Journal · LightWhy ChatGPT and a TikTok filter gave you the wrong season
Same face, new lighting, new season. The face was not the variable.
By Ximena · 5 min read · Updated Oct 19, 2026
If ChatGPT gave you one season and a TikTok filter another, you may not have a wrong season, only two different photos.
The rest of the answer
A photo carries the color of the light it was taken in, plus whatever the camera and any filter did to it, so each app, filter or AI tool reads a different picture and can name a different season. Research also describes AI systems learning from pictures that include fewer deep skin tones, which can add error. A home method can control some of this: daylight, a fixed gray strip, your own clothes and a repeat on another day. It cannot control all of it.
Local woman receives new season with each new lighting
Her face, reportedly, stayed the same.
A photo carries the light it was taken in, so two photos can give two seasons. The test here uses none.
What changes photo to photo
A phone photo is not a neutral witness.
Myth
Why, and how solid
Several things shift before any app looks at it.
| What changes | What it does to your result |
|---|---|
| Camera white balance | The camera sets its own, so the whole picture can lean warmer or cooler than the room did. |
| Room light | Light tints the whole picture, skin included. |
| Filters | A filter, a TikTok color filter included, can shift your skin tone along with the rest of the frame. |
| Compression | Video compression and automatic brightness tend to flatten mid tones and edges, so a face can read quieter on screen than in the room. That is camera logic and photographer habit, and I found no study of it. |
| The model's guess | A tool turns pixels into a season guess, so it inherits the light, the camera and the filter. Research also describes training pictures with fewer deep skin tones, which can add error for deep skin. |
"A photo is more objective than a window test."
Myth The camera sets its own white balance, filters can shift skin tone and the room changes a lot, so a photo is not a neutral witness.
Two honest readings can differ without either one being wrong.
Neighboring seasons make it worse
The twelve seasons form a ring, and each one borders the next.
Moderate
Why, and how solid
Four pairs sit right where warm meets cool: Deep Autumn and Deep Winter, Soft Summer and Soft Autumn, Bright Winter and Bright Spring, Light Spring and Light Summer. A result between two of them is expected, so two honest readings can differ without either one being wrong.
If two tools give you two seasons, check whether they are neighbors before you decide either one is wrong.
Why: light and camera change from photo to photo and neighboring seasons share a border, so two honest readings can differ.
Moderate The effect of light on color is well understood. Where the borders between seasons sit is a convention.
Two seasons? Compare them
#8E2D1FOne flat color#5C2A5FOne flat colorWhen two tools name two neighbors, a third photo will not settle it.
Tradition
Why, and how solid
Your own face in window light can.
Take the free test, then tap "Hold it to your face" on your card.
Why: your phone shows one flat color from each of the two seasons, and you hold it under your chin by a window. The light falls on your face and the color together, so it cannot favor one of them.
Tradition Comparing drapes is how consultants tell neighbors apart. This phone version has not been tested.
Compare real colors against your face by a window.
What a home method controls
- 1
- 2
- 3
- 4
- fixed gray5
- 6
- 7
- 8
- 9
Measure next to your own face, so the light hits both the same way.
ModerateTradition
Why, and how solid
The test here is built on one idea: put what you measure next to your own face, in the same light, so the light affects both the same way. You compare against a fixed gray strip and a few color cards by a window, and you change one thing at a time. Nothing is uploaded.
Skip filters when you want to know your season. Compare real colors against your face by a window.
Why: a filter can shift your skin tone along with the rest of the frame, so the color you are judging is not the color you have.
Moderate The reasons are well understood. Whether a window comparison finds your season is Tradition.
It still has limits. Light, screens and printers shift color. Makeup, a tan and dyed hair change what you see. The tests that exist were mostly built with lighter skin in mind, and deep and olive skin are where they tend to miss. I do not know yet how often my test is right. In 2027 I plan to ask 60 to 100 readers of different skin tones and backgrounds to help measure it, and to publish the numbers.
Read the confidence label on your card, and if it says low or mixed signals, repeat the test by a window on another day.
Why: confidence drops when your answers sit near a cut point or point two ways, because a home test cannot control light the way a studio does. The label says how sure the card is, not how good you look.
Tradition The test has not been validated yet, and the label is how it says so.
no. that photo lied.
but try again by a window, then compare the two.
Questions people ask
Why did I get one season in the morning and another at night?
The light may have changed more than you did. The camera sets its own white balance and the room tints the whole picture, so the same face can give two readings. That goes for an app, a TikTok filter or a chat tool like ChatGPT given a selfie.
Do AI tools work worse on deep skin?
They can. A paper in the ACM Digital Library describes AI training data that includes fewer deep skin tones, and Essence has reported on the erasure of Black beauty in AI images. I have not tested any one tool, so I make no claim about ChatGPT or any other by name. If a tool gave you a result that did not fit, the photo is a likelier suspect than you are. The deep skin and olive skin pages say what I do instead.
Did ChatGPT or a TikTok filter give me the wrong season?
Maybe, and maybe not. If it named a season next to the one you expected, both readings can be honest, because neighbors share a border and a photo moves the line. Start with the free test, which uses no photo, and hold your two seasons to your face by a window.
Is one app more accurate than the others?
I do not know, and I found no study that measures it. I also found no study of how often a season guess is right, or how often two human analysts agree, for any season system. Many lists of the best apps are written by the apps themselves, so I would read them with that in mind.
Sources
- ACM Digital Library, paper on training data and dark skin tones in AI systems. dl.acm.org/doi/full/10.1145/3632120
- Essence, article on the erasure of Black beauty in AI images. essence.com
I read both only as search summaries, so I describe them in my own words and quote neither. The camera and compression points are camera logic and photographer habit, not studies.
The free test uses no photo and says how sure it is.
