# Accessible B2B Login: Review Password Managers, Pasting, and Every Authentication Step

<a id="main-content"></a>

Operations &amp; Systems

October 28, 2026  ·  8 min read  ·  Victor Valentine Romo

On this page

1. [Put the Existing Account at the Start](<#put-the-existing-account-at-the-start>)
2. [Name the Criterion and Its Four Provisions](<#name-the-criterion-and-its-four-provisions>)
3. [Prepare Evidence Without Collecting Secrets](<#prepare-evidence-without-collecting-secrets>)
4. [Run the Password and Code Checks Separately](<#run-the-password-and-code-checks-separately>)
5. [Work Through a Fictional Supplier Login](<#work-through-a-fictional-supplier-login>)
6. [Copy the Authentication Worksheet](<#copy-the-authentication-worksheet>)
7. [Check the Alternative From Its Own Starting State](<#check-the-alternative-from-its-own-starting-state>)
8. [Return a Finding Another Reviewer Can Repeat](<#return-a-finding-another-reviewer-can-repeat>)
9. [Sources](<#sources>)

A password field accepts a pasted value. The next screen rejects a pasted verification code. The login review is still open.

Treat authentication as a route through the product. Write down what a person must remember, solve, recognize, or transcribe at each stop. Then record the assistance or alternative available at that stop.

The worksheet below reviews an existing account in a synthetic environment. It makes no claim about a particular vendor, customer, or completed accessibility assessment. Use it to give a product owner a reproducible question and a bounded decision.

<a id="put-the-existing-account-at-the-start"></a>

## Put the Existing Account at the Start

Begin the review with a test identity that already exists. Define the intended destination, such as a fictional supplier document area. Keep account creation outside this review.

W3C's [Understanding Accessible Authentication guidance](<https://www.w3.org/WAI/WCAG22/Understanding/accessible-authentication-minimum.html>) focuses on existing users. It includes authentication during recovery and requires attention to every step. The Understanding document is informative guidance. The linked success criterion defines the requirement.

Draw the ordinary route first: account identifier, password, verification, destination. Add separate branches for a forgotten password, [an expired code](</articles/session-timeouts-b2b-portal-accessibility/>), and a challenge that appears after an unsuccessful attempt. Mark branches you cannot exercise as untested.

Use one row for each distinct demand. A password and a verification code belong in separate rows even when the interface places them together. Their assistance mechanisms can behave differently.

Keep permission questions in a separate record. [An access review](</articles/small-team-access-review-evidence/>) asks whether an identity should retain access. This review asks what the identity must do to authenticate. A retained permission does not answer that question.

<a id="name-the-criterion-and-its-four-provisions"></a>

## Name the Criterion and Its Four Provisions

[WCAG 2.2 SC 3.3.8, Level AA](<https://www.w3.org/TR/WCAG22/#accessible-authentication-minimum>) restricts required cognitive tests at authentication steps. It permits a noncognitive alternative, assistance, object recognition, or recognition of non-text content previously supplied by the user. Those last two provisions can involve images, audio, or video.

In the worksheet, distinguish an alternative method from assistance within the current method. Write the actual action in either case. “Another option exists” leaves too much unresolved.

The informative [Understanding guidance](<https://www.w3.org/WAI/WCAG22/Understanding/accessible-authentication-minimum.html>) describes password-manager support and pasting as assistance. Remembering passwords, solving puzzles, and transcribing codes can create cognitive demands. An audio transcription challenge still involves transcription. Personal text does not qualify as the personal-content exception. Object recognition does not cover an added calculation involving those objects.

Do not stretch an exception to fit a convenient label. If the product owner selects “personal content,” ask what the user supplied. If the answer is a memorized phrase, record the mismatch for review.

Keep the claim narrow. This article supplies a review method for one criterion. It does not establish whole-site conformance, legal compliance, or behavior across every browser and [assistive technology](</articles/accessible-status-messages-b2b-websites/>).

<a id="prepare-evidence-without-collecting-secrets"></a>

## Prepare Evidence Without Collecting Secrets

Use a designated test environment, invented records, and test credentials approved for that environment. Do not ask a customer to reveal a password. Do not paste a real recovery code into a ticket.

Give each observation a run identifier. Record the product build, browser version, operating system, and password-manager version when one is used. Include the date and the path selected. Those details describe the conditions of the observation.

Record results as behavior: both fields received values, submission remained available, the code field rejected the paste, or the next screen appeared. The values themselves add nothing to that finding.

For screenshots, use a synthetic interface containing invented labels and dummy values. Crop unrelated windows before capture. A finding about a paste handler does not need an inbox, a personal profile, or a credential list beside it.

Keep the reproduction steps separate from the evidence attachment. A reviewer should understand the defect without opening an image containing sensitive content. If the attachment needs a secret to make sense, redesign the attachment.

<a id="run-the-password-and-code-checks-separately"></a>

## Run the Password and Code Checks Separately

For a synthetic password check, start with empty fields. Try the configured test password manager. Record whether it identifies the intended fields, fills them, and leaves a usable submission path.

Reset the fixture before the next attempt. Paste the dummy identifier and password from an approved local test source. Record the result independently from the manager attempt. A successful fill in one configuration does not prove every entry method works.

Continue through the next screen. If the fixture requests a code, paste the entire dummy code. Check the resulting field state before submission. In a split-field design, record whether every position receives the intended character.

W3C's [informative guidance](<https://www.w3.org/WAI/WCAG22/Understanding/accessible-authentication-minimum.html>) specifically discusses code pasting and split inputs. It assumes the code is available on the clipboard for evaluating the web content. It does not assess transfer between devices.

