Try for FREE
Respondent quality

Real people, verified at every step.

Every answer comes from a real person. AI, algorithms and human review check who is answering, how they answer and what they say.

1
Respondent verificationWe check who is answering.
2
Session verificationWe check every session in the app.
3
Response scanningWe scan every open-text answer.
4
Human reviewPeople review everything that is flagged.
1. Respondent verification

We check who is answering.

Automatic checks on each community member’s account and device identify and exclude bad actors. We also verify that every member’s email address is genuine.

Member verificationVerified
  • Geographic region
  • IP address background check
  • VPN check
  • Email check
  • PayPal verification
  • Device time zone
  • Device type
  • Device model and ID

The results are cross-referenced against each other.

2. Session verification

We check every session.

Every session in the OnePulse app is checked for authenticity and security. This helps us detect and prevent unauthorized access and protects the integrity of the data.

Session checksEvery session
  • Cryptographic checks on every operation
  • Answers confirmed as coming from the official OnePulse mobile app
  • IP address background check
  • Geographic region verification
  • VPN check
3. Response scanning

We scan every open-text answer.

Every open-text answer is scanned for answers that don’t fit the question, gibberish and keyboard smashing. Anything flagged goes to our community team.

Question 3 · Open textWhy would you do that?
“Orange is orange. If it costs more, I’ll grab whatever’s cheaper.”Passed
“dfgkjh sdfg”Flagged: keyboard smashing
“Yes good nice ok blue”Flagged: gibberish
“My flight was delayed today.”Flagged: doesn’t fit the question
AI detection

Built to catch AI-written answers.

We built our own behavioral fraud detection system. By analyzing suspected AI responses, we identified the consistent behavioral signals that give them away.

1
Signals are capturedOur system captures these behavioral signals as each answer is given.
2
A risk score is calculatedThe signals are combined into one weighted risk score that shows whether an answer is AI-generated.
3
Caught at submissionAI-assisted behavior is detected at the point of submission. Every result is stored for ongoing monitoring and refinement.
4. Human review

People check what the systems flag.

Our research shows that human oversight gives a fairer and more accurate assessment of answers than AI alone.

Every flagged answer is reviewed

Our dedicated community team reviews every flagged answer and takes action.

Poor behavior leads to a ban

Members who give low-quality open-text answers or show signs of inattention are temporarily banned.

Manual spot checks

Ad hoc manual checks on profiles and answers validate our automated systems and keep refining them.

Profile accuracy

Profiles that stay accurate.

Every profile answer has an expiry date and is automatically reconfirmed with the member, so the audience you target stays accurate over time.

Smoking statusExample
Non-smokerReconfirmed every 6 months
SmokerReconfirmed every month

Until an answer is reconfirmed, the member receives no studies that target it.

More in the Trust CenterSecurity and data protectionWhere your data lives, who can see it and how we comply with the GDPR.More in the Trust CenterResponsible AIHow OnePulse AI keeps every claim tied to real answers, and how it handles your data.

Questions about security, privacy or data quality?

Email us and our team will answer.

info@onepulse.comBook a demo