AI Acceptance Criteria Reviewer

AI Requirement Analysis

Scores each criterion, points out vague terms, missing negative paths and untestable statements, and proposes improved wording.

Privacy: Requires a free Mutqan account. Your input is sent to the Mutqan AI service to generate the result; prompts and results are not stored — only your account's usage counts for the daily allowance. How the AI tools work

What are acceptance criteria?

Acceptance criteria are the conditions a user story must satisfy to be considered done, agreed before development starts and written so that each can be checked: "Given a registered user with an expired card, when they pay, then the order is not created and the card error is shown." Each criterion becomes at least one test, which is how testers, developers and product owners end up testing the same thing.

Read the guide: How to Write Test Cases

How to use AI Acceptance Criteria Reviewer

  1. Paste the acceptance criteria — numbered, bulleted or as Given/When/Then blocks.
  2. Optionally add the user story they belong to as context.
  3. Pick the output language and press Review.
  4. Start with the lowest-scoring criteria: read the problems, adopt the improved wording and add the missing criteria the review lists. Download the review as CSV, Markdown or JSON.

AI Acceptance Criteria Reviewer features

  • Testability score per criterion and overall
  • Detects vague terms, untestable statements, compound criteria, missing data and missing negative paths
  • Improved wording for every weak criterion
  • List of criteria that are missing entirely

AI Acceptance Criteria Reviewer example

Review

Input:

The page should load fast.

Output:

AC-1 score 20 — vague, untestable → "The lookup page renders its first 50 values within 2 seconds on 4G (p95)."
The rewrite makes load fast testable: it names the page, what the user sees and a limit in seconds, then adds the network and percentile the original never stated.

Before: AC-1, score 20 — vague, untestable

Part 1:The page should Part 2:load Part 3:fast.

After: the improved wording

Part 1:The lookup page Part 2:renders its first 50 values Part 3:within 2 seconds Part 4:on 4G Part 5:(p95).

  1. 1Which pageThe page becomes a named page.
  2. 2Observable resultload does not say what the user sees; the rewrite names what appears.
  3. 3Measurable limitfast cannot be measured, which makes it vague; 2 seconds can.
  4. 4NetworkMissing before: the connection the limit applies to.
  5. 5PercentileMissing before: p95 means 95% of loads must meet the limit.

Frequently asked questions about AI Acceptance Criteria Reviewer

How is the score calculated?

Each criterion is scored 0–100 for testability and completeness; criteria without problems score 85 or more. The overall score summarises the whole set.

What makes a criterion vague?

Words that cannot be measured — "fast", "user-friendly", "appropriate", "etc." — several outcomes in one statement, and criteria without concrete data values or an error path. The rewritten version replaces them with actors, inputs, observable results and measurable limits.

What are the "missing criteria"?

Criteria the set clearly needs but does not have — typically negative paths, permissions or error handling — listed separately so you can add them to the story.

Technical notes

Each criterion gets an ID (AC-1, AC-2 … in the order you pasted them), a 0–100 testability score, the issues found — vague, untestable, incomplete, ambiguous, compound, missing negative path, missing data or inconsistent, or "ok" — a plain-language explanation and a rewritten, testable version. The set also gets an overall score and a list of criteria that are missing entirely.

References