L1
·
Quiz
·
Lab
L2
·
Quiz
·
Lab
L3
·
Quiz
·
Lab
L4
·
Quiz
·
Lab
Module Test
Module 2 · Lesson 1

The Six-Fingered Giveaway

How AI images betray themselves through anatomy — and why your eye can learn to catch what cameras can't fake
If an image looks real at first glance, what would make you look again?

In February 2023, an image of Pope Francis wearing a large white puffer jacket went viral on Twitter. Within hours it had been reshared millions of times. News outlets in Italy, Spain, and the United States embedded it without comment. People assumed it was a candid photo taken outside the Vatican.

It was not a photo. It had been generated by Midjourney version 5, which had just been released days earlier. The person who created it, a Chicago construction worker named Pablo Xavier, posted it in a Facebook group almost as a test — and watched it escape into the internet as fact.

When researchers at BuzzFeed News and Bellingcat went back and studied the image closely, they found it: the Pope's right hand had too many fingers. The lace detail on his white garment dissolved into a soft blur at the edges. His glasses had no visible temples — the arms that hook behind the ears had simply been forgotten by the model. And his skin had the smooth, slightly plastic quality that AI images had not yet learned to lose.

The image fooled millions. But once you knew what to look for — once you knew the specific places AI models consistently get wrong — you couldn't unsee it. That's what this lesson teaches you to do.

Why AI Struggles with Human Bodies

AI image generators like Midjourney, DALL-E, and Stable Diffusion work by learning statistical patterns. They train on billions of images and learn what things tend to look like near other things. A face usually has eyes above a nose. A hand usually appears near an arm. A street usually has buildings on both sides.

The problem is that hands are incredibly complex. A human hand has 27 bones, dozens of tendons, and can form hundreds of distinct shapes. When you look at a hand in a photo, you're seeing a precise mechanical structure. When an AI generates a hand, it's predicting what a hand-shaped cluster of pixels usually looks like — and it frequently predicts wrong. Extra fingers. Fingers that merge into each other. Thumbs on the wrong side. Nails that face different directions.

This isn't a bug that's going to disappear tomorrow. As of 2024, even the best AI image generators still produce hands that require close inspection. The more fingers visible in the image, the higher the chance something is wrong.

ArtifactA visible glitch or error in an AI image — something that shouldn't be there or looks wrong. Artifacts are the clues you're learning to find.
Statistical patternThe way AI learns: by finding what things usually look like, rather than understanding how they actually work. This is why AI can seem convincing but still make structural errors.
The Anatomy Checklist

After the Pope jacket image went viral, several digital forensics researchers published guides to spotting AI anatomy errors. They identified a consistent set of places where AI models fail. You can run this checklist on any suspicious image in under thirty seconds.

Fingers and hands. Count visible fingers. Check whether they look independently jointed. Look for fingers that are too long, too short, that merge at the base, or that have the wrong number of knuckles. Look for thumbs that appear on both sides of a hand.

Eyes. AI-generated eyes are often slightly asymmetrical in ways real eyes aren't. The reflections in the eyes — called catchlights — sometimes don't match. One eye might reflect a window; the other reflects nothing. Look at the pupils: in real photos they're round or slightly oval; in AI images they sometimes have subtle irregular shapes.

Ears. Ears are another structure AI consistently fails at. They're often blurred, simplified, or on close inspection structurally impossible — the inner cartilage folds in the wrong direction or doesn't exist at all.

Teeth. When mouths are open, AI often generates teeth that blend into a single smooth surface instead of showing distinct individual teeth. Or conversely, it generates too many teeth for the space.

Hairlines and hair at edges. Where hair meets the background of an image, AI often produces a soft, unnatural blending — or individual hairs that seem to multiply and fork in impossible ways.

Pause Point

You've now got the first tool. Before you read further: think of one image you've seen recently that seemed surprising or unusual. Could any of these features have been off? You don't need to answer definitively. The habit of asking is what matters.

What You Can Now See

Here's the honest truth: most people who saw the Pope jacket image did not look at his hands. They looked at his face, registered "Pope," and moved on. That is the normal way of seeing. We're pattern-matchers too — and we mostly look for confirmation of what we expect to see.

What forensic analysts and fact-checkers do differently is train themselves to look at the parts of an image that carry the most information. Not the thing that catches your eye — but the thing that's hardest to fake.

You can now do that. You know that hands are the hardest thing for AI to generate correctly. You know what to look for in eyes, ears, and teeth. Most adults who share AI images online have never been told this. You have been.

Identity Moment

Knowing where AI images break down puts you ahead of the vast majority of people who encounter them. Not because you're smarter — because you've learned to look in the right places. That is a specific, real skill. The next time someone shares an image claiming to be a real photo, you'll check the hands. That changes everything.

Ethical Question — No Clean Answer

Pablo Xavier didn't claim the Pope jacket image was real. He posted it in a hobby group. But millions of people shared it as real anyway. Who is responsible for the harm caused by a convincing AI image — the person who made it for fun, the platforms that allowed it to spread, or the people who shared it without checking? Does intent matter if the damage is the same?

