A conditional form changes its questions when a visitor makes a choice. Testing it means checking those changes as a connected journey: which questions appear, which answers are required, what happens when the visitor changes their mind, and what the receiving system actually accepts. A screenshot of the first screen cannot establish that the other paths work.

For a website owner, the practical starting point is a small record of expected behaviour for each choice. Agree that record before reviewing the interface, then test the visible result as well as the underlying settings. A recent Vinove form correction shows why this distinction matters, even when the change is only a few lines of styling.

A hidden setting did not guarantee a hidden question

Vinove's release record for 15 November 2026 documents a correction to conditional fields on its Contact form. The form used a hidden state to control questions associated with the enquiry topic. A styling rule could still leave a field displayed. The correction gave the form's hidden state precedence over its display rules and added a computed-style regression check.

The useful observation is specific: inspecting whether a field had been marked hidden was not enough to establish what a visitor would see. The browser's final rendering also needed to be checked. This is not a claim that conditional forms are unreliable, or that one styling rule is a complete solution for every implementation.

A fresh browser observation on 17 November checked the live form without submitting it. Selecting Careers displayed the Role question and the careers-link section while hiding Organisation. Selecting investment, partnership, press or the remaining general option reversed that arrangement. The observation checked the visible field states. It did not test delivery, validate an application or establish accessibility conformance.

The underlying browser behaviour is documented independently: MDN's hidden attribute reference explains that a CSS display rule can override the hidden state. That is established platform behaviour, not a new finding by Vinove. The dated release and subsequent observation provide a concrete example of why a test should examine the result rather than only the intended setting.

Write the choice-to-question record first

Begin with the visitor's purpose. Each branch should have a reason to exist and a clear owner who can explain why its extra questions are needed. Avoid making the form responsible for discovering information the receiving team does not use.

For every choice, record five things in plain language:

  • Visible questions: the labels and supporting instructions the visitor should encounter.
  • Required answers: what must be supplied before that branch can continue.
  • Inactive questions: what should disappear or become unavailable when another choice is selected.
  • Changed answers: whether information from a previous choice is retained, cleared or excluded from submission.
  • Next step: what the visitor should see after a successful submission and what the receiving team needs to act.

This record is a design decision, not something to infer from whichever behaviour the current code happens to produce. If the team cannot agree what should happen to an answer when its question becomes irrelevant, testing cannot resolve the disagreement for them.

Keep the record small enough to review beside the form. A short list of branches with explicit differences is more useful than a large specification that repeats the same common fields on every page. Link each exception to the reason it exists, so later changes do not accidentally remove it.

Separate appearance, validation and submission

A question has several states that should not be treated as interchangeable. It can be absent from the visible layout but still exist in the page. It can be visible but optional. Its value can remain in browser memory after the visitor switches to a different branch. The application needs an intentional policy for each case.

Ask three separate questions during review. Can the visitor see and reach the field? Does the current branch require a valid answer? Will its value be included and accepted when the form is submitted? A correct answer to one does not prove the others.

For example, imagine a partnership form that asks for an organisation and a recruitment path that asks for a role. This is an illustrative design exercise, not a description of a completed Vinove submission. If a visitor enters an organisation and then switches to recruitment, the team needs to decide what happens to that earlier answer. Silently treating it as current recruitment information may be wrong; clearing it immediately may also be frustrating if the visitor switches back.

There is no universal retention rule for every form. Choose deliberately, explain surprising behaviour and test the chosen policy. Browser behaviour and server-side acceptance should agree about which information belongs to the selected purpose. Hiding a field is not a substitute for deciding that policy.

Test transitions, not just separate starting points

Opening a fresh form once for each option is a useful first check. It does not cover the person who makes an initial choice, types something, changes their mind and returns to an earlier branch.

Build a few transition paths around realistic uncertainty. Start with one topic, move to another, then return. Repeat after entering values and after encountering a validation error. Check whether old error messages remain attached to questions that are no longer relevant, and whether newly visible questions have the instructions needed to answer them.

Include the initial state before any choice is made. It should be clear what the visitor needs to do first. Also inspect the state reached through an approved link or preselected topic, if the site offers one. Do not assume that a path entered through a link behaves exactly like a path selected manually.

Use both a narrow and a wide viewport. Watch the position of the newly revealed questions, the amount of scrolling required and whether the next action remains easy to find. Record the actual browser and viewport used. A desktop screenshot is not evidence of the mobile path, even when both are generated from the same code.

Check the experience beyond the screenshot

Visual review should be paired with interaction checks. A field that looks correctly placed can still have an unclear label or an unexpected place in the keyboard sequence. Newly revealed content also needs to make sense to a person using assistive technology.

Section508.gov's dynamic-forms guidance treats changing forms as interactive experiences and distinguishes automated checks from manual evaluation of focus, updates and error recovery. Use that guidance as a testing reference, not as grounds to claim that a particular form has passed an accessibility assessment.

For the actual review, name the journeys and assistive-technology combinations that were tested. Record any manual checks that remain outstanding. A report that says exactly what was examined helps the next reviewer reproduce the result; a broad “accessible” label without that evidence does not.

The same restraint applies to delivery. A browser-only field-state check cannot prove that a message reached its intended recipient. Delivery acceptance needs its own authorised test and receipt. Keep those records separate so a successful visual check is not accidentally reported as an end-to-end submission result.

Give a failed check a precise owner and outcome

When a branch fails, describe the trigger, expected result and observed result together. For example: “After changing from topic A to topic B, question C remains visible although the agreed branch record excludes it.” Add the browser, viewport and relevant evidence. Avoid labelling the issue only as “form broken”.

Then identify which decision or implementation needs correction. The problem might be an unclear content rule, conflicting presentation, validation that applies to the wrong branch, or a mismatch in the receiving system. These are different tasks and may belong to different people.

After the correction, repeat the failing transition and the neighbouring paths it could affect. Preserve the original failing case as a regression check where practical. In Vinove's dated example, checking computed display made the test sensitive to the presentation issue that a hidden-state check alone could miss. That narrow check complements broader journey testing; it does not replace it.

Keep the acceptance record useful for the next change

A concise acceptance record should identify the form revision, the agreed branches, the tested transitions and the observed results. Attach the limitations beside the pass result: no submission performed, delivery not tested, or a named manual check still pending. Those statements prevent later readers from assuming a broader review than actually happened.

Keep the record with the release evidence. A future styling, content or validation change can then be assessed against known behaviour. The team can decide which checks need to run again instead of relying on memory of how the form once looked.

For broader acceptance questions, see Vinove's B2B website accessibility guide and release-readiness evidence guide. The conditional-form decision is narrower: make the expected questions explicit, test the changes between them, and report only the behaviour the evidence establishes.