Answering the manual checks

An automated scan measures what can be measured: contrast, missing labels, operating the page with the Tab key. Whether an existing alternative text describes the image accurately, whether the focus order makes sense or whether videos have captions is beyond it. For these, the service asks you manual checks: questions with instructions, whose answers go into the report and into the accessibility statement.

What the manual checks are for

The report puts every success criterion into exactly one of four states:

Of the 50 success criteria of WCAG 2.1 AA, the scan measures 4 completely. All the others can only be listed as fulfilled through your answers. That is why there is at least one question for each of them.

A defect from the measurement always outweighs an answer: if the scan has a finding for a criterion, it stays “with defects”, even if you answered the question on it with “met”.

The same goes for an uncertain result nobody has decided yet: as long as an occurrence for a criterion is listed under “To review”, the criterion stays “not evaluated” – even if you answered the question on it with “met”. The question tells you so and leads to the occurrence: Answering the manual checks.

What the states mean for the statement: The conformance status – and when you may deviate.

Where to find the manual checks

• In the scan result via the “Manual checks” tab.
• From every page of the scan via the bar “Your path to the statement”: as long as a question blocks publication, its button “Answer the questions” leads straight to the first of them. More about the bar: Reading the scan report.
• From the overview via the link that says how many success criteria are still not evaluated. The link only appears if a question in this scan is waiting for your answer.
• In the “Statement” tab, as long as a question blocks publication: the note above the preview of the draft names the open questions, and “To the manual checks” leads straight to the first of them.

A scan from before the complete catalogue of questions only has the questions that existed then. Nothing is added there afterwards; the remaining questions are asked by the next scan of the same website.

Without an account

You can read the questions of a free scan without an account, but not answer them: an answer goes into a statement and needs someone who stands behind it. Instead of the answer buttons you see a note; the button “Create an account and continue” sits in the bar “Your path to the statement” at the top of the page. In the view of a single question, and when the result of the scan is not reliable, the note itself carries it. That moves the scan into your account, and you answer the questions there – your answers then stay available for later scans too. If you do not have an account yet, you create one or sign in on the way; afterwards this tab opens again, with the same scan. You can still assess findings without an account: The free scan without an account.

The tab opens with an overview. At the top you see where you stand, counted in success criteria: how many are assessed and how they are spread across “without defects”, “fulfilled by own assessment”, “with defects” and “not evaluated”. Below that, it says how many questions are answered, followed by a button – “Answer the questions”, later “Continue answering” – that opens the next open question and names it.

What does your website contain?

Many questions concern content that not every website has. So that you do not have to choose “not applicable” several times in a row, “What does your website contain?” lists up to 9 preliminary questions – for example about video and audio, or about forms and individual input and selection controls such as a search field. You answer each with “Yes” or “No”; below each one it says how many questions a No saves you. A preliminary question only appears if it can save at least one question in this scan. In the overview, the section is one line with the state of the preliminary questions; a click expands it. As long as none is answered, it is open.

• No answers the related questions that are still open with “not applicable”. Those questions then no longer stand among the others: at the end of their group, one line says how many are settled by the preliminary questions, and “show” brings them up. They remain reachable and changeable and are marked as answered by the preliminary question. Questions you have already answered yourself stay as they are.
• Yes changes nothing about the questions – you answer them one by one. If you had chosen No before, a Yes reopens exactly the questions the No had answered.

Two limits apply to every No:

• A measurement overrides your own statement. If the scan found such content – a video, for example – the preliminary question says “Found in this scan.”, it is answered with Yes, and No cannot be chosen – not even after you have clicked Yes once more. The reverse does not hold: if the scan finds nothing, that never means No. The scan covers a selection of your pages, not the whole website.
• A No never answers a question that blocks publication. Such a question stays open until you answer it yourself – with “not applicable” where that fits.