Lesson 1 Quiz

5 questions · Anatomy artifacts in AI images
1. In February 2023, the AI-generated image of Pope Francis was created using which tool?
Correct. The image was made with Midjourney v5, which had just launched. The creator was Pablo Xavier, a Chicago construction worker who posted it in a Facebook hobby group.
Not quite. The image was generated with Midjourney version 5, which had just been released in early 2023.
2. Why do AI image models specifically struggle with generating hands?
Correct. AI learns statistical patterns — what things usually look like — rather than understanding the underlying mechanical structure of a hand. That structural complexity is what trips it up.
Not quite. The real reason is that hands have 27 bones and complex jointing that pattern-matching can't reliably reproduce — so AI generates "hand-shaped pixels" that often have errors.
3. A friend shows you a news photo. You notice one person has two thumbs on the same hand and the teeth in their open mouth form a single smooth surface. What is the most accurate conclusion?
Correct. Two thumbs on one hand and blended teeth are both classic anatomy artifacts. Multiple artifacts appearing together is a strong indicator of AI generation, not camera error or medical conditions.
Think about it differently. Two thumbs and blended teeth are specific anatomy artifacts — structural errors consistent with AI generation. When multiple artifacts appear in one image, that's a significant signal.
4. What does the word "artifact" mean in the context of AI images?
Correct. An artifact is a visible error — something that looks wrong structurally. In AI images, these are the clues: extra fingers, impossible ear shapes, blended teeth.
In forensics, an artifact means a visible error or glitch — something that shouldn't be there. In AI images, artifacts are structural mistakes like extra fingers or eyes with mismatched reflections.
5. You see an image of a crowd at a concert. No faces are visible, but you notice in the front row that several people have hair that seems to fork and multiply into impossible strands where it meets the bright stage lighting. This is most likely:
Correct. Hair-to-background blending is a recognized AI artifact. AI models struggle to render the precise edge where individual strands meet a contrasting background — they often produce multiplying, forking, or melting strands.
Hair-to-background blending is a known AI artifact. AI models struggle at the precise edge where hair meets background — generating strands that fork, multiply, or melt in ways real hair doesn't. This isn't a camera or lighting effect.

Lab 1: Anatomy Auditor

Role: Image forensics investigator · AI partner: peer analyst, not teacher

Your Assignment

You're reviewing images flagged for possible AI generation before a news outlet publishes a story. Your AI partner has studied the same forensics literature you have. They will push back on weak reasoning and ask you to justify your calls.

Start by describing a scenario: pick any type of image — a portrait, a crowd photo, a political image, a product ad — and describe what you'd examine first and why. Your partner will respond as a peer investigator, challenge your approach, and add their own analysis.

Opening prompt suggestion: "I'm looking at a portrait of a politician that was shared without a photo credit. Here's what I'm checking first and why..." — or describe your own scenario.
Forensics Lab — Anatomy Artifacts
PEER ANALYST
I've got the same training you do on anatomy artifacts. Give me the scenario — what image are we auditing, and what's your first move? I'm going to push back if your reasoning is weak, so be specific.
Module 2 · Lesson 2

Light That Doesn't Know Where It Came From

Lighting physics is the one test AI almost always fails — and the one most people never think to apply
Can light in a picture lie? And what does that tell you about who made the picture?

In March 2023, Eliot Higgins, the founder of the open-source investigative group Bellingcat, generated a series of images purporting to show Donald Trump being arrested by police officers. He created them in Midjourney and posted them on Twitter explicitly labeled as AI-generated, as a demonstration of what the technology could now produce.

Several of the images circulated beyond Higgins's posts without his labels. Fact-checkers at AFP and Reuters were called in to verify them. One of the first things their analysts noted — before checking fingers, before checking metadata — was the lighting.

In one image, Trump's face was lit from the left. The police officers surrounding him were lit from the right. The pavement between them had a shadow pattern that matched neither. Three different scenes had been averaged together by the AI into a single image — and physics didn't care that the statistics said it looked right. The light said something different.

Lighting inconsistency became one of the primary diagnostic tools that major wire services used in 2023 and 2024 to flag AI-generated images before publication. It is not obvious to a casual viewer. It is obvious once you know how to look.

How Light Actually Works in Photos

In a real photograph, there is one physical light source — or a coherent set of them. The sun, a lamp, an overhead fluorescent. Every object in the scene is lit by that same source, which means every shadow falls in the same direction, every highlight appears on the same side of a curved surface, and every reflective surface catches the same color of light.

AI image generators don't simulate physics. They predict what a scene usually looks like based on millions of training images. The problem is that those training images come from thousands of different scenes with different lighting. When the AI averages them together, it sometimes produces a scene where different parts of the image were lit by different imaginary light sources — sources that can't coexist in the same physical space.

This produces specific, detectable errors.

