How We Verify The Work

What we tested, what it rests on, and what it does not prove.

A report that gives you a number and no method asks you to take its word for it. This page shows the method. It covers what software settles, what a person has to look at, and where we stop.

A clinician explaining a document to a seated patient across a desk.
Every finding in your report names the tool and the published standard behind it, so your own web team can re-run it.
Short answer

How does OptiSite verify the work it reports?

Every finding says what was tested and what it is based on: a rule from the accessibility standard that courts and regulators point to (called WCAG 2.2 AA), published search guidance, or an OptiSite recommendation for the patient experience. Checks software cannot settle are flagged for a person to review and are never recorded as a pass. A clean automated result is reported as "nothing found," not as proof that your site meets the standard.

Get My Free Audit
  • Every finding names its tool and its source
  • Software-checked, person-checked and advisory items are labeled separately
  • A clean scan is reported as nothing found, not as proof
  • Read a complete sample report (PDF)
FirstA machine reads itSettles what software can settle, on every page it reaches.
ThenA browser uses itPresses the buttons, follows the links, records where a patient lands.
LastA person looks at itTakes what is left. Flagged for review, never scored as a pass.
Three kinds of evidence. We label each one instead of adding them into a single number.
In Order

Each layer answers what the one before it could not.

A scanner cannot tell whether the order a keyboard moves through the page makes sense to a blind patient. A person cannot check four thousand items by hand. Telling you which layer produced a finding lets you separate a measured failure from a judgment call, and both from a guess.

What Actually Runs

Three different things, and they are not interchangeable.

Most website reports blur these together and present the total as one score. Yours keeps them apart, because they carry different weight.

A machine reads the pageAutomated inspection

Our scanner opens every page it can reach. It tests what software can settle on its own: a picture with no written description, a form that is not protected, a broken link, a page that loads too slowly.

  • Runs on every page the scanner can reach, not a sample
  • Each finding names the tool that produced it
  • Repeatable: your web team can run the same tools and get the same answer
A browser uses the pageInteraction checks

Some things only fail when something is actually clicked. A real browser presses the buttons and follows the links to see where a patient ends up.

  • Book and Call buttons are followed to the page they lead to
  • We try to move through the page using only a keyboard
  • If a step is blocked, we record what we saw and which button or link failed
A person looks at the pageHuman review

The checks no scanner can settle. These appear in your report as items needing review — never counted against your site, and never counted as passes.

  • Whether the order a keyboard moves through the page makes sense to someone who cannot see the screen
  • Whether a button's label on screen matches what a screen reader says aloud (a screen reader is software that reads a page out loud)
  • Whether an error message tells a patient how to fix the problem
A phone showing a Google search for eyecare beside a monitor showing an AI assistant’s answer.
Where A Finding Comes From

Four kinds of backing, each one labeled.

A form box with no label breaks a published accessibility rule. A thin service page does not. Both are worth fixing, but only one breaks a standard. Your report says which is which. You can tell legal risk apart from our opinion about what works.

The Four Labels

What each finding rests on.

Accessibility ruleWCAG 2.2 Level A and AA

A specific rule from the published accessibility standard, named by number, with the pages where it failed. This is the category the lawsuits are about.

  • The strictest level (called AAA) is not assessed and is not required by law
  • The rule number is printed on the finding
  • The report states which pages the scanner reached
Published search guidanceGoogle's own documentation

Advice a search company has published. Following it makes your pages easier to understand. It does not buy you a spot in the results, and we never say it does.

  • Advice, not a guaranteed way to rank higher
  • Google's own documentation says the same basic practices apply to its AI results
  • We claim no secret formula anywhere in your report
Patient-experience recommendationOptiSite's judgment

Our opinion, based on watching how patients use practice websites. Examples: a Book button you can see right away, insurance answers that do not need a phone call, and a portal link that goes where it says.

  • Labeled as our recommendation, not an external requirement
  • You are free to disagree with any of it
  • Never presented as a law, a rule or a way to rank higher
AdvisoryEmerging, unsettled

Ideas the industry is still debating, such as new file types for AI assistants (tools like ChatGPT that answer patient questions). Nobody has agreed on a standard yet. Worth knowing about. Not counted as a failure.

  • Kept separate from measured failures
  • Does not affect your score
  • Flagged so you can decide for yourself
The Honest Part

What your report does not prove.

Someone in this industry will make each of these claims to you. We will not make any of them.

It is not a certificate of complianceAccessibility

A clean automated result means we looked for that problem and did not find it on the pages we reached. Whether a site meets the standard is a judgment about the whole site, and part of it can only be settled by a person.

  • Nothing found is not the same as proof
  • We do not issue certifications and neither can a scanner
  • Items needing a person are listed, not quietly passed
It is not a promise of trafficSearch and AI

We can remove what stops search tools from understanding your practice. Where you rank after that, and whether an AI assistant names you, is up to them.

  • No ranking is promised anywhere in your report
  • We measure how ready your site is. We do not predict how visible it will be
  • Nobody can guarantee an AI recommendation, including us
It is not a count of lost patientsFindings

A finding can show that a patient hit an obstacle. It cannot tell you how many patients left because of it. We will not invent a dollar figure to make the point land harder.

  • Obstacles we saw, not estimated lost revenue
  • No projected patient counts appear in your report
  • The occurrence count is places found, not people lost
It is not an independent auditWho ran it

We run the audit and we build the new site, so we benefit if you say yes. We show you an audit you can verify anyway, with the tool named on every measured finding.

  • Your own web team can re-run it
  • Same engine, same rules, on your site and on the one we build
  • If our rebuild scores worse somewhere, it shows
Somebody Else's Code

When the problem is not on your website.

Your scheduler, your portal and your payment pages belong to other companies. We can show you exactly what a patient runs into, and we can fix how your site hands them over. We cannot change software we do not control. And a vendor's failure does not shift your responsibility onto the vendor.

Questions, Answered

More on this page's topic

Is this an independent audit?
No, and we do not claim it is. OptiSite runs the audit and OptiSite builds the new site, so we benefit if you say yes. What we offer instead is a report you can check. Every measured finding names the tool and the published standard behind it, so your own web team can re-run it and get the same answer.
Does a clean result mean my website is ADA compliant?
No. An automated pass means we looked for that problem and did not find it on the pages we could reach. Whether a site meets the standard is a judgment about the whole site, and parts of it can only be settled by a person. Anyone who tells you a scan makes you compliant is selling you something we would not sign our name to.
What does a person actually look at?
The checks software cannot settle. Does the order a keyboard moves through the page make sense? Does a button's label on screen match what a screen reader (software that reads a page aloud) says? Does an error message tell a patient how to fix the problem? These appear in your report as items needing review. They are not counted against your site, and they are not counted as passes either.
Can you guarantee a Google ranking or that an AI assistant names my practice?
No. We can make your doctors, services, locations and answers easy for search tools to read, and we check that nothing is blocking them. Whether Google ranks you, or an AI assistant names you, is their decision, and nobody can promise it. Any company that does is guessing.
What happens when the problem is not on my website?
Your scheduler, your portal and your payment pages belong to other vendors. We can show you exactly what a patient runs into, and we can fix how your site links to them, but we cannot change software we do not control. Findings like that are labeled so you know who has to act.
Do you re-check the site after it is built?
Yes. The same audit runs on your new site every month, and what it finds gets fixed inside the monthly price. The monthly run is automated. A person reads the result before anything is changed on your site.

See the method on your own website.

The free audit runs the same way this page describes, on the site you have now. No card, no obligation.