One question deliberately hangs on no preliminary question, although you might expect it to: the one about flashing (“Does nothing flash more than three times per second?”). The preliminary question “Does anything move, blink or update by itself?” means what runs without anyone doing anything. But a video your visitors start themselves can flash too – so you always answer this question individually.

Your answers to the preliminary questions appear at the end of the “Findings” tab and in the PDF under “Questions about the website's content”, each with its source: “Self-declaration” or “found in this scan”.

The questions

The catalogue holds 45 questions on the success criteria of WCAG 2.1 AA. They are not ordered by number but sorted into 8 groups by what you do to check them – for example everything that shows up during one pass with the Tab key, or everything for which you fill in a form.

The overview

In the overview, each group is a block with its name and the number of its open questions. The groups are collapsed – except the one holding the next open question. Expanded, a group shows each question as one line: the question and, beside it, where it stands – your answer, or “open”. A click on the line opens the question. Explanations, instructions and answer buttons are not in the overview but with the individual question.

Answering a question

An opened question stands alone on the page. Above it you see which group it belongs to and which question of the group it is. The question shows

• the success criterion it belongs to,
• under “How to check”, the steps that lead you to a judgement,
• behind “What this is about”, a short explanation of the requirement,
• the answer buttons,
• and, where there is something, what the scan has already found – for example which ways to reach a page it detected. That supports your judgement; it does not replace it.

Your answer leads on. A click on an answer saves it and shows the next open question straight away; the sentence below the buttons announces this. The order: first the questions that block publication, then the groups one after the other. Questions settled by a preliminary question are not part of the walk-through. After the last question you are back in the overview.

Once every question is answered, the card with your status becomes the conclusion – “I cannot judge this” counts as a given answer here, too. It states how many questions you answered with “not met” and leads on in two ways under “How would you like to continue?”: “Create the statement” leads with “Continue to the statement” to the “Statement” tab and says beforehand what is still to come there – how many details are missing, or that your draft is ready –, “Download the report” leads to the PDF at the end of the “Findings” tab. If a criterion stays “not evaluated” despite all answers, the reason is given there as well: an uncertain measurement result you still have to decide in the “Findings” tab, or an answer “I cannot judge this”.

The statement depends only on the questions that block publication. Once they are answered, this is stated below the button to the next question even before that – with the link “Continue to the statement”. The bar “Your path to the statement” then also shows the station “Manual checks” as done, although it still counts open questions. You can still go on answering the remaining questions.

Clicked the wrong one? Above the new question you see the one you just answered, with your answer; “change” takes you back to it with one click, and so does “Back” at the foot of the page. There you choose again – or withdraw the answer with “Withdraw answer”; the question then stays where it is.

“Skip” moves on without answering; the question stays open. “To the overview” takes you back at any time, and the overview then shows the line of the question you came from. Every question has its own address: reloading, and the browser's back and forward, land where you were.

When a requirement does not apply to your website is stated with every question, above the answers.

You do not have to answer everything in one go. Every answer is saved immediately.

The five answers

Every question asks “Is this requirement met?”. You can choose:

“I cannot judge this” is an honest answer and better than a guessed “met”. It is not offered for questions that block publication.

If several questions belong to one success criterion, it only counts as fulfilled by your own assessment once all of them are answered with “met” or “not applicable”.

The note appears in the scan report next to your answer, not in the statement. Open the field with “Add a note”; if there is a note already, it is open. Use it to record what you checked and what your answer rests on. In the statement, a fixed sentence names the defect for “not met”. If you write the note before clicking the answer, it is saved together with it. If you write it afterwards – you are then already at the next question – go back with “change”; the note is saved as soon as you leave the field, and below the field it then says “Note saved.”. A note alone does not move you on. It may be at most 2,000 characters long.

What blocks publication

Not the whole catalogue. Publication is only blocked while one of the questions listed at the top of the tab under “Required for publication” is open. Each of them is marked “Must be answered before publication”. There can be at most 6:

The questions about captions and about more than one way also block when they blocked in the previous scan and you answered them there. Your answer is then carried over, so nothing is open. That also holds for the question about more than one way: if you answered it with “met”, it stays that way even if the next scan again detects fewer than two ways: Answering the manual checks.