CatchlightThe small reflection visible in a person's eye from the light source in the scene. In real photos, both eyes reflect the same light. In AI images, catchlights sometimes don't match between eyes.
Shadow directionThe direction a shadow falls from an object. In a real scene, all shadows point away from the same light source. In AI images, different objects sometimes cast shadows in different directions.
The Lighting Checklist

Step 1: Find the main light source. Look at the image and identify where the light is coming from. Is the sun visible? Is there a window? A lamp? Once you've identified it, check whether the shadows in the image are consistent with it.

Step 2: Check shadow directions. Pick two or three objects in the scene — a person, a building, a chair — and trace the direction their shadows fall. In a real photo, they all point away from the same source. In AI images, they sometimes point in different directions.

Step 3: Check the catchlights. If faces are visible, look at the eyes. Each eye should have a small bright reflection. In a real photo, both eyes show the same reflection. In AI images, one eye might show a window and the other might show a blank wall — or one might have a reflection that doesn't exist anywhere in the scene.

Step 4: Check highlights on curved surfaces. Noses, cheeks, shoulders, and round objects like glasses should all have highlights on the same side — the side facing the light source. If highlights appear randomly placed, or on the wrong side of a surface, that's an artifact.

Step 5: Check reflected color. In real scenes, surfaces near a colored light source pick up that color in their shadows (this is called color bleeding or color cast). AI images sometimes get this right and sometimes produce surfaces that are completely neutral when they should be tinted.

Why This Matters at Scale

In 2024, Adobe, Getty Images, and the Associated Press all updated their editorial policies to require photographers to disclose AI-generated content. One of the methods these organizations use to audit submissions is automated lighting analysis — software that checks whether shadow directions and catchlights are consistent. You just learned to do by eye what professional tools do computationally.

The Harder Question

Bellingcat's Eliot Higgins made the Trump arrest images to demonstrate capability — to show what was now possible so that the public would be prepared. He labeled them clearly. They still escaped and spread as real.

Researchers who study synthetic media call this the demonstration problem: showing people what a fake looks like sometimes makes better fakes more plausible, not less. The more you demonstrate that AI can produce convincing images, the more people assume any surprising image might be AI — which can also be used to dismiss real photos as fake.

You now have a way to check. But knowing how to check is different from having time to check every image you see. Most people make sharing decisions in seconds. The lighting analysis you just learned takes a full minute of careful attention.

Ethical Question — No Clean Answer

If you know that demonstrating an AI image trick causes some people to use it for harm, but also causes some people to become better at detecting it — what's the right call? Is it better to show the public how sophisticated AI fakes have become, or does that demonstration itself become a tool for bad actors? Who gets to decide?

What You Can Now See

Wire service analysts at Reuters and AFP use lighting consistency as a primary detection method. You've now learned the same five-step process they apply. This is not general media literacy — this is a specific technical skill used by professional fact-checkers. Most people who encounter AI images never apply it once.

Lesson 2 Quiz

5 questions · Lighting inconsistency as a detection tool
1. Which organization did Eliot Higgins found, and what was he demonstrating when he made AI images of Trump's arrest in 2023?
Correct. Higgins founded Bellingcat, an open-source investigative group. He made the images explicitly to demonstrate what Midjourney could now produce — and labeled them as AI, though they spread without those labels.
Higgins founded Bellingcat, a well-known open-source investigative group. His Trump arrest images were a deliberate demonstration of AI's new capabilities — though they spread without his labels anyway.
2. Why do AI images sometimes have shadows pointing in different directions for different objects in the same scene?
Correct. AI doesn't simulate physics — it averages statistical patterns from training data. When those training images came from different lighting environments, the AI blends them in ways that violate physical consistency.
AI doesn't simulate physics. It predicts pixels based on statistical patterns from millions of training images — images that had many different real light sources. The AI blends them, producing physically impossible lighting combinations.
3. You're examining an indoor portrait. The subject's left cheek is highlighted, suggesting light from the left. But their left eye's catchlight shows a window reflection on the right side of the iris. What does this suggest?
Correct. A highlight on the left cheek means the light is from the left — so the catchlight in the eye should also be on the left side. A catchlight on the right contradicts this and is an artifact of AI averaging different lighting environments.
Think about what the highlight and catchlight each tell you about the light source direction. If the highlight says "light from the left" and the catchlight says "light from the right," they can't both be true in the same physical scene. That contradiction is an artifact.
4. What is the "demonstration problem" in the context of AI synthetic media?
Correct. The demonstration problem is a genuine dilemma: educating the public by showing what AI can do may also teach bad actors what to aim for, or make people more willing to dismiss real photos as fake.
The demonstration problem is about the double-edged nature of revealing AI capabilities: it educates some people while potentially enabling others — and can also make people dismiss real photos as AI-generated.
5. In 2024, the Associated Press updated its editorial policy to require photographers to disclose AI-generated content. Which of the following technical methods do organizations like AP use to audit image submissions for lighting consistency?
Correct. Professional organizations use automated lighting analysis — software that checks shadow direction consistency and catchlight matching — as part of their AI detection workflow. You've now learned to do by eye what that software does computationally.
Organizations like AP use automated lighting analysis tools that check whether shadow directions and catchlight positions are physically consistent. File size and reverse search are separate tools — they don't address lighting physics.

