# How do I check whether a LinkedIn profile photo is AI generated? Source: https://global100.org/forum/q/how-do-i-check-whether-a-linkedin-profile-photo-is-ai-generated/ Site: Global 100 Forum, category AI images and photos Published: 2026-10-24. Updated: 2026-10-25. Replies: 8. ## Question by Ingrid V., 2026-10-24 I run trust and safety at a marketplace, and a growing share of the seller accounts we review have a polished headshot and a thin, slightly generic professional story. Several reviewers have started pasting profile photos into image detectors and treating the answer as a decision. I am uneasy about that. The same question reaches me from the other side too. Colleagues get connection requests and recruiter messages on LinkedIn and ask me whether the sender is real. They want a quick way to check a photo. What I would like to understand is what actually works and what is theatre. Are the old tells, like mismatched earrings or warped backgrounds, still useful with current generators? Do the free detector sites hold up on a compressed profile picture that has been resized by the platform? And if the photo cannot be settled either way, what are the better signals to check on the profile itself? I need something I can write into a reviewer checklist, including what to do with an uncertain result. I would rather have a checklist that says "this is not enough on its own" than one that gives staff false confidence. ## Reply 1 by Marco Ferri, 2026-10-24 Newsroom habit: I never judge a face. I judge the account. Where else does this person exist? Does the employer's own site or press page list them? Does the photo appear anywhere else, under another name? A reverse image search takes thirty seconds and answers more than staring at pixels does. The earring and background tells were real in the early GAN era. I would not put them in a checklist now. A checklist item that only works on old generators is worse than no item. ## Reply 2 by Dan Okafor, 2026-10-24 Practical thing from the school side: ask the person to join a short live video call. We do it for parent contacts we cannot verify. It is not foolproof, but a real person with a real face on camera is a very different signal from a photo and a polite message. ## Reply 3 by Hana Sato, 2026-10-25 Careful with detectors on profile photos, Ingrid. They are trained on particular generators and tested on particular image conditions. Profile images get recompressed and resized, and detectors tend to do worse once that happens. If a detector says "real", that tells you very little, and if it says "fake" it should open a review, not close one. ## Reply 4 by Sam Whitlock, 2026-10-25 Adversarial view: whoever is going to the trouble of generating a face is also going to build the rest of the profile. So the face is the least interesting part. What gives these accounts away is lack of history: a connection list that grew in one burst, no posts older than a few months, endorsements from other thin accounts, a work history with no corroboration anywhere else. You can fake a face in a minute. Faking six years of mutual connections is expensive. ## Reply 5 by Marco Ferri, 2026-10-25 @Hana Sato agreed, and one practical add: if you do run a detector, save the original file as well as the result and note the tool name and date. If someone challenges the decision you can rerun it, and you may find the answer changes. ## Reply 6 by Kwame A., 2026-10-25 One thing worth writing into the checklist: the legal exposure of a wrong call. Calling a real person's photo "AI" and removing their account or rejecting their application is a decision with consequences for them. I would require a second reviewer, and I would record which signals were used. "The detector said so" is a poor reason to give if it is challenged later. ## Reply 7 by Tomas Reyes (staff), 2026-10-25 You cannot settle this from the photo alone. People guess at about chance, automated detectors lose accuracy on compressed images from generators they were not built for, and the stronger signals are on the account: who else vouches for this person, and whether anything about them exists elsewhere. Start with what humans can do. NPR's report [That smiling LinkedIn profile face might be a computer-generated fake](https://www.npr.org/2022/03/27/1088140809/fake-linkedin-profiles) cites a study finding people have just a 50% chance of guessing correctly whether a face was computer made, and says the faces were rated slightly more trustworthy. The same report describes how Stanford researchers found more than 1,000 profiles: the photo oddities (one earring, vanishing hair, eyes centred in the frame, blurry background) were the trigger, but the confirmation came from checking the claims. Employers said the named people never worked there, and universities contacted had no record of the supposed graduates. That is the 2022 GAN era, so treat the photo tells as dated, as Marco says. For automated tools, NIST's [Reducing Risks Posed by Synthetic Content (AI 100-4)](https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-4.pdf) says detector accuracy varies with image compression, style and content, that detectors do better on images from the generator they were trained on, and that in the cross-generator setting error rates are substantial, especially after compression and resizing. It cites one 2023 paper reporting 61% to 70% accuracy and another from 2022 reporting 50% to 62%. For your checklist: treat a detector result as a prompt for review, never a decision; check the employer and school claims directly; look for the same photo elsewhere; look at account history; require a live check for high-risk cases; and have a second reviewer sign off on any rejection. The [thread on whether image detectors are reliable](/forum/q/are-ai-image-detectors-reliable-and-what-do-the-tests-show/) covers the testing in more detail. ## Reply 8 by Ingrid V., 2026-10-25 @Tomas Reyes this is the checklist skeleton I was missing, thank you. I will drop the earring item, make detectors a "prompt for review" signal only, and put employer and school verification first. The two reviewer sign-off on rejections goes in as well. --- Cite as: Global 100 Forum, "How do I check whether a LinkedIn profile photo is AI generated?", https://global100.org/forum/q/how-do-i-check-whether-a-linkedin-profile-photo-is-ai-generated/, accessed 2026-11-10.