For employers wondering whether Truffle's AI scores a specific signal (language proficiency, tone, confidence, and so on).
Truffle's AI scores the content of a candidate's video responses against the role criteria you've defined. The match percentage, key insights, and Candidate Shorts all come from that analysis. What candidates say, not how they look or sound. It's not an accent checker, a confidence detector, a personality profiler, a lie detector, or a language-proficiency test. For trait-level signals beyond response content, add one of the three assessments.
What Truffle scores
The match percentage runs across 32 competency dimensions calibrated from your job description and any additional context you've given it. Dimensions include areas like Career alignment, Problem-solving approach, Relevant experience, Customer orientation, Communication clarity, Self-awareness, Independent work, Team collaboration. The mix of dimensions used depends on your role.
For each candidate, the AI reads their transcripts and rates them on each relevant dimension. Those scores roll up into the match percentage (0–100%) shown on the candidate profile. The scoring is strictly on the content of what the candidate said.
What it explicitly doesn't score
These are outside the AI's design scope. If these signals matter for your role, Truffle's main match score isn't the tool.
Language proficiency. If a candidate's response is in English but not fluent, the match score scores the content, not fluency. It doesn't grade grammar, vocabulary range, or accent. For roles where language proficiency is a must-have, use a dedicated language test alongside Truffle, or add a qualification question up front ("Would you rate your spoken English as conversational, professional, or native?").
Accent, tone, or voice quality. Not a factor. By design, accent scoring carries high discrimination risk and Truffle doesn't do it.
Confidence, anxiety, or "executive presence." Not scored. The AI doesn't interpret body language, pauses, or hesitation as positive or negative signals.
Truthfulness / lie detection. Truffle doesn't claim to know whether a candidate is telling the truth.
Any protected characteristic. The AI is specifically trained not to factor age, gender, race, disability, nationality, religion, pregnancy, or any protected category into its scoring.
Video quality, lighting, or background. Candidates in quiet offices and candidates recording from their kitchens with kids in the background get the same treatment. Scoring is on transcript content.
When to use an assessment instead
If you need a signal the match score doesn't cover, one of the three optional assessments may be the right tool.
Personality (Big Five traits). Where stable personality traits matter for the role. Validated IPIP questionnaire, scored against trait preferences you set. See How do I add a personality assessment to a position?.
Work Scenarios (situational judgment). Where you want to see how a candidate would handle specific workplace situations. Scenario-ranking format. See How do I add Work Scenarios to a position?.
Environment Fit (work-style preferences). Where alignment with the role's actual pace, structure, or communication style matters. Binary A/B question format. See How do I add Environment Fit to a position?.
Each produces its own independent signal on the candidate profile. See What's the difference between Truffle's assessments? for an overview.
How to make the AI sharper for your role
The match score is a function of two inputs. The candidate's responses, and the role context you've given Truffle.
Write a specific job description. Generic JDs produce generic scoring.
Fill in the AI personalization step. Two short questions. "What does success look like in the first 6–12 months?" and "What traits help someone thrive or struggle on the team?". Measurably improve match score quality. See How do I personalize Truffle's AI for my position?.
Write focused screening questions. A question that asks candidates to demonstrate a specific competency produces sharper per-question scoring than a broad "tell me about yourself". See How does Truffle evaluate each question?.