---
type: source
id: frank-2024-representative-study
title: A Representative Study on Human Detection of Artificially Generated Media Across Countries
authors:
- Joel Frank
- Franziska Herbert
- Jonas Ricker
- Lea Schönherr
- Thorsten Eisenhofer
- Asja Fischer
- Markus Dürmuth
- Thorsten Holz
year: 2024
venue: IEEE Symposium on Security and Privacy (S&P 2024)
volume: null
pages: null
doi: null
url: https://arxiv.org/abs/2312.05976
arxiv_id: '2312.05976'
peer_review: peer-reviewed-conference
evidence_tier: B
design: Online survey with representative samples in USA, Germany, China; audio, image and text; regression on individual predictors
modality:
- image-face
- audio
- text
stimuli: State-of-the-art (2023) generated audio, images and text vs human-made counterparts
n_participants: 3002
n_note: ''
topics:
- human-detection
- cross-cultural
- hyperrealism
summary: Most participants were effectively guessing; AI-generated media was rated more human-like than real media across all media types and countries.
results:
- metric: Overall
  value: near guessing
  note: abstract gives no percentages
  source_location: abstract
- metric: AI vs real 'human-like' ratings
  value: AI rated more human-like
  note: all media types, all three countries
  source_location: abstract
- metric: Predictors
  value: generalised trust, cognitive reflection, self-reported deepfake familiarity significantly influenced decisions
  note: across all media
  source_location: abstract
moderators: []
limitations:
- No video
- Fixed stimulus set
verification:
  status: verified-abstract
  confidence: high
  checked: 2026-10-07
  via: arXiv 2312.05976 abstract; S&P 2024 venue and DOI not confirmed
pmid: null
pmcid: null
openalex_id: null
oa_url: null
cited_in:
- MR-0001
updated: '2026-10-07'
---

# A Representative Study on Human Detection of Artificially Generated Media Across Countries

> **Source record** `frank-2024-representative-study` · 2024 · IEEE Symposium on Security and Privacy (S&P 2024)  
> **Evidence tier:** B (peer-reviewed-conference) · **Verification:** verified-abstract (compiler confidence: high)  
> **Bottom line:** Most participants were effectively guessing; AI-generated media was rated more human-like than real media across all media types and countries.

## Citation {#citation}

Joel Frank, Franziska Herbert, Jonas Ricker, Lea Schönherr, Thorsten Eisenhofer, Asja Fischer, Markus Dürmuth, Thorsten Holz (2024). A Representative Study on Human Detection of Artificially Generated Media Across Countries. *IEEE Symposium on Security and Privacy (S&P 2024)*. https://arxiv.org/abs/2312.05976

Verification entry (identifiers, resolvers, claims with source locations): [/data/citations.json#frank-2024-representative-study](/data/citations.json#frank-2024-representative-study)

## Study design {#design}

| Field | Value |
|---|---|
| Design | Online survey with representative samples in USA, Germany, China; audio, image and text; regression on individual predictors |
| Modality | image-face, audio, text |
| Stimuli | State-of-the-art (2023) generated audio, images and text vs human-made counterparts |
| Participants (N) | 3002 |
| Topics | human-detection, cross-cultural, hyperrealism |

## Results {#results}

| Metric | Value | Note |
|---|---|---|
| Overall | near guessing | abstract gives no percentages |
| AI vs real 'human-like' ratings | AI rated more human-like | all media types, all three countries |
| Predictors | generalised trust, cognitive reflection, self-reported deepfake familiarity significantly influenced decisions | across all media |

## Limitations {#limitations}

- No video
- Fixed stimulus set

## Cited in {#cited-in}

- [MR-0001](/metareports/human-detection-accuracy.md): Human Detection of Deepfakes: Accuracy, Bias, Confidence, and Training

## Verification {#verification}

| Field | Value |
|---|---|
| Status | `verified-abstract` |
| Compiler confidence | high |
| Checked on | 2026-10-07 |
| Via | arXiv 2312.05976 abstract; S&P 2024 venue and DOI not confirmed |

Values on this page have not all been checked against the full text. Confirm them against the primary source before high-stakes use.