---
type: source
id: rehman-2025-training-synthetic-faces
title: Training humans for synthetic face image detection
authors:
- Ramlah Sara Rehman
- Ewald Meier
- Mathias Ibsen
- Christian Rathgeb
- Robert Nichols
- Christoph Busch
year: 2025
venue: Frontiers in Artificial Intelligence
volume: '8'
pages: '1568267'
doi: 10.3389/frai.2025.1568267
peer_review: peer-reviewed-journal
evidence_tier: B
design: 'Online randomised experiment: 16 trials, then artefact/semantic-region training (experimental) or a 5-min break (control), then 16 trials'
modality:
- image-face
stimuli: StyleGAN2 and diffusion (dezgo.com) synthetic faces plus FFHQ and online real faces; 8 GAN and 8 diffusion images in total
n_participants: 184
n_note: Experimental n=95, control n=89 (254 started)
topics:
- human-detection
- training
- criterion-shift
summary: One-shot artefact training raised trained participants' accuracy on synthetic faces (79.08% vs 58.29% for controls) but lowered accuracy on real faces, so overall accuracy did not differ significantly between groups.
results:
- metric: Overall accuracy, experimental pre vs post
  value: 67.43% vs 71.12%
  note: p = 0.01
  source_location: null
- metric: Overall accuracy, control pre vs post
  value: 68.12% vs 68.33%
  note: p = 0.88
  source_location: null
- metric: Post-intervention, experimental vs control
  value: difference 2.79 points
  note: p = 0.08, n.s.
  source_location: null
- metric: Post-intervention synthetic faces correct
  value: experimental 79.08% vs control 58.29%
  note: p < 0.001
  source_location: null
- metric: Post-intervention real faces correct
  value: experimental lower than control (control 78.37%)
  note: 't(182) = -5.37, p < 0.001: consistent with a shift toward ''synthetic'''
  source_location: null
- metric: Pre-intervention accuracy on GAN vs diffusion fakes
  value: GAN 25.26% / 24.72% vs diffusion 88.68% / 90.45% (experimental / control)
  note: the diffusion fakes used were easy; GAN fakes were mostly judged real
  source_location: null
moderators:
- 'Generator: GAN fakes far harder than the diffusion fakes used'
limitations:
- No d' or criterion reported
- 8 images per generator
- Single session, no retention
- Online, uncontrolled
verification:
  status: verified-fulltext
  confidence: high
  checked: 2026-10-12
  via: frontiersin.org full text
url: null
pmid: null
pmcid: null
arxiv_id: null
openalex_id: null
oa_url: null
cited_in:
- MR-0001
updated: '2026-10-12'
---

# Training humans for synthetic face image detection

> **Source record** `rehman-2025-training-synthetic-faces` · 2025 · Frontiers in Artificial Intelligence  
> **Evidence tier:** B (peer-reviewed-journal) · **Verification:** verified-fulltext  
> **Bottom line:** One-shot artefact training raised trained participants' accuracy on synthetic faces (79.08% vs 58.29% for controls) but lowered accuracy on real faces, so overall accuracy did not differ significantly between groups.

## Citation {#citation}

Ramlah Sara Rehman, Ewald Meier, Mathias Ibsen, Christian Rathgeb, Robert Nichols, Christoph Busch (2025). Training humans for synthetic face image detection. *Frontiers in Artificial Intelligence*, 8, 1568267. https://doi.org/10.3389/frai.2025.1568267

Verification entry (identifiers, resolvers, claims with source locations): [/data/citations.json#rehman-2025-training-synthetic-faces](/data/citations.json#rehman-2025-training-synthetic-faces)

## Study design {#design}

| Field | Value |
|---|---|
| Design | Online randomised experiment: 16 trials, then artefact/semantic-region training (experimental) or a 5-min break (control), then 16 trials |
| Modality | image-face |
| Stimuli | StyleGAN2 and diffusion (dezgo.com) synthetic faces plus FFHQ and online real faces; 8 GAN and 8 diffusion images in total |
| Participants (N) | 184 |
| N detail | Experimental n=95, control n=89 (254 started) |
| Topics | human-detection, training, criterion-shift |

## Results {#results}

| Metric | Value | Note |
|---|---|---|
| Overall accuracy, experimental pre vs post | 67.43% vs 71.12% | p = 0.01 |
| Overall accuracy, control pre vs post | 68.12% vs 68.33% | p = 0.88 |
| Post-intervention, experimental vs control | difference 2.79 points | p = 0.08, n.s. |
| Post-intervention synthetic faces correct | experimental 79.08% vs control 58.29% | p < 0.001 |
| Post-intervention real faces correct | experimental lower than control (control 78.37%) | t(182) = -5.37, p < 0.001: consistent with a shift toward 'synthetic' |
| Pre-intervention accuracy on GAN vs diffusion fakes | GAN 25.26% / 24.72% vs diffusion 88.68% / 90.45% (experimental / control) | the diffusion fakes used were easy; GAN fakes were mostly judged real |

## Moderators {#moderators}

- Generator: GAN fakes far harder than the diffusion fakes used

## Limitations {#limitations}

- No d' or criterion reported
- 8 images per generator
- Single session, no retention
- Online, uncontrolled

## 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-fulltext` |
| Compiler confidence | high |
| Checked on | 2026-10-12 |
| Via | frontiersin.org full text |