Skip to main content

TalkScore Reading Detection

Identify candidates who may be reading their responses during voice assessments.

Perry Santiago avatar
Written by Perry Santiago
Updated today

Reading Detection is a feature within TalkScore that helps identify whether a candidate may be reading their responses during a voice assessment. While some answers may sound clear and fluent, they don’t always reflect a candidate’s actual speaking ability, especially when responses are scripted.

This feature focuses on how a candidate speaks rather than what they say. By analyzing speech patterns such as pacing, pauses, and rhythm, it can flag responses that appear to be read aloud rather than spoken naturally. These insights provide recruiters with additional context when reviewing assessments, enabling them to make more confident and informed decisions about a candidate’s communication skills.

Score Ranges and Labels

Here’s how the labels are applied based on the Overall Reading Score:

Label

Overall Reading Score (ORS)

Color

Likely Reading

70 or higher

🔴 Red

Chance of Reading

Between 30 and 69

🟠 Orange

Natural Speech

Below 30

(No label shown)

This tool is available only for open-ended (free speech) questions, meaning it does not apply to “listen and repeat” or multiple-choice questions. Each open-ended answer receives a Reading Score, and these are combined to calculate an Overall Reading Score (ORS) for the assessment. If the ORS falls within specific ranges, the system adds a color-coded label to the candidate’s profile to help recruiters spot responses that may have been read aloud.

In the Candidate’s General View, the Overall Reading Score appears on the left side. If the score is 30 or higher, a label will be shown to help you identify whether the candidate may have been reading. When flagged, a short explanation is provided.

In the Detailed View, you’ll see a Reading Score for each open-ended answer. This makes it easier to find which specific responses may have triggered the flag.

Reading detection is also saved as a candidate attribute called Talkscore Reading, with one of the following values:

  • Likely Reading

  • Chance of Reading

  • Natural Speech

This attribute can be used for filtering candidates, tracking patterns, or setting up automation inside your workflow.

Note:

If a candidate only answers one open-ended question, the Overall Reading Score will be based on that single answer. In cases where the audio data is incomplete or the question type isn't supported, a score may not be shown.


Q: Does the reading flag affect the candidate’s TalkScore?

No. The reading flag does not change the candidate’s scores for Pronunciation, Fluency, Grammar, or other categories. It is simply an additional indicator to help you interpret spoken responses more effectively.

Q: Can I use the reading flag in automation or filters?

A: Yes. The result is saved in a candidate attribute called Talkscore Reading, which you can use in Autoflow or filters to take actions based on the candidate’s speaking behavior.

Q: Will accents or regional dialects affect the reading detection?

A: No. The detection is based on timing and speech patterns—not on how words are pronounced. Candidates with strong accents or regional dialects will not be incorrectly flagged for reading.

Q: How does the system handle candidates with speech impediments or stutters?

A: The system is designed to be cautious in these cases. Stuttering typically results in longer pauses and shorter speaking segments, which are more commonly associated with natural speech than with reading.

Q: How do I enable the Reading Detection feature?

A: To enable this feature, it must first be activated in your company’s Admin settings. Please contact your Account Manager, who can assist with the setup and make sure it’s properly configured for your team.


Did this answer your question?