---
type: source
id: gray-2025-super-recognizers
title: Training human super-recognizers' detection and discrimination of AI-generated faces
authors:
- Katie L. H. Gray
- Josh P. Davis
- Carl Bunce
- Eilidh Noyes
- Kay L. Ritchie
year: 2025
venue: Royal Society Open Science
volume: 12(11)
pages: '250921'
doi: 10.1098/rsos.250921
peer_review: peer-reviewed-journal
evidence_tier: B
design: Two online experiments (Exp1 single-face detection; Exp2 two-face 2AFC discrimination); super-recognisers vs Prolific and database controls; with vs without ≈5-minute artefact training with 10 feedback trials
modality:
- image-face
stimuli: 80 StyleGAN3 faces and 80 FFHQ real faces, demographically matched
n_participants: 664
n_note: Super-recognisers n=283, controls n=381 across conditions after catch-trial exclusions
topics:
- human-detection
- training
- individual-differences
- hyperrealism
summary: Untrained controls labelled only 31% of StyleGAN3 faces as AI (below chance) and super-recognisers 41%; ≈5 minutes of artefact training raised these to 51% and 64%, with no retention test.
results:
- metric: Exp1 AI faces detected, untrained
  value: super-recognisers 41%; Prolific controls 31%; database controls 30%
  note: 'd'' 0.024 / -0.489 / -0.466: controls below chance'
  source_location: null
- metric: Exp1 AI faces detected, trained
  value: super-recognisers 64%; Prolific controls 51%
  note: 'd'' 0.738 / -0.066: only super-recognisers above chance'
  source_location: null
- metric: Exp2 2AFC discrimination, untrained
  value: super-recognisers 54%; controls 42% / 42%
  note: ''
  source_location: null
- metric: Exp2 2AFC discrimination, trained
  value: super-recognisers 64%; Prolific controls 48%
  note: ''
  source_location: null
moderators:
- Super-recogniser advantage persisted after training; training effect similar in both groups
limitations:
- No retention test
- Single generator (StyleGAN3)
- Training relies on visible rendering artefacts
- Online testing
verification:
  status: verified-fulltext
  confidence: high
  checked: 2026-10-12
  via: White Rose eprint 241331 (typeset published PDF)
url: null
pmid: null
pmcid: null
arxiv_id: null
openalex_id: null
oa_url: null
cited_in:
- MR-0001
updated: '2026-10-12'
---

# Training human super-recognizers' detection and discrimination of AI-generated faces

> **Source record** `gray-2025-super-recognizers` · 2025 · Royal Society Open Science  
> **Evidence tier:** B (peer-reviewed-journal) · **Verification:** verified-fulltext  
> **Bottom line:** Untrained controls labelled only 31% of StyleGAN3 faces as AI (below chance) and super-recognisers 41%; ≈5 minutes of artefact training raised these to 51% and 64%, with no retention test.

## Citation {#citation}

Katie L. H. Gray, Josh P. Davis, Carl Bunce, Eilidh Noyes, Kay L. Ritchie (2025). Training human super-recognizers' detection and discrimination of AI-generated faces. *Royal Society Open Science*, 12(11), 250921. https://doi.org/10.1098/rsos.250921

Verification entry (identifiers, resolvers, claims with source locations): [/data/citations.json#gray-2025-super-recognizers](/data/citations.json#gray-2025-super-recognizers)

## Study design {#design}

| Field | Value |
|---|---|
| Design | Two online experiments (Exp1 single-face detection; Exp2 two-face 2AFC discrimination); super-recognisers vs Prolific and database controls; with vs without ≈5-minute artefact training with 10 feedback trials |
| Modality | image-face |
| Stimuli | 80 StyleGAN3 faces and 80 FFHQ real faces, demographically matched |
| Participants (N) | 664 |
| N detail | Super-recognisers n=283, controls n=381 across conditions after catch-trial exclusions |
| Topics | human-detection, training, individual-differences, hyperrealism |

## Results {#results}

| Metric | Value | Note |
|---|---|---|
| Exp1 AI faces detected, untrained | super-recognisers 41%; Prolific controls 31%; database controls 30% | d' 0.024 / -0.489 / -0.466: controls below chance |
| Exp1 AI faces detected, trained | super-recognisers 64%; Prolific controls 51% | d' 0.738 / -0.066: only super-recognisers above chance |
| Exp2 2AFC discrimination, untrained | super-recognisers 54%; controls 42% / 42% |  |
| Exp2 2AFC discrimination, trained | super-recognisers 64%; Prolific controls 48% |  |

## Moderators {#moderators}

- Super-recogniser advantage persisted after training; training effect similar in both groups

## Limitations {#limitations}

- No retention test
- Single generator (StyleGAN3)
- Training relies on visible rendering artefacts
- Online testing

## 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 | White Rose eprint 241331 (typeset published PDF) |