How Old Do I Look ChatGPT? A Prompt That Actually Works

The prompt to paste into ChatGPT

The wording decides whether you get a range, a single number, or a hedge.

Open a temporary chat and turn personalization off before you attach the photo. OpenAI's data controls in ChatGPT say a personalized temporary chat can still use memories, custom instructions, and plugins. A birthday saved in memory would anchor the reply.

Paste this prompt, then attach one clear photo of your own face:

• Look only at this photo. Ignore my name, any age I have stated before, and anything you remember about me. Estimate the age I appear to be in this picture. Reply with one best apparent age, a likely age range, a confidence of low, medium, or high, and 3 to 5 visible cues from skin texture, lines, the eye area, hair or hairline, face shape, and skin tone. Mark each cue older, younger, or neutral. Note if blur, light, a filter, glasses, or the camera angle may have shifted the reading. Do not name me, guess ethnicity, or give medical advice.

What an answer can settle

A useful reply names an apparent age and a range: how old the face looks in that photo. A second photo can move it.

OpenAI's separate feature, age prediction, decides whether an account should get ChatGPT for Teens. Age prediction in ChatGPT lists the signals: conversation topics, time of day, how the account is used, and how long the account has existed. An adult corrects that category with Persona, by a live selfie or a government ID. OpenAI does not receive the ID or selfie, and Persona deletes them within 7 days. The approach to age prediction describes that same account check. Asking how old you look from a photo you attach is a different request.

OpenAI does not publish an error margin for this kind of photo question. Public write-ups describe both a written estimate and a refusal, so one chat is one try. If the model declines, stop. Pressing past a refusal is a poor test.

OpenAI's usage policies disallow facial recognition databases without consent, real-time remote biometric identification in public places, and classifying people from biometric data, including inferring sensitive attributes. Use your own photo, or a photo someone asked you to check. Leave hiring, medicine, and age gates to processes built for those decisions.

ChatGPT beside a dedicated photo report

Neither side has published a score for this question. Compare the answer, the account, and where the photo goes.

Run a fair side-by-side test

Use one original file both times, and do not state your real age. Light, a close wide-angle phone, a filter, or a tired expression can move either reading.

• Use a front-facing photo in soft light, one face, no filter, no sunglasses. Keep the original file, not a screenshot.
• In Settings, then Data controls, turn off Improve the model for everyone if you want the chat out of training. Start a temporary chat and turn personalization off.
• Paste the prompt and attach the photo. Leave out your birthday, your city, and your age.
• Upload that same file above. You get a range, a confidence level, and the cues.
• Compare the ranges and which cues each one called older or younger.
• For a second round, change only the light, the expression, or the glasses.
• If ChatGPT declines, stop. The report on this page still reads the file.

Where to go after the two readings

Cue by cue: the main photo report. An existing picture or old print: the age detector. How the report is built: the AI age guesser. A round with friends: guess my age.

Limits both readings share

NIST's face age estimation evaluation measures systems built for that job. ChatGPT is a general assistant, and OpenAI's age-prediction documents describe an account check for teens. We have not run that evaluation, so this page does not quote a NIST score for either one. Use both readings to compare your own photos. Keep access checks, medicine, and decisions about other people on processes built for those jobs.

Questions about how old do I look ChatGPT