The service names the questions that are still open – in the “Statement” tab above the preview of the draft and on the publishing page – and leads from there to the first of them.

Once you have answered one of these questions, it no longer appears at the top of the overview but as a line in its group.

All other questions are voluntary. You can publish without answering them. The report and the statement show the consequence openly: the criteria concerned count as not evaluated, the statement names their number, and the status then cannot be “fully conformant”: The conformance status – and when you may deviate.

Why a question can start blocking

Whether the questions about captions and about more than one way block is decided anew by every scan. The captions question only blocks once a scan finds a video on the pages it examines to which it cannot match any captions – not simply because you added one. If you have already answered the question yourself – not with “I cannot judge this” – your answer stays and nothing is open. Otherwise it blocks the next publication until it is answered. A statement that is already published is not affected.

The question about explanations in German Sign Language and Easy Language is only put to public bodies in Germany, and only where federal law or the law of their state requires these explanations. As long as nothing is chosen under “Federal or state level”, it is asked. If the chosen law does not require the explanations, the question is dropped from the next scan on, even if it had already been answered – an earlier scan keeps it.

What carries over to the next scan

On the next scan of the same website, the service carries over your answers including notes. They are then marked “Carried over from the previous scan”. Your answers to the preliminary questions carry over as well; a No again answers the related open questions in the new scan. You do not start from scratch on every run. A note on a question you have not answered yet carries over as well.

Three exceptions:

• A measurement proves the opposite of your answer. This only exists for the question about more than one way, and only in one direction: you had answered “not met”, and the new scan detects two ways. Your answer is then not carried over; the question is open again, shows which ways were detected, and asks you to answer once more. If you stay with “not met” – because the search does not cover every page, for example – the service does not ask again in the following scans. The reverse does not hold: if you answered “met” and a scan detects fewer than two ways, your answer stays. The scan has then merely found nothing – it does not recognise a search behind a magnifier icon, for example – and so has refuted nothing. What it did detect is still shown at the question. If you do not answer a question reopened this way, it stays open with its notice in the following scans too and blocks publication until you have answered.
• The question blocks publication in the new scan. Then neither “I cannot judge this” is carried over nor an answer that came from a preliminary question. The question is open, and you answer it yourself.
• A preliminary question is now answered by a measurement. If you had chosen No and the new scan finds such content, Yes applies; the questions your No had answered are open again.

What a self-assessment is – and what it is not

Your answers are a self-assessment: the own statement of the body making the declaration. They are not an audit by a third party and not an expert opinion. The report, the statement and the Accessibility Conformance Report therefore mark them as what they are:

• The report lists “fulfilled by own assessment” next to “without defects”, never inside it.
• The statement says that it is based on a self-assessment and names how many success criteria were not evaluated.
• The Accessibility Conformance Report notes on every row that rests on your answer that it is self-assessed.

Whether a self-assessment is enough for your legal duty is not something the service judges. It replaces neither legal advice nor a full test by people.

Manual checks and WordPress

The WordPress plugin only shows a published statement. Before the first publication, the shortcode therefore shows no text as long as a question blocks it. If a version is already published, it stays visible; only the next version is blocked. In the plugin you answer the questions that block publication; the preliminary questions and all other questions are answered in the service. More: Publishing the statement from WordPress and a notice instead of the statement in WordPress.

Not to be confused with “to review”

Uncertain results of the automated scan appear in the report in their own section. Those are spots on your pages, grouped into building blocks, not questions about the whole website. They do not block publication, but they do not count as passed either: Assessing uncertain results.

The two are connected at one point: as long as such an occurrence is undecided, its success criterion stays “not evaluated”, however you answer the question on it. The question then carries a notice with the link “To the results under “To review””. Decide the occurrence as a defect or as no defect – only then does your answer count.