Lab 2: Lighting Detective

Role: Wire service image auditor · AI partner: peer analyst

Your Assignment

A wire service editor has received three images for a breaking news story. You need to audit them for lighting consistency before they go to print. Your AI partner will analyze your reasoning and challenge any calls you can't back up with specific observations.

Describe a specific image scenario — political, celebrity, natural disaster, sports — and walk through your lighting audit step by step. Your partner will push back, add analysis, and ask you to go deeper on anything vague.

Opening prompt suggestion: "I'm auditing a photo of a protest crowd at night, lit by what looks like streetlights. Here's what I'm checking..." — or set up your own scenario and start your audit.
Lighting Analysis Lab
PEER ANALYST
Editor's got three images waiting. Let's start the audit. Set up the scene — what's the image, what's the claimed light source, and what's your first move? If you're vague, I'll ask you to be specific.
Module 2 · Lesson 3

The Background That Forgot the Rules

Text, signs, and backgrounds in AI images reveal the model's deepest limitations — and the most overlooked clues
What happens when you zoom out from the face — and start reading everything else in the picture?

In October 2023, following the outbreak of fighting in Israel and Gaza, social media platforms were flooded with images claiming to document events on the ground. Researchers at the Stanford Internet Observatory and the Israeli newspaper Haaretz analyzed hundreds of viral images in the first two weeks of the conflict.

Many images that had circulated as documentation of current events turned out to be either archive photos from earlier conflicts, or AI-generated. The detection of AI images in this context often came not from faces or hands — many images showed rubble, crowds at a distance, or smoke — but from text and signage in the background.

In one widely shared image purporting to show a damaged building in Gaza, a researcher at Stanford noticed that a sign on a wall in the background contained letters that did not spell any real Arabic or Hebrew word. They looked like Arabic script — the curves and dots were approximately correct — but the characters had been blended into nonsense. Another sign appeared to be in English but read something like "MEDEICAL ENTRACE." Not "MEDICAL ENTRANCE" — a garbled approximation.

AI image models, as of 2023 and 2024, do not understand text. They generate images of text — shapes that look like letters — without knowing what those letters mean or whether they form real words. This is one of the most reliable tells in any AI image that contains writing.

Why AI Can't Read Its Own Writing

When you read the word "EXIT" on a sign, you understand it as a sequence of meaning: four letters, one word, one instruction. You know the letters are in an order that produces a specific word in a specific language.

An AI image generator, when it produces a sign that says "EXIT," is not doing any of that. It is generating a cluster of pixels that statistically resembles what a sign with text on it looks like in the training data. If the text happens to be legible, that's coincidental. The AI has no model of language that it applies when generating images — even though the same company's language model understands text perfectly.

Image generation and language understanding are separate systems that don't yet talk to each other well. This is why you will often find, in AI images that contain signs, shop fronts, newspapers, graffiti, or name badges, text that is partially legible but contains errors, invented characters, or near-miss words.

Garbled textText in an AI image that looks like real writing from a distance but contains invented letters, misspellings, or character combinations that don't form real words in any language.
Background coherenceWhether the background of an image follows the same rules as the foreground — consistent architecture, readable signs, and logical spatial relationships between objects.
The Background Checklist

Read every piece of text visible in the image. Signs, menus, name tags, newspapers, bumper stickers, graffiti, labels on equipment. If any text is garbled, partially invented, or uses characters that don't exist in the language it's supposed to be written in, that's a significant artifact.

Check architectural logic. AI images frequently produce buildings where windows don't align, doors appear at impossible heights, or brickwork patterns start and then arbitrarily change. Look at corners where walls meet — AI models often produce walls that don't join cleanly or that continue with different textures on each side.

Look for objects that merge. In real photos, distinct objects are distinct. In AI images, objects near each other sometimes merge — a crowd of people where individual bodies blend into each other at the edges, a bookshelf where the spines of books melt together. Look at the edges between objects, especially in busy scenes.

Check for repeated elements. AI models sometimes tile or repeat visual patterns in backgrounds — the same window appearing three times with slightly different warping, a crowd where you notice the same face appearing twice on different bodies. This is because AI generates from patterns, and sometimes the same pattern gets generated multiple times.

Check for impossible spatial relationships. In real scenes, objects that are behind other objects are partially occluded — you can't see the part that's blocked. In AI images, sometimes an object appears to be behind another object but is fully visible anyway, or a hand holding a cup appears in front of a wall but the wall's shadow somehow falls on the hand from both sides.

Real Stakes — Conflict Zones and Policy

In conflict zones, AI-generated images can shape public opinion, justify military decisions, and affect aid organization responses within hours. When the Stanford Internet Observatory published its analysis in October 2023, they noted that some AI images had been shared by verified accounts with hundreds of thousands of followers. The detection skill you're building right now is something governments, newsrooms, and human rights organizations are actively trying to scale.

