When you create a custom screening question, Truffle runs an automatic check on it and rates it pass, warn, or fail. Pass adds the question silently. Warn flags it with a suggested rewrite and lets you accept the rewrite, edit, or proceed with a recorded acknowledgement. Fail means the question can't be added as written — you have to edit it, and there's no override. The check surfaces questions that reference protected characteristics (age, race, gender, disability, national origin, religion, pregnancy, and similar) or are phrased in a way that could be read as doing so.
The three tiers
Every custom question goes through the same check when you click Create question in the custom-question tab.
Pass. The question is clear. It's added with no flag shown.
Warn. The question is likely fine but could be read as touching on a protected characteristic. Truffle flags it as a "Potential bias risk" and suggests a safer rewrite. You can accept the rewrite, edit the question, or proceed with a recorded acknowledgement (optionally noting a business reason).
Fail. The question directly references a protected characteristic or is otherwise high-risk. Truffle shows "This question can't be added as written" with a required change, and the only option is Edit question. A fail can't be acknowledged or overridden — you have to rewrite it before it's added.
While the check runs, you'll see a status message in the dialog.
What to do when a question is flagged
When you see a warn or fail message, the dialog shows the original question and a suggested rephrase.
Read the suggested rewrite. Truffle's rewrite usually preserves what you're trying to learn about the candidate and removes the phrasing that caused the flag.
Accept the rewrite if it still captures what you want to ask. The safer version gets added.
Edit the question manually if the rewrite isn't quite right. Change the wording, then re-run the check.
For a warn, you can also proceed with the question as written by acknowledging the risk. Truffle records the acknowledgement, and you can add a short business reason. A fail doesn't offer this — you have to edit it before it can be added.
Why Truffle flags these questions
Screening questions get shown to every candidate, so a single risky phrasing can reach dozens or hundreds of people before a reviewer notices. The check surfaces the obvious cases (direct questions about age, family status, or health) and the subtle cases (questions that correlate with a protected characteristic by proxy).
AI flags. Humans decide. Every decision to keep, rewrite, or override a question stays with you.
Disclaimer
Truffle's question check is an automated pattern-matching tool, not legal advice. It does not certify that your questions are compliant with any specific employment law, regulation, or case-law precedent in your jurisdiction. Employment law varies by country, state, industry, and role. Local rules, collective bargaining agreements, and industry-specific requirements may apply beyond what this check looks for.
You are responsible for the final content and legality of every question you publish. For roles or jurisdictions with complex requirements, have your custom questions reviewed by a qualified attorney or HR professional before publishing. Truffle makes no warranty that the classifications (pass, warn, fail) are accurate or complete in every context, and is not liable for any decision you make to add, rewrite, or override a flagged question.
When the check doesn't run
The check fires on custom questions only. Questions picked from Truffle's built-in question catalog have already been reviewed and don't get re-checked. AI-suggested questions generated from your job description are pre-screened before they're shown to you.