AI Image Detection

AI Image Detection

Analyze visual signals linked to AI generation
and editing to support informed image review.

Same scene, different production paths

Same scene, different production paths

Compare an original photo with an AI-edited image and review representative signals from MORI.

Compare an original photo with an AI-edited image and review representative signals from MORI.

Original Photo

No AI Signals Found

A representative example with no AI-generation or editing signals detected under the review criteria.

AI-Edited Image

Further Review Recommended

A representative example where AI-related editing signals were detected and flagged for further review.

Examples are illustrative. Results can vary by image and context, and do not independently confirm authenticity or source.

Examples are illustrative. Results can vary by image and context, and do not independently confirm authenticity or source.

Review hard-to-judge images from multiple angles

Review hard-to-judge images from multiple angles

As creation and editing tools advance, a single visual cue is rarely enough. MORI analyzes multiple signals to bring consistency to repeatable review.

As creation and editing tools advance, a single visual cue is rarely enough. MORI analyzes multiple signals to bring consistency to repeatable review.

Multi-signal analysis

Review multiple signals instead of relying on a single clue.

Built for real-world distribution

Account for image size, compression, and file format in real-world publishing environments.

Review-ready results

Organize findings and file details for efficient follow-up review.

Visualize analysis attention

See areas that received greater attention during analysis.

Heatmaps are visual references that show areas of relative analysis attention. They do not independently determine whether an area was generated or edited.

AI Detection Performance Evaluated
on Public Benchmarks

AI Detection Performance Evaluated
on Public Benchmarks

We evaluated MORI’s AI detection performance across five public benchmarks, covering 331,859 images from diverse datasets.

We evaluated MORI’s AI detection performance across five public benchmarks, covering 331,859 images from diverse datasets.

Weighted Average Performance

98.1

%

Weighted by image count per benchmark

Images Evaluated

331,859

Total images evaluated across five benchmarks

Public Benchmarks

5

Datasets used in published research

Results by Benchmark

Benchmark

Image Count

Performance (%)

UnivFD diffusion

Ojha 2023

10,000

97.4%

AIGCDetectBenchmark

AIGCDetect Benchmark

PatchCraft 2023

113,963

99.0%

DIRE / DiffusionForensics

2023

28,501

98.8%

GenImage val

NeurIPS 2023

99,963

99.1%

ForenSynths test

CNNDetection 2020

79,432

95.2%

Evaluation Criteria and Notes

Calculation

Performance is averaged across benchmarks, weighted by image count and rounded to one decimal place.

Scope

Images generated by Gemini and GPT models were excluded. Actual performance may vary depending on the generative model and image conditions.

Interpretation

Results support your review and should not be used alone to establish an image’s authenticity or origin.

From image upload to review records

From image upload to review records

1

Select an Image

Upload the image you want to review.

2

Analyze the Image

Review AI-generation and editing signals from multiple angles.

3

Review Results

Check key findings and whether further review is needed.

4

Put Findings to Work

Connect findings to internal review and follow-up workflows.

Analysis reports when you need them

Analysis reports when you need them

Request a report that organizes analysis results and file details around your review process and operating environment.

Request a report that organizes analysis results and file details

around your review process and operating environment.

*Representative example flagged for further review under operating criteria. It does not identify a specific creation tool or editing method.

*Representative example flagged for further review under operating criteria.
It does not identify a specific creation tool or editing method.

Build more trusted content operations

Prioritize images that need further review

Use consistent criteria at scale

Use findings and file details as review records

Fit existing review workflows

Built for real-world use

Built for real-world use

Use AI image detection to strengthen content trust and manage operational risk.

Use AI image detection to strengthen content trust and manage operational risk.

Content Platforms

Prioritize uploaded images that need additional review.

Newsrooms and Content Review

Add supporting analysis for images that need source verification or fact-checking.

Brands and Commerce

Review AI-generation and editing signals in ads, product images, and user-submitted content.

Start AI image review
for your workflow

Start AI image review
for your workflow

AI for Content Trust

Business Registration: 722-88-03255

293 Bupyeong-daero, Bupyeong-gu, Incheon, South Korea (TechCity #907)

AI for Content Trust

Business Registration: 722-88-03255

293 Bupyeong-daero, Bupyeong-gu, Incheon, South Korea (TechCity #907)

AI for Content Trust

Business Registration: 722-88-03255

293 Bupyeong-daero, Bupyeong-gu, Incheon, South Korea (TechCity #907)