Keep that boundary visible in your report. Write “whole test code accepted from clipboard in this run.” Avoid replacing it with a broad claim that the organization has solved every second-device problem.

Use [the lead-form review](</articles/accessible-b2b-lead-forms/>) for labels, instructions, and error repair. Record those findings separately when the same screen raises them.

<a id="work-through-a-fictional-supplier-login"></a>

## Work Through a Fictional Supplier Login

Consider an invented portal called Supplier Review. Its existing test identity opens a document area through an identifier, a password, and a six-position verification screen. These are fictional observations, not a live product test.

In run A, the test password manager fills the identifier and password. Submission opens the verification screen. Pasting a six-character dummy code fills only the first position. The remaining positions stay blank.

The fixture then asks the reviewer to type the remaining characters. No other verification method appears. The reviewer stops the ordinary route there and records the unanswered assistance question.

In run B, the reviewer starts from the initial screen and pastes the password manually. That step succeeds again. The same code behavior follows. Two successful password attempts do not repair the later step.

The fixture also provides a recovery link. Selecting it opens a question requesting a remembered project phrase. No assistance or alternative appears in this invented branch. The recovery row stays open with its own evidence.

Do not count the recovery link as an alternative until its route reaches the intended authenticated destination. In this example, it introduces another demand. Moving the barrier to a different screen leaves the finding unresolved.

The decision is specific: revise the code-entry behavior and review the recovery branch. Nothing in these invented results justifies removing a security control or changing the organization's authentication policy.

<a id="copy-the-authentication-worksheet"></a>

## Copy the Authentication Worksheet

Use a small header for the environment, run identifier, reviewer, date, starting state, and intended destination. Then use one row per step.

| Field | What to record |
| --- | --- |
| Step | Screen, branch, and required action |
| Cognitive demand | Recall, transcription, puzzle, recognition, or none observed |
| Assistance | Mechanism attempted and its observed behavior |
| Alternative | Separate method and tested destination, or none found |
| Exception evidence | Exact provision and supporting behavior, or not established |
| Observed result | What occurred under the recorded conditions |
| Decision | Accept the bounded observation, investigate, revise, or retest |

For the fictional supplier fixture, the completed rows read:

| Step | Cognitive demand | Assistance | Alternative | Exception evidence | Observed result | Decision |
| --- | --- | --- | --- | --- | --- | --- |
| Password | Recall | Test manager fill and manual paste | None reviewed | Assistance observed | Next screen opens in both runs | Retain scoped evidence |
| Verification | Transcription | Whole-code paste attempted | None found | None established | Only first position receives input | Revise and retest |
| Recovery | Recall of project phrase | None found | None found | Personal-content provision unsupported | Branch remains blocked | Investigate another usable path |

These rows organize the review. They are not a W3C-mandated template. Add the evidence location without copying any credential into the table.

If a branch is unavailable in the fixture, write “not exercised.” [An empty cell](</articles/null-blank-zero-business-data/>) hides whether the reviewer found nothing or never looked.

<a id="check-the-alternative-from-its-own-starting-state"></a>

## Check the Alternative From Its Own Starting State

Ask the product owner to identify each supported alternative. Repeat the route using the starting conditions that method requires. Record whether the test identity can select it before encountering the unresolved demand.

Suppose the fictional portal gains a test-device confirmation option. Define a new run with that device already configured. Record the selection, confirmation, and destination. Leave unsupported configurations outside the conclusion.

Do not describe that proposed option as implemented until it appears in the tested build. Keep a proposed remedy in the decision column and an observed result in the evidence column.

Review the handoff between methods too. If the alternative sends the test identity back to the same broken code screen, record the loop. A second button can lead to the same unresolved step.

Ask who can use the alternative under the agreed review scope. If it requires earlier setup, document that prerequisite. Keep that setup question distinct from a claim about new-account creation.

Use [the keyboard review](</articles/keyboard-navigation-b2b-website-review/>) to examine how someone reaches and activates the option. This worksheet still owns the cognitive-demand question.

<a id="return-a-finding-another-reviewer-can-repeat"></a>

## Return a Finding Another Reviewer Can Repeat

[Write a short finding](</articles/b2b-usability-findings-triage/>) with the environment, exact branch, attempted assistance, observed outcome, and review question. Link it to the row and the synthetic evidence.

For the fictional code screen, describe the six-position input and the whole-code paste attempt. State that only the first position receives input. Ask the owner to provide usable assistance or establish another supported path.

Avoid a vague ticket such as “login is inaccessible.” That wording omits the step, the conditions, and the action a reviewer needs to repeat. It also invites a fix at the password screen that leaves the code screen untouched.

After a revision, rerun the route from its starting state. Confirm the destination and revisit the other branches included in the agreed scope. Keep the original observation beside the new result.

Close only the rows whose evidence supports closure. List remaining branches, configurations, and assistance mechanisms as untested. A release decision can then distinguish a demonstrated change from an unanswered question.

Keep every authentication step visible.

<a id="sources"></a>

## Sources

- [W3C: WCAG 2.2, SC 3.3.8 Accessible Authentication (Minimum)](<https://www.w3.org/TR/WCAG22/#accessible-authentication-minimum>). Normative criterion and provisions. Checked 2026.10.28.
- [W3C: Understanding Accessible Authentication (Minimum)](<https://www.w3.org/WAI/WCAG22/Understanding/accessible-authentication-minimum.html>). Informative scope, assistance, transcription, and exception guidance. Checked 2026.10.28.
