Saturation-Based AI Image Detection
Raise saturation to expose structural artifacts hidden in realistic AI images.
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 4
- Confidence
- 97%
This detection method transforms a subtle visual-authenticity problem into a targeted artifact inspection. Start with the highest-quality image available, increase its saturation, and then examine areas whose geometry should remain physically coherent: teeth, bone structures, microphones, signs, badges, and background text. Artificially generated details may appear convincing at normal color levels but reveal hollowness, malformed edges, or inconsistent internal structure after saturation is raised. The method is diagnostic rather than definitive, so suspicious findings should trigger further verification instead of automatically proving fabrication. Its main limitation is compression: social networks commonly alter uploaded files enough to erase or distort the signals on which the test depends.
Origin
Extracted from Marketing Against The Grain, based on a detector demonstrated by Didi Das and discussed during the episode.
Core principles
- 01Inspect structure rather than overall realism.
- 02Exaggerated color can reveal concealed generation errors.
- 03Teeth, bones, text, and rigid objects provide strong diagnostic signals.
- 04Compression can destroy the evidence required for detection.
How to run it
- 1
Acquire the Best Source
Find the original image or the least-compressed copy available. Preserve its native resolution before applying any changes.
Pro tip Ask the sender for the original file rather than using a screenshot.
Watch out A social-media download may already have lost the necessary forensic detail.
- 2
Increase Saturation
Raise saturation enough to exaggerate color and structural inconsistencies without resizing the image.
Pro tip Compare the normal and saturated versions side by side.
Watch out Do not interpret color distortion alone as evidence of AI generation.
- 3
Inspect Diagnostic Regions
Zoom into teeth, skeletal contours, text, patterns, and rigid objects such as microphones or badges. Look for hollow areas, malformed geometry, and inconsistent lettering.
Pro tip Prioritize details governed by clear physical or typographic constraints.
Watch out Organic backgrounds and soft textures produce weaker signals.
- 4
Assess the Evidence
Record each anomaly and decide whether several independent inconsistencies point toward generation. Seek corroborating provenance or metadata before reaching a firm conclusion.
Pro tip Require multiple anomalies rather than relying on one unusual pixel region.
Watch out This test is a screening method, not conclusive proof.
In the wild
A marketer receives a high-resolution photograph purportedly showing a packed conference. After increasing saturation, they inspect the speaker's teeth, microphone, badge, and background signage. The microphone appears hollow and several letters lose coherent shapes, prompting the marketer to request original event records before sharing the image.
→ The image is withheld pending stronger provenance evidence.
Common mistakes
Testing a Compressed Screenshot
Compression may remove the fine artifacts the method relies on, making a clean-looking result falsely reassuring.
Treating One Artifact as Proof
Real photographs can contain blur, reflections, and compression errors. Require multiple signals and supporting provenance.
Is it for you?
Best for
It is best for examining original or high-resolution images whose authenticity is uncertain.
Not ideal for
It is not ideal for heavily compressed images downloaded from social-media platforms.
From the transcript
“if you bump up the saturation and look at certain parts of the picture, especially like teeth and bone structures or like ananimate objects, you…”
“the images get too compressed when they're uploaded to social media sites, so that you're not gonna be able to do this.”
From the episode
Ai Images: Flux, Spotting Fakes & 4 Marketing Use Cases