The Zoom-Out Habit

Most people look at an image's subject. They look at the face, the action, the dramatic element in the center of the frame. This is natural — it's where composition techniques direct the eye.

Developing the background-reading habit means training yourself to zoom out of the emotional core of the image and ask: what's happening at the edges? What does the sign on the wall say? Do the buildings make architectural sense? Are the objects in the background distinct from each other?

This requires deliberate effort. It goes against how images are designed to be read. That's exactly why it works — AI models are trained on images designed to look compelling at the center, and they put less effort into coherence at the periphery.

What You Can Now See

You've now added background analysis to your toolkit. Most fact-checkers working AI image cases describe the same moment: they were checking a face or hands, then noticed something weird in the background — a sign with nonsense text, a window that didn't exist on the right side of the building. You now know to look there first, not last.

Ethical Question — No Clean Answer

The Stanford researchers who published their analysis of AI images in the Israel-Gaza conflict were helping the public understand what was real. But publishing a detailed guide to how those images were detected also gives future image creators a roadmap to avoid those specific errors. Is publishing detection methods always the right call? Or does it accelerate an arms race between fakers and checkers that the checkers can't win?

Lesson 3 Quiz

5 questions · Background artifacts and text detection
1. What organization analyzed AI-generated images circulating during the October 2023 Israel-Gaza conflict, and what was the key artifact that identified many AI images?
Correct. Stanford Internet Observatory researchers identified garbled text on background signs as a key artifact in many AI images circulating during that conflict.
The Stanford Internet Observatory conducted that analysis. Garbled text on background signs — Arabic or Hebrew characters that looked right from a distance but formed no real words — was a primary detection artifact.
2. Why do AI image generators produce garbled or misspelled text on signs and objects, even though the same company's language models understand text perfectly?
Correct. Image generation and language understanding are separate systems. The image model generates pixels that statistically resemble text without any understanding of what the letters mean or whether they form real words.
The reason is architectural: image generation and language understanding are separate systems. The image model generates pixel patterns that look like writing without understanding language — so text becomes an approximation, not real words.
3. You're analyzing an AI image of a busy café. The faces look realistic. But you notice the menu on the wall contains the word "COFEE" and a third item that uses a character combination that doesn't exist in any language. What do you conclude?
Correct. Garbled text — especially invented characters that don't exist in any language — is a strong AI artifact. Realistic faces don't cancel out text artifacts. Each artifact is independent evidence.
Garbled text is a strong, independent artifact. Realistic faces elsewhere in the image don't cancel it out — each element is its own piece of evidence. "COFEE" and invented characters are classic AI text errors.
4. What does it mean when a crowd in an AI image shows people at the edges "merging" into each other rather than being distinct individuals?
Correct. AI generates from patterns, not physics. At the edges between distinct objects, the model often blends them rather than maintaining the clean separation that real physical objects have.
This is an AI background artifact. The model generates patterns that look like a crowd, but at object boundaries it doesn't maintain physical distinctness — so bodies merge. Real JPEG compression blurs uniformly; it doesn't merge specific object edges selectively.
5. A researcher publishes a detailed guide to detecting AI images, including all the specific artifacts to look for. A critic argues this is harmful because it teaches image fakers what to fix. The researcher argues it's necessary because the public needs to know. This disagreement is an example of:
Correct. This is exactly the demonstration problem: publishing detection guides educates some people and may give others a roadmap to evade those detections. There is no clean answer about whether publishing is right or wrong.
This is the demonstration problem — the same concept from Lesson 2. Educating people about AI detection may simultaneously educate fakers about what to avoid. It's a genuine ethical tension without a clean resolution.

Lab 3: Background Reader

Role: Conflict zone image analyst · AI partner: peer investigator

Your Assignment

A human rights organization has flagged 20 images circulating on social media, claiming to document events in a conflict zone. You're the analyst assigned to triage them by examining backgrounds, text, and object coherence. Your AI partner is on the same team — but will challenge any call you make without specific evidence.

Describe an image scenario involving a location (street, building, crowd, rubble) and walk your partner through your background analysis. Be specific: what text do you see? What architectural details? What object boundaries?

Opening prompt suggestion: "I'm looking at an image of a damaged apartment building. The foreground shows rubble and looks realistic. But in the background I notice..." — describe what you'd examine and what you find suspicious.
Background Analysis Lab
PEER INVESTIGATOR
Twenty images, clock's ticking. Organization needs a priority list by end of day. Walk me through your first image — describe the scene, what's in the background, and what you're flagging. Give me specifics. I'll argue back if your reasoning is thin.
Module 2 · Lesson 4

Putting the Clues Together

Real detection doesn't happen artifact by artifact — it happens when you build a case and know when you have enough evidence
When you've found three things wrong, is that enough? And what do you do when everything looks right?

In January 2024, explicit AI-generated images of Taylor Swift were posted on the platform X (formerly Twitter) and spread to millions of viewers before they were taken down. The images had been generated using a Microsoft AI tool and distributed initially through a Telegram channel dedicated to creating non-consensual synthetic images of celebrities.

