The best website audit tools in 2026, and where each loses
Website audit tools are not competitors. They are four families measuring four different things, and buying from the wrong one is why your last audit went unused.
Most website audit tools are not competitors. They belong to four families that measure four different things, and buying from the wrong family is why your last audit produced a list nobody acted on.
Every roundup ranks them against each other anyway. Semrush against Screaming Frog against Hotjar, as though one of them wins.
None of them wins. Only one of them answers the question you asked.
Which website audit tool do you actually need?
Match the family to your question. Crawlers answer whether the site is technically correct. Lab speed tools answer how fast one page loaded on one run. Behaviour tools answer what visitors did. Conversion audits answer which problem is costing you the most money.
Founders almost always want the fourth and buy the first, because the first is what appears when you search for a website audit.
- Crawlers. Screaming Frog, Semrush Site Audit, Ahrefs. Broken links, duplicate titles, redirects, orphaned pages.
- Lab speed tools. PageSpeed Insights, Lighthouse, WebPageTest. One URL, one simulated device, one run.
- Behaviour tools. Microsoft Clarity, Hotjar. Clicks, scrolls and recordings, with no reasons attached.
- Conversion audits. Revslip, and the human CRO audit it partly replaces. Ranked problems with money attached.
What does a crawler like Screaming Frog find?
A crawler walks your pages and reports what is mechanically wrong: broken links, duplicate titles, missing meta descriptions, bad redirects, pages nothing links to. That is the whole of its published feature list. None of it measures whether a visitor understood what you sell.
It is the best free thing in the category, with one hard edge. Screaming Frog's pricing page caps the free version at 500 URLs per crawl and charges $279 per licence per year to lift it. A 40 page site never notices. A 4,000 page store hits the wall on the first run.
Worth saying: a crawler that comes back clean is good news. It means the leak is somewhere a crawler cannot look.
Want the other kind of list? Revslip ranks findings by cost.
Can you trust a single PageSpeed score?
No. Google's own Lighthouse documentation states that performance scores change because of inherent variability in web and network technologies even when nothing on the page has changed, and it tells you to run the test several times and beware of that variability before concluding anything.
Read that again, because every audit report you have ever been sent contains exactly one score.
The variability document names the causes: A/B tests changing the page between runs, network conditions, the hardware you tested on, and browser nondeterminism, which it rates as certain in every environment. Its fix is the median of five runs, because the median of five is twice as stable as one.
A single performance score is one sample from a distribution, presented as a measurement.
PageSpeed has a real job: comparing your page against itself over time. Producing a number for a slide is not it.
What do behaviour tools show, and what can they never see?
Clarity and Hotjar record what people did: where they clicked, how far they scrolled, where the session ended. That genuinely finds the screen where attention dies. What they cannot record is why, because intention leaves no trace in a click log.
We covered the structural limits in why switching heatmap tools rarely helps. A heatmap tells you the button was ignored. It has no opinion on whether the button was wrong, the price was wrong, or the visitor was never going to buy.
How much does one automated tool miss?
More than any roundup admits. The UK Government Digital Service built a page carrying 142 deliberately planted accessibility barriers and ran 13 automated checkers across it to compare them. The best tool found 40% of the barriers. The worst found 13%.
- 142 barriers deliberately planted on one page
- 40% caught by the best of 13 automated checkers
- 13% caught by the worst
Two caveats before anyone quotes that back at us. It is an accessibility test rather than a conversion test, and the GDS results page was last updated in April 2018, so individual tools have moved. What survives is structural: one automated pass over a hard page returns a fraction of what is on it, which is why GDS told its own teams to combine tools rather than trust one.
Our own numbers carry the same shape of problem. Revslip has audited 134 businesses, and every one arrived because somebody already suspected the site was leaking. That is not a random sample of the web.
Where does Revslip lose?
In three places, and they are limits of the whole automated family rather than bugs in ours. Revslip cannot judge whether your offer is worth its price. It has no history on your site, so it sees today and not the trend. Its euro figures are a ceiling, not a forecast.
Which of those limits should send you to a different tool, and which rival is the better buy in each case, is set out in Revslip alternatives.
The formula is published so you can check it by hand: traffic × CVR gap × AOV × mobile weight. Take 8,000 monthly visitors, a 0.9 point gap against your own best page, and a €70 average order. That is 72 orders, about €5,040 a month. If the gap is optimistic, so is the answer.
Read this section against us We sell the fourth family, so an article arguing the fourth family matters most is the article we would write regardless. The test is whether the sources survive being clicked. Two of the three above sell nothing in this category, and one is a government report we have no control over.
In what order should you run them?
Broadest and cheapest first, then narrow. Each pass removes a category of explanation, so the next tool starts from a shorter list. Reverse the order and you pay expert rates for problems a free crawler hands you in four minutes.
- Crawl it. Screaming Frog free, or your platform's built in report.
- Measure speed properly. PageSpeed five times, median only.
- Price what is left. A conversion audit, ours or a person's, to rank the survivors by money.
- Watch people last. Recordings once you have a hypothesis, because open ended watching is where weeks go.
Step three, free: paste your URL and see the ranked list.
When none of this is your problem
No tool in any family helps when the question sits upstream of the page. If you cannot yet say who the product is for, an audit returns a tidy list of fixes to a page that should not exist yet.
Same when the traffic is the fault rather than the site. If one source drags your average down while the rest convert normally, the four check split settles it faster than any tool will.
Questions people ask when choosing a tool
These four come up constantly, and they share one root. The category name promises a whole audit, while each product in it delivers a slice.
Is a free website audit tool enough to start with?
Usually yes, for the mechanical layer. What varies is what each one checks, and how much of it is a form pretending to be a scanner. What free website audits actually check goes through it.
Do I really need more than one tool?
If you have a real question, no. Pick the family that answers it. The case for running several is coverage, which the GDS numbers support, and it applies when you have no hypothesis at all.
What does a paid audit add over the free pass?
Judgement and prioritisation. What a CRO audit costs in 2026 covers the price bands, and the machine versus consultant split covers which half each is good at.
Will one tool find every leak on its own?
No. Revslip checks roughly 200 conversion signals across copy, offer, trust, UX, technical and tracking, and still misses judgement calls. The conversion leak index ranks the common leaks by cost.