1. Home
3. Ethical Advertising
5. How to Detect AI-Generated Images and Videos: A Practical Guide

How to Detect AI-Generated Images and Videos: A Practical Guide

  Jeroen Seynhaeve     2025-10-27

As artificial intelligence becomes increasingly sophisticated, distinguishing between real and AI-generated content has become a critical skill for consumers, marketers, and media professionals alike. With synthetic media flooding social feeds and advertising campaigns, knowing how to spot AI-generated images and videos isn’t just useful—it’s essential for maintaining trust and authenticity in digital communications.

Why Detection Matters

AI-generated content has exploded across advertising, social media, and news platforms. While these tools offer creative possibilities, they also raise ethical concerns around misinformation, deepfakes, and misleading brand communications.

For businesses committed to ethical advertising, transparency about AI-generated content is paramount, and understanding detection methods helps maintain transparency with audiences and protects brand reputation. In addition, advertising standards authorities increasingly scrutinise misleading content, including undisclosed AI-generated advertisements.
 

Visual Signs: What to Look For in AI-Generated Images

AI image generators, despite their impressive capabilities, often leave telltale signs that trained eyes can catch. Look out for unnatural details, textures and inconsistencies.

• Hands and fingers remain one of the most reliable indicators. AI struggles with hand anatomy, frequently producing extra fingers, missing digits, fused fingers, or hands in anatomically impossible positions. When reviewing images, count the fingers and examine how hands interact with objects.
• Eyes and facial symmetry can reveal AI origins. Look for mismatched eyes, inconsistent reflections in pupils, or unusual iris patterns. AI sometimes creates faces that are too symmetrical or have subtle asymmetries that don’t follow human facial structure.
• Look closely at textures like skin, hair, and fabric. AI-generated images may show hair that lacks individual strand definition or blends into the background, skin that appears too smooth or plastic-like, missing natural pores and imperfections, fabric textures that don’t follow the physics of how materials drape and fold, and teeth that merge together or vary unnaturally in size and shape
• Text and writing in AI-generated images often appears as gibberish or distorted letters. Signs, product labels, book spines, and screen displays frequently contain nonsensical text or letters that look like language but don’t form real words.
• Lighting and shadows may behave illogically. Check whether shadows fall in consistent directions, whether reflections match light sources, and whether lighting on faces corresponds with the environment. AI can struggle with complex lighting physics.
• Background details sometimes blur or merge unnaturally. Architectural elements may defy geometry, patterns might not align properly, or objects in the background may morph into each other in ways that seem dreamlike rather than photographic.
• Jewellery, accessories, and clothing patterns often show inconsistencies. Earrings may not match, necklaces might phase through clothing, or fabric patterns may distort or fail to follow the contours of the body naturally.

 

Red Flags in AI-Generated Videos

Video presents additional challenges because it adds motion and temporal consistency to the equation.

Temporal Inconsistencies

• Morphing and warping occur when objects or people shift unnaturally between frames. Background elements might change subtly, accessories may appear and disappear, or facial features could shift slightly from frame to frame.
• Facial expressions and movements may seem disconnected from natural human behaviour. Blinks might occur at odd intervals, lip movements may not perfectly sync with audio, or micro-expressions could appear unnatural.
• Physics violations reveal AI generation when objects don’t follow expected physical laws. Hair might move independently of head motion, clothing may flutter without wind, or liquids could behave strangely.

Audio-Visual Mismatches

In AI-generated or manipulated videos, particularly deepfakes, watch for:

• Lip-sync issues where mouth movements don’t quite match speech patterns
• Audio quality that differs from video quality (pristine audio with lower-quality video or vice versa)
• Voice tones that remain unnaturally consistent without natural variation
• Background sounds that don’t match visible environments

 

Manual Detection Methods

Metadata Analysis

Genuine photographs and videos contain metadata (EXIF data) showing camera settings, device information, location data, and timestamps. AI-generated content often lacks this metadata or contains inconsistent information. However, metadata can be stripped or manipulated, so this shouldn’t be your only verification method.

Reverse Image Search

Using tools like Google Images, TinEye, or Bing Visual Search can help determine if an image appears elsewhere online, potentially revealing its origin. If an “original” photo appears in multiple contexts or predates its claimed creation date, that raises red flags.

 

AI Detection Tools and Software

Several specialised tools have emerged to help identify AI-generated content. Here are just a few in a rapidly growing list:

AI Image Detectors

• Hive Moderation: offers a free AI-generated content detection tool that analyses images and provides probability scores. It’s trained on multiple AI generators and regularly updated.
• AI or Not: provides simple image analysis, distinguishing between human-created and AI-generated images with reasonable accuracy for obvious cases.
• Illuminarty: examines images for signs of AI generation and can identify which AI model likely created the content.

Video and Deepfake Detectors

• Sensity: specialises in deepfake detection, offering enterprise solutions for media verification.
• Reality Defender: provides detection services for both images and videos, helping organisations verify content authenticity.

Important Limitations

No detection tool is perfect. AI generators evolve rapidly, and detection software constantly plays catch-up. False positives and false negatives occur regularly. These tools work best as part of a comprehensive verification strategy rather than as definitive answers.
 

Best Practices for Verification

• Cross-Reference Multiple Sources: Never rely on a single indicator or tool. Combine visual inspection, metadata analysis, reverse image searches, and AI detection tools for the most reliable assessment.
• Consider Context: Ask critical questions: Does this image or video make sense in context? Is the source reliable? Does the content align with other verified information? Is there a reason someone might create synthetic content here?
• Check the Source: Investigate who published the content and their track record for authenticity. Reputable news organisations and established brands typically verify content before publication, while anonymous social media accounts warrant more skepticism.
• Look for Verification Markers: Some platforms and creators now voluntarily label AI-generated content. Adobe’s Content Credentials and the Coalition for Content Provenance and Authenticity (C2PA) are developing standards for transparent content labeling. Look for these markers when available.

 

 
For more insights on advertising ethics and digital media responsibility, explore our other resources on transparent brand communications and ethical marketing practices.