Vendors · Monitoring platforms

Monitoring at Scale Is a Different Job

Siteimprove gets grouped with overlays and it should not be. It crawls, it reports, and it does not touch your pages, so its limits are the ordinary limits of automation rather than anything worse.

Last reviewed August 30, 2026

What it is

Siteimprove is a web governance platform. It crawls your site on a schedule and reports on accessibility, search, content quality, and broken links, tracks the numbers over time, and gives content teams a way to assign and follow issues. The accessibility scanning now covers PDFs and images as well as pages, which most monitoring platforms historically did not.

It does not inject a repair script or a visitor toolbar into your site, and that distinction is the reason this page exists. One qualification is worth carrying so nobody widens the claim too far. The analytics module, like any analytics product, does need a snippet on your pages. That snippet measures. It does not remediate, and visitors never see it.

The product has moved since most descriptions of it were written. There is a plugin for designers working in Figma, checks that run in editors and in a build pipeline, and newer agent products that surface and remediate PDF and image problems. So it is no longer only the thing that crawls your site overnight and mails you a number.

What it does well

It looks at thousands of pages over and over, which no human team can do and no audit is trying to. That is automation used for the thing automation is genuinely good at.
Trend data across months is useful to an organization running a program, because the question there is whether the site is getting better, not what is wrong on Tuesday.
The monitoring product reports problems rather than claiming to have fixed them. A tool that tells you what is wrong is a fundamentally different purchase from a script that says it handled it. The company does sell separately licensed remediation tooling alongside it, so check which of the two you are being quoted for.
Catching a contrast or target-size problem in the design file, before it reaches sixty pages, is the cheapest moment there is, and moving the check to the designer is what makes that possible.

What it cannot do

It is automated, so it inherits the ceiling. Our own editorial classification grades 356 of the 432 techniques and failures we publish, and puts 10 as fully machine-checkable, with 228 visible to a scanner only in part. Read that as our reading rather than a measurement of any product, including this one. We have not tested Siteimprove against it, and no classification of ours sets a limit on what somebody else's engineering can reach.
Any score or percentage is the tool's own invention. WCAG has no scoring formula at all. It defines conformance for full pages and for complete processes, and it requires every responsive variation of a page to conform, which is more than a crawler at one width has looked at.
It will not walk a purchase flow, trip a keyboard trap inside a modal, or decide whether alt text describes the thing the image is doing.
Nothing is priced in public, and the buyer this is built for has a procurement process rather than a credit card.
Lighthouse and WAVE both print the caveat inside their own output, telling you automated detection cannot establish accessibility on its own. The Siteimprove marketing page we read does not carry it, though the company's support documentation does say automated and manual testing are both needed. So the sentence exists, it just is not next to the number. Read the dashboard with it in mind.

Our read

This is a sound tool doing an honest job, and it pairs with human auditing rather than competing with it. Machines are good at watching a thousand pages forever. People are good at judgment. Use each for the thing it is actually good at and neither of them looks disappointing.

The mistake to avoid is reading the dashboard as a verdict. A score is a weighted summary, not a statement that every machine-checkable rule passed, and whatever it says applies only to the pages it visited at the width it used. It also says nothing about the criteria no published test rule reaches. Count W3C's own ACT rules against the 55 Level A and AA criteria and 24 of them have no live rule behind them at all, which is a gap in the published rule catalogue rather than proof that no tool can check them.

The newer remediation agents are worth watching rather than assuming. A platform that has started writing fixes is moving into the judgment half of the split its own success rested on, and that is a different product from the one described above.

What we would do instead

Sources

We have no commercial relationship with any company named on this page, and no affiliate arrangements. We sell audits, so we compete with some of them, and you should weigh what we say accordingly. Nothing here is legal advice. Product names and trademarks belong to their owners.

Other vendors

Whatever you are running, find out where you stand.

The free scan checks 10 pages in a real browser and reports what your markup actually contains. No widget, no subscription, no signup.

Run the free scan