What made this case different from earlier AI image controversies was how quickly analysts at several organizations, including Wired, 404 Media, and the Stanford Internet Observatory, were able to confirm the images were AI-generated. It wasn't one artifact — it was the convergence of several.

Researchers noted: skin texture that was slightly too smooth and uniform across lighting variations. Background elements (crowd figures in some images) that merged at their edges. Proportional inconsistencies in how different body parts related to each other spatially. And in several images, subtle text on clothing or merchandise that was partially garbled.

No single artifact was definitive. But four independent signals pointing in the same direction made the case as solid as forensic analysts could make it without access to the original generation logs. This is how professional image forensics actually works: you don't need a confession; you need a pattern of evidence that has no other plausible explanation.

Building a Case, Not Finding a Smoking Gun

When you've been through the previous three lessons, you have a substantial toolkit: anatomy artifacts (hands, eyes, ears, teeth), lighting inconsistency (shadow directions, catchlights, highlights), and background analysis (text, architecture, object merging, repeated elements).

The professional approach is to treat each of these as an independent line of evidence. When you find one artifact, you have a flag — something worth investigating. When you find two, you have a serious concern. When you find three or more, especially across different categories, you have a pattern of evidence that is very difficult to explain any other way.

This is important for a specific reason: real photos can have apparent artifacts too. A real photo taken in unusual light might have odd shadows. A real photo taken through glass might distort a sign. A real photo of a crowd taken with motion blur might make bodies seem to merge. The difference is that these real-photo artifacts have specific explanations — you can point to the glass, the blur, the lighting angle. AI artifacts tend to be unexplainable by any single cause.

Convergent evidenceWhen multiple independent observations all point toward the same conclusion. In forensics, convergent evidence is much stronger than a single piece of evidence — because the chance that all indicators are wrong simultaneously is very low.
Alternative explanationA different reason that could account for the same observation. Good forensic thinking requires eliminating alternative explanations before making a conclusion.
When Everything Looks Right

Here's an uncomfortable truth about where AI image generation is heading: the artifacts you've learned to spot in this module are getting harder to find. Midjourney v6, released in December 2023, produces hands significantly better than v5 — the hands in many v6 images require much more careful scrutiny. The lighting consistency has improved. Some text on signs is now legible.

This means that the visual artifact checklist you've built is a snapshot of AI's current limitations — not a permanent set of tells. As AI improves, some of these tells will become less reliable. This is why professional forensic organizations increasingly combine visual inspection with other methods: metadata analysis (checking EXIF data embedded in image files), reverse image search (checking whether the image or similar images appear elsewhere), and AI detection software (tools trained specifically to recognize patterns in AI-generated images at a statistical level the human eye can't perceive).

The visual skills you've built are the first line of analysis — the thing you apply in the first 30 seconds when you encounter a suspicious image. They are not the last line. When visual analysis is inconclusive, the professional response is not to guess — it's to say "inconclusive" and apply additional methods.

What Institutions Do With This

In 2024, the European Union passed the AI Act, which among other provisions requires that AI-generated images be labeled as such at point of creation. Major social media platforms, including Meta and X, announced policies requiring disclosure of synthetic media in political advertising. These policies exist because visual detection alone is insufficient at scale — millions of images per day can't all be hand-checked. The skills you've built matter for the images you personally encounter. At the institutional level, the same principles are being encoded into law and platform policy.

Your Complete Toolkit — and Its Limits

You've now covered four lessons and three major categories of AI image artifacts. Here is the honest summary of what you can and can't do with this knowledge.

You can: Apply a systematic three-category artifact check to any still image in about two to three minutes. Identify specific, nameable artifacts rather than just a vague feeling that something is "off." Build a case from convergent evidence rather than relying on a single indicator. Recognize when your analysis is inconclusive and further methods are needed.

You can't: Definitively prove any image is AI-generated through visual inspection alone — you can only say the evidence is consistent with AI generation and inconsistent with the alternative explanations. Keep pace with improving AI without continuing to update your knowledge. Inspect video using these still-image techniques — video has its own separate artifact categories (motion artifacts, temporal inconsistency, audio sync) that go beyond this module.

The goal of this module was never to make you certain. It was to make you careful. And careful, systematic inspection — even when it ends in "I can't tell" — is infinitely more useful than confident guessing.

What You Can Now See

You've built a three-category forensic framework that professional fact-checkers and image analysts use in real newsrooms and human rights organizations. You understand not just what to look for, but why AI fails in those specific places — and why those failures are getting harder to spot over time. You also understand the limits of visual detection, which is something many confident "AI detector" articles don't tell you. Knowing the limits of your own tools is what separates a careful analyst from an overconfident one.

Ethical Question — No Clean Answer

The Taylor Swift deepfake images prompted calls in the US Congress for legislation against non-consensual AI-generated images of real people. Some proposed laws would make creating or distributing such images a federal crime. But the same technology that generated those images can also create valuable art, satire, and fiction involving real public figures. Where is the line between protected creative expression and harmful synthetic imagery — and who has the authority to draw it?

