What Jennifer Did: Were the Images AI-Manipulated?

Yes. Netflix’s true-crime documentary What Jennifer Did used AI-manipulated versions of photographs of Jennifer Pan in a sequence during which her face appears to change. The images were altered with generative AI, according to Netflix and the production company; the change was not evidence that the real photographs or the documentary’s underlying case footage had been digitally fabricated.

What viewers noticed

The controversy emerged after viewers spotted apparent differences in Pan’s facial features across images shown in the documentary. In particular, some versions appeared to give her a different smile, teeth, or facial expression. The shots looked like ordinary still photographs in the film, so viewers questioned whether they were authentic images or had been altered.

Online comparisons drew attention to the discrepancies, and reporting by outlets including 404 Media identified the images as AI-altered. The issue was not merely that a documentary had used a visual effect: the images were presented without an on-screen explanation that they had been changed. That raised reasonable questions about authenticity, editorial transparency, and how documentary viewers interpret photographs.

What Netflix and the filmmakers said

Netflix and the documentary’s production company, RAW, acknowledged that generative AI had been used on the images. Their explanation was that the photographs had been enhanced to improve resolution, rather than to create new evidence or change the story. The clarification came after the images attracted public scrutiny.

That distinction matters, but it does not make the concern disappear. AI “enhancement” can go beyond making a blurry image clearer: generative tools may synthesize visual details that were not legible, or even present, in the source. If a face is reconstructed while increasing resolution, viewers cannot assume every visible feature is a faithful record of what the camera captured.

The filmmakers’ stated purpose was not to alter the meaning of the case. However, the finished images were not simply untouched photographs, and the film did not clearly tell viewers at the moment they appeared that AI had been used. The most accurate summary is therefore that the images were AI-manipulated for visual enhancement, according to the production’s explanation, and that their presentation lacked adequate transparency.

How the AI image changes work

Generative image tools can estimate and add detail when enlarging or cleaning up a photograph. Rather than recovering information that is certainly present in the original, a model may generate plausible-looking pixels based on patterns learned from other images. This can make a small or low-resolution picture look sharper while also changing details such as skin texture, teeth, eyes, or the outline of a smile.

There are several different processes that people may loosely call “AI enhancement”:

Conventional enlargement: pixels are scaled up and may be sharpened, without a model inventing substantial new facial detail.
Generative upscaling: software predicts and creates details to make an image look higher-resolution.
Face alteration or synthesis: facial features or expressions are deliberately modified or generated.

The public explanation for What Jennifer Did described enhancement, not an attempt to fabricate a new photograph or change Pan’s expression for a particular argument. Still, viewers comparing the images reported visible facial differences. Without access to the original files, editing workflow, and a documented comparison, an outside viewer cannot independently establish exactly which tool or process produced every changed feature.

Why it matters in a true-crime documentary

Photographs in a documentary can feel like direct evidence, even when they are used only to illustrate a person or a period in their life. A digitally generated detail may influence how a viewer reads someone’s appearance, mood, or credibility. That makes disclosure especially important when the subject is a real person connected to a criminal case.

AI alteration does not, by itself, prove that a documentary’s account of events is false. The images in question are not the same thing as the police evidence, court record, or the facts established at trial. But documentary makers should make clear when a visual is reconstructed, enhanced with generative software, staged, or otherwise changed in a way that could affect its evidentiary appearance.

A useful way to assess a disputed image is to ask what it is being used to establish. If it is merely decorative, alteration may still deserve disclosure, but it may have limited bearing on the factual claims. If it is presented as proof of a person’s appearance, actions, or state of mind, the standard for faithful representation should be much higher.

Background on Jennifer Pan’s case

What Jennifer Did is a Netflix documentary about Jennifer Pan, a Canadian woman whose parents were attacked at their home in Markham, Ontario, in 2010. Her mother, Bich Ha Pan, was killed, and her father, Hann Pan, survived with serious injuries. The investigation and criminal proceedings established that the attack was connected to a murder-for-hire plot involving Pan and other people.

Pan was convicted in 2014 of first-degree murder and attempted murder. In 2023, the Supreme Court of Canada ordered a new trial after finding legal errors in the jury instructions concerning the possibility of parole ineligibility. That ruling did not declare Pan innocent or erase the convictions automatically; it ordered a new trial. This legal context is separate from the documentary’s image-editing controversy, and the two should not be conflated.

What is known—and what is not

The confirmed point is that generative AI was used to enhance photographs in the documentary, and Netflix and RAW acknowledged the use after questions were raised. Viewers also identified visual differences between versions. The production described the purpose as improving image quality, while critics argued that the apparent changes and lack of clear disclosure made the images misleading.

Publicly available information does not settle every technical question. It does not provide a complete, independently verified account of the software, settings, source files, or exact edits for each image. Nor is there public evidence that the AI images were used to fabricate the attack, alter court evidence, or prove a disputed fact in the case. Claims beyond those points should be treated cautiously unless supported by new documentation.

Frequently Asked Questions

Did Netflix admit that the images were AI-generated?
Netflix and production company RAW acknowledged that generative AI was used to enhance photographs shown in the documentary. They said the aim was to improve resolution.

Were the images completely fake?
They were based on photographs, according to the production’s explanation, but were altered using generative AI. Calling them wholly fabricated goes beyond what has been established; calling them untouched originals is also inaccurate.

Did AI change Jennifer Pan’s expression?
Viewers observed differences in facial details, including her apparent smile and teeth. The production described the work as enhancement, but the precise edits and tools have not been fully documented publicly.

Did the AI controversy change the facts of the case?
No. The editing controversy concerns documentary images, not a change to the court record or the legal history. It does raise questions about how the film presented visuals.

Is What Jennifer Did still on Netflix?
Availability can vary by country and over time. Check the Netflix listing in your region for the current status.

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.

More cards in this drawer

The whole Numbers and Powers drawer →