URLs.

When an audit makes sense and when it is an expense

Not every site needs one, and we would rather say so before the sale.

In the first two cases on the second list we say so in conversation and do not sell an audit. The third leads straight to a conversation about a project.

There is also a question of how often. There is no reason to run one every year off the calendar — a routine repeat describes the same site in the same words. Three moments are worth it: after a change of system or of addresses, after a noticeable drop in traffic, and before a decision on a larger investment. Everything else is a habit, not a diagnosis.

What you can check yourself before ordering anything

The first layer can largely be walked on your own, and there is no reason not to. We built a self-audit checklist for exactly that — no registration, no email address.

Five checks to start with:

1. Walk the site on your own phone — the home page, one offer page, the contact form.
2. Open Search Console, range three months, and look at two numbers: impressions and clicks. Many impressions with few clicks is a titles problem. No impressions is an indexing problem.
3. Send an enquiry through your own form and check whether it arrived, to whom, and after how long.
4. Do the same in every language version you run — switch language on a deep page, see where you land, then send that version's form too.
5. Check who has access to the domain, the hosting, the system and the analytics.

If those five come out well, you probably need work on the content, not a technical audit. That is a result too, and we would rather say it than sell a service.

How the audit runs

Five working days and three things on your side. Worth knowing what you are agreeing to, because two of them can move the date.

Day 0 — access. We need read rights to Search Console and Analytics, plus the site address. Not the server, not the content management system; if either turns out to be needed for something specific, we ask separately and say what for. This is the stage that most often delays the start — not because anything is difficult, but because nobody at the firm knows who holds the access.

Days 1–2 — technical and visibility. Here the tools work and so do we. A crawler collects data on every URL; we check the mismatches: addresses answering with something other than what they should, redirects that do not work, pages competing for the same phrase.

Days 3–4 — content and the customer path. Walking the site by hand on a phone, testing the form, checking whether it answers the questions sales gets asked. This part cannot be automated, which is why it costs the most. On a multilingual site it is also the part that is done more than once — and the reason we ask up front how many language versions are live, rather than counting them ourselves on day three.

Day 5 — document and conversation. You get the report, then an hour for questions. The conversation is part of the audit, not an extra: half the value appears when you can ask "and what exactly does that mean for us".

What we do not do along the way: change anything on your site. An audit is read-only. If we find something needing immediate action — an unresponsive page, an expired certificate — we say so at once, not on day five.

What a free audit gives you, and what it does not

The free audit is an industry standard, and it is worth saying plainly what it is.

A free audit is a sales tool and there is nothing improper about that — provided both sides know it. Its purpose is to show that problems exist, specifically enough to build trust and generally enough that they cannot be fixed without buying the service. It is usually an export from an automated tool, wrapped in a conversation.

What you actually get from it:

• A list of machine-detected first-layer defects — real value, because some genuinely need fixing.
• The scale of the problem: whether there are five or two hundred.
• A reason for a conversation in which an offer will be made.

What you will not get from it, which matters more:

• Priority. A tool lists fifty observations and never says which three matter. That is precisely what a paid audit is for.
• A judgement of content and usability. The third layer cannot be measured by machine, so a free audit does not contain it.
• The answer "you do not need this service". An audit that is a sales tool cannot end that way.

When it is worth taking: when you want to see the scale and get to know a supplier before paying anyone. That is a sensible use, and we have two tools for it with no registration — a self-audit checklist and a website speed test.

When it is not worth stopping there: when the site has traffic and no enquiries. The answer lies in the second and third layers, which no free audit covers.

How to compare two audit offers

Six questions that settle whether you are comparing the same thing. Each follows from something above.

1. "Which of the three layers does this price cover?" A cheap offer and one at four times the price can both be honest — if the first covers SEO alone and the second the whole thing. Without this question you compare numbers, not scope.
2. "How many language versions does this price cover?" On a two- or three-language site this changes the amount of work more than anything else on the list, and it is the item most often left unsaid. An offer priced for one version and delivered against three will be renegotiated halfway through.
3. "How much is done by a tool and how much by a person?" If the answer is "we have advanced software", you are buying an export from a machine you could run yourself.
4. "Do I get a list of priorities or a list of defects?" Three hundred problems with no order is useless — and easy to generate.
5. "What of the recommendations can I do myself?" An honest answer always includes some part. "Everything requires our team" says more about the sales model than about your site.
6. "Is the audit fee deducted if I commission the fixes?" With us, yes. Worth asking everywhere, because the answer changes the real cost of the decision.

What it costs

CHF 1,250 net, the result in five working days, an hour discussing the report. There is one price and we give it before the start, not after.

If you decide on a project with us afterwards, we deduct that CHF 1,250 from the quote. The audit exists so that you know what to do — not so that you pay twice for the same diagnosis.

What the price does not include: implementing the recommendations. The audit ends with a document and a conversation; fixing is a separate decision and a separate quote, including when somebody else does it. The report is written so that it can be handed to anyone — your current supplier included. That is deliberate: a report only you can act on is a sales document with a different name.

What you get at the end

A document, not a tool export. Inside:

• A list of findings with priorities — what to fix first and what can wait a quarter, with the reasoning behind the order.
• What can be done in-house, separated from what needs a supplier. Some recommendations usually do not need us, and that is written down explicitly.
• An effort estimate against every item, so it can be budgeted rather than merely read.
• An hour of conversation walking through the document and answering questions.

What it does not contain: a hundred pages of machine-generated charts. A report is meant to be read, not weighed.

And one outcome worth naming in advance, because people assume it cannot happen: sometimes nothing is broken. Then that is what the document says. "The site is technically sound, the problem is in the content" is a result worth paying for — it is the sentence that saves the cost of a rebuild that would have changed nothing.