Lesson 4 Quiz

5 questions · Building a forensic case from convergent evidence
1. In January 2024, AI-generated images of Taylor Swift originated from which platform, and what was notable about how analysts confirmed they were AI-generated?
Correct. The images spread from a Telegram channel to X. Analysts confirmed AI generation through convergent evidence across multiple categories — no single artifact was definitive, but several independent signals all pointed the same direction.
The images were initially distributed on Telegram, then spread to X. Analysts used convergent evidence — multiple independent artifact categories — rather than any single smoking gun. That's how real forensic work operates.
2. What does "convergent evidence" mean in image forensics, and why is it stronger than a single artifact?
Correct. Convergent evidence means multiple independent signals all pointing the same way. Each single artifact might have an alternative explanation, but three separate artifact categories all indicating AI generation is very hard to explain otherwise.
Convergent evidence means multiple independent observations all pointing the same direction. A single artifact might have an innocent explanation. But when anatomy, lighting, and background artifacts all appear in the same image, the combined probability that they're all coincidental is very low.
3. A real photo of a construction site has shadows that seem to fall in slightly inconsistent directions. A careful analyst would:
Correct. Real photos can have apparent artifacts with specific real-world causes. A good analyst looks for alternative explanations — multiple light sources, glass reflections, architectural shadows — before attributing an anomaly to AI generation.
Real photos can have shadow anomalies with specific real-world explanations: a reflective building nearby, a second light source, a time-lapse composite. Good forensic thinking eliminates alternative explanations before concluding AI generation.
4. Why is the visual artifact checklist covered in this module likely to become less reliable over time?
Correct. Midjourney v6 already produces better hands than v5. The specific artifact categories you've learned reflect AI's current limitations — and those limitations are actively being reduced with each new model version.
AI models improve rapidly. The artifacts in this module reflect AI's limitations as of 2023–2024. Each new model version — Midjourney v6, newer DALL-E versions — specifically addresses the most visible failure points. Visual detection must be supplemented with metadata analysis and detection software.
5. You've examined an image thoroughly and found: (1) no obvious hand errors, (2) consistent shadow directions, (3) no garbled text. But something still feels "off." What is the most professionally responsible conclusion?
Correct. Passing the visual checklist doesn't prove an image is genuine — it means visual inspection was inconclusive. The professional response is to escalate to additional methods rather than force a conclusion from insufficient evidence.
Passing visual checks doesn't prove genuineness — it means visual inspection is inconclusive. Good forensic practice says: when a method doesn't give you a clear answer, apply additional methods (metadata, reverse search, detection software) rather than guessing.

Lab 4: Full Case Build

Role: Lead forensic analyst · AI partner: peer reviewer

Your Assignment

You're the lead analyst at a digital verification unit. An image has been submitted to your organization with a claim that it shows a real event. You need to build a full forensic case — documenting what you examined, what you found, and what your conclusion is — across all three artifact categories: anatomy, lighting, and background.

Your AI partner is acting as peer reviewer. They will challenge your reasoning, ask you to justify calls you can't support with specific evidence, and push you to address alternative explanations before you reach a conclusion.

Opening prompt suggestion: "I'm building a full forensic case on an image showing a celebrity at what's claimed to be a charity event. No photo credit, no timestamp. Here's my systematic analysis..." — run all three artifact categories and build your case.
Full Case Analysis Lab
PEER REVIEWER
I'm reviewing your case before it goes to the editor. Walk me through the full analysis — anatomy, lighting, background — in order. Show your work. If your conclusion doesn't follow from your evidence, I'll send it back. Start when you're ready.

