Ask users for sex and gender

When to use

Ask only when a specific decision in your product depends on the answer, and be precise about which of two distinct things you need.

• Sex, in the clinical sense, is needed because some reference ranges, screening programmes and risk calculations differ. What is clinically relevant is usually narrower than "sex": it is often a specific organ, a specific hormone profile, or a specific screening eligibility.
• Gender is needed to address the reader correctly, and to know which pronouns and title to use. It is an identity question, not a clinical one.

These are two fields or none. A single "Gender: Male / Female" control used to select a haemoglobin reference range is the most common version of this mistake, and it produces both a clinical error and an exclusion at the same time.

When not to use

• You are personalising content or a greeting. You need a name and possibly pronouns; you do not need sex. See Name.
• You need one specific clinical fact. Ask for that fact. "Do you have periods?" or "Are you pregnant or might you be?" answer the actual question more accurately than any sex field, are easier to answer honestly, and do not require the reader to disclose an identity to a product that did not need it.
• The answer changes nothing you do. Do not collect it. This field carries a real cost for trans, non-binary and intersex readers, and collecting it without a use is a cost with no benefit.
• You want the wording rules. Asking sensitive questions.

How it works

flowchart TD
A["We think we need sex or gender"] --> B{"What exactly changes?"}
B -->|"a reference range or screening eligibility"| C{"Can we ask the specific clinical question instead?"}
C -->|"yes"| D["Ask that: periods, pregnancy, a specific organ, current HRT"]
C -->|"no"| E["Ask sex for clinical purposes, and say so at the field"]
B -->|"how we address the reader"| F["Ask for pronouns and title, not sex"]
B -->|"nothing specific"| G["Do not ask"]
E --> H["Allow intersex, prefer not to say, and a route to talk to a human"]
F --> I["Free text plus common options; never a closed two-item list"]

Practical rules:

• Label the clinical field for its purpose, not its identity. "Sex recorded for your clinical records" or "Sex used for reference ranges" tells the reader which question they are answering and why the answer might not be the one they would give socially.
• Never a two-item closed list. Intersex people exist; readers who do not wish to disclose exist; and a reader whose clinical history is more complex than a binary needs a route that is not "pick the closest".
• Gender is free text with common options offered, not an enumeration imposed on the reader. Where you need pronouns, ask for pronouns. It is a smaller, more useful and less intrusive question.
• Never derive one from the other. A gender answer must never select a reference range, and a clinical sex answer must never set a pronoun.
• Both must be editable, at any time, without contacting support. People transition; records are corrected. A field of this kind that is write-once is a design that expects the reader not to change.
• Say what happens to the answer. Who sees it, whether it appears in an export, and whether it is used to select what is shown. See Sharing with a clinician.
• A reader whose situation does not fit needs an exit that is not a form field. Provide a route to a human, and do not block progress on the answer.

Content

That last "don't" is the failure this page exists to prevent. Forcing a reader to trade accurate care against accurate address is a design failure, and it is entirely avoidable by keeping the two fields separate.

Accessibility

• autocomplete="sex" exists in the HTML specification's autofill list and applies to the administrative field; there is no token for gender identity or pronouns, and none should be invented. WCAG 2.2 SC 1.3.5.
• A radio group needs a fieldset and legend carrying the question, so every option is announced in context.
• "Prefer not to say" is a real option in the group, not an absence. Leaving the group unanswered and treating that as the refusal is invisible to a screen-reader user and indistinguishable from a mistake.
• The explanation precedes the options in the DOM and is referenced by aria-describedby on the group.
• Free-text alternatives are labelled controls, not an "other" radio that reveals an unlabelled box.
• Do not re-ask on subsequent sessions (WCAG 2.2 SC 3.3.7); do make it editable.

Research

Updates to this page

Last read through against the system on 2026-09-20. Due for review every 6 months; expiry is reported by pnpm run check:freshness.