AI makes surveys and interviews more personal and efficient. At feedbk.ai, we use AI so that real people have a better user experience. But: What happens when AI bots themselves participate in surveys and corrupt the results?

A recent study shows: AI systems can perfectly imitate human survey respondents. This manipulation risk threatens the integrity of research in market and opinion research.

The Problem: AI Can Manipulate Surveys

According to a report in Der Standard, large language models (LLMs) are capable of:

• Perfectly imitating human respondents
• Manipulating survey results in specific directions
• Doing this without leaving detectable traces

The consequence: Researchers and companies can no longer be certain whether their data comes from real people or was generated by AI bots.

For feedbk.ai, which uses AI to increase the quality of interviews, it’s especially important that only real humans participate. We don’t want manipulated data – we want authentic feedback.

The Solution: Tests Humans Can Do, But AI (Still) Can’t

To distinguish real humans from AI bots, feedbk.ai uses a multi-layered strategy. One component is the Clock Test.

Source: clockbench.ai

What is the Clock Test?

The Clock Test is a visual test where participants must read the time on an analog clock. Sounds simple? For humans, yes – for AI, surprisingly difficult.

A recent study by the ClockBench research group shows:

• Weak Performance: Even the best AI models achieve only 30% exact match accuracy in clock reading
• Systematic Errors: All tested models often misread even simple times like 2:00, 4:00, or 5:00
• Deep-Seated Issues: AI has fundamental difficulties with detecting hand angles and interpreting them

Humans, on the other hand, can often easily read analog clocks (90%). This cognitive ability is – as of today – a reliable indicator of human intelligence.

Why we use multiple tests

The Clock Test is just one of many measures we use to prevent bot participation. Real humans show natural behavioral patterns: they hesitate, correct themselves, think. AI bots often respond too quickly, too consistently, too perfectly.

Important: The Clock Test alone is not a perfect solution. AI is evolving rapidly. That’s why feedbk.ai uses an adaptive, multi-layered system that is continuously improved.

Why Bot Protection is Essential

feedbk.ai uses AI to make interviews more personal. We want:

• Real people to benefit from intelligent follow-up questions
• Researchers and companies to trust authentic data
• AI to serve as a tool, not a manipulation instrument

Anyone using AI surveys must ensure that the results are genuine. Only then can AI reach its full potential: better insights, deeper conversations, authentic feedback.

Conclusion: The Future of AI Surveys is Authentic

AI bots are getting better at imitating humans. Through tests like the Clock Test, behavioral analysis, and adaptive security mechanisms, we ensure that our AI-moderated interviews come from real people.

Expert Insight: In 2026, it’s no longer enough to just create good surveys. You must also ensure they’re not filled out by bots. feedbk.ai combines both: intelligent AI moderation and robust bot protection.

Sources

• Der Standard: “AIs Can Manipulate Surveys Without Leaving a Trace”
• ClockBench: AI Performance on Analog Clock Reading
• ArXiv: ClockQA Research Paper