Module Test

15 questions · All lessons · Score 80% or higher to pass
1. The AI-generated image of Pope Francis in a puffer jacket went viral in February 2023. What was the creator's name and original platform?
Correct. Pablo Xavier, a Chicago construction worker, created and posted the image in a Facebook hobby group — from which it escaped into the wider internet.
Pablo Xavier, a Chicago construction worker, created the image and posted it in a Facebook hobby group. It spread from there without his labels.
2. Which specific anatomy artifact was found in the AI-generated Pope Francis image that was identified by fact-checkers?
Correct. Fact-checkers identified extra fingers, missing glasses temples, and smooth skin texture as artifacts in the Pope jacket image.
The identified artifacts included too many fingers and glasses with no visible arm/temple pieces — among other subtle tells like unusually smooth skin texture.
3. AI image generators produce hand errors because:
Correct. The 27-bone structure of human hands is mechanically complex — and AI generates statistical approximations, not structural models. The complexity means errors occur frequently.
Hands have 27 bones and extremely complex jointing. AI doesn't model that structure — it predicts pixel patterns. That gap between structure and pattern produces frequent errors.
4. In a real photograph, all shadows in the scene point away from:
Correct. In any real scene, one physical light source (or coherent set of sources) governs all shadows. All shadows point away from that source — in the same direction.
In a real scene, all shadows point away from the same light source. If shadows in an image point in different directions for different objects, that's an artifact — physics doesn't allow it.
5. What is a "catchlight" and why does it matter for AI image detection?
Correct. Catchlights are the light source reflections in the eyes. In real photos, both eyes catch the same source. AI images sometimes show different reflections in each eye, or reflections of light sources that don't exist anywhere in the scene.
A catchlight is the reflection of the light source visible in a person's eyes. Because both eyes see the same scene, a real photo shows matching catchlights. AI images sometimes produce different or impossible reflections in each eye.
6. Eliot Higgins labeled his AI Trump arrest images as synthetic when he posted them. They still spread as real. This illustrates:
Correct. Labels are easily stripped when images are saved and reshared. Once an image enters circulation without its original context, it spreads based on visual impact — not on what the creator originally said about it.
Labels are stripped when images are downloaded and reshared. An image spreads based on its visual impact, not what the creator said. This is why detection skills matter — labels can't be relied on to follow an image everywhere it goes.
7. In October 2023, researchers analyzing AI images from the Israel-Gaza conflict found that many were identifiable because of garbled text in backgrounds. Why does this specific artifact occur?
Correct. Image generators and language models are separate systems that don't fully communicate. The image model generates what text-on-a-surface looks like statistically — not what specific words mean.
Image generation and language understanding are different systems. The image model generates pixel patterns resembling writing without understanding language — so it produces plausible-looking but often garbled text.
8. When checking an AI image for background artifacts, which of these would be the strongest single indicator?
Correct. Invented characters that exist in no real language have no alternative explanation — they can't be caused by focus, compression, or unusual architecture. This is the cleanest possible artifact because it has no other cause.
Invented characters that exist in no real language can't be explained by camera settings, compression, or cultural unfamiliarity. An unusual architectural style could be real. Obscured faces are normal in crowds. Background blur is a real camera technique. Invented characters are unambiguous.
9. You're inspecting an image and find: (1) a hand with six fingers, (2) a shadow that points the wrong direction for one object, (3) a shop sign with partially garbled letters. What is your conclusion?
Correct. Three independent artifact categories all pointing the same direction is convergent evidence. The probability that a real photo happens to have anatomy, lighting, and background artifacts simultaneously from coincidental causes is extremely low.
Three independent artifact categories all pointing toward AI generation is convergent evidence — much stronger than any single artifact. The chance that anatomy, lighting, and text errors all appear in a real photo by coincidence is very low. This is a strong case.
10. The AI-generated images of Taylor Swift in January 2024 were initially distributed through which platform?
Correct. A Telegram channel dedicated to creating non-consensual synthetic images distributed them, from which they spread to X (formerly Twitter).
The images originated in a Telegram channel dedicated to creating synthetic images and then spread to X, where they reached millions of viewers before being removed.
11. Which EU legislation, passed in 2024, requires that AI-generated images be labeled as such at point of creation?
Correct. The EU AI Act, passed in 2024, includes provisions requiring labeling of AI-generated content at point of creation. This is one of the first major legal frameworks addressing synthetic media.
The EU AI Act (2024) includes requirements for labeling AI-generated content. It represents one of the first significant legal frameworks addressing synthetic media at a policy level.
12. A real construction site photo has shadows that appear slightly inconsistent because of a reflective glass building nearby bouncing light. An analyst applying the lighting checklist should:
Correct. Good forensic analysis eliminates alternative explanations. A glass building creating secondary reflections is a real physical cause for shadow anomalies — one that removes this from the AI artifact column.
Forensic analysis requires eliminating alternative explanations. If reflected light from a glass building explains the shadow anomaly, that anomaly is not an artifact — it has a real physical cause. Artifacts are anomalies with no other plausible explanation.
13. Why is the visual artifact detection framework taught in this module expected to become less reliable over time?
Correct. Each new model version specifically addresses the most visible failure points. Midjourney v6 already improved significantly on hands compared to v5. The artifact checklist tracks current limitations, not permanent ones.
AI models iterate rapidly, specifically fixing their most visible failures. Legal labeling requirements also don't guarantee compliance. Visual detection must be supplemented with metadata analysis and detection software as the technology improves.
14. After checking all three artifact categories on an image and finding no clear problems, you still feel uncertain. The most professionally responsible response is to:
Correct. Passing visual checks means visual inspection is inconclusive — not that the image is real. Professional practice calls for escalating to additional methods rather than forcing a conclusion.
Passing visual checks doesn't confirm genuineness — it means visual inspection is inconclusive. Next steps are metadata analysis, reverse image search, and AI detection software. Forcing a conclusion without sufficient evidence is not responsible analysis.
15. The "demonstration problem" in synthetic media research means that publishing detailed AI detection guides:
Correct. The demonstration problem is a genuine ethical dilemma without a clean resolution — publishing detection methods educates some and enables others, and the same information does both simultaneously.
The demonstration problem is a real dilemma: detection guides simultaneously inform the public and potentially inform bad actors about which specific errors to avoid. There's no clean answer about whether publishing is always right or always wrong.