Can People Actually Tell AI Headshots from Real Photos? Here's What the Research Says

If you've considered using an AI headshot generator for your professional photo, you've probably had this thought: will people be able to tell it's not a real photograph? It's the question that stops most people from clicking "generate."

The research might surprise you.

Most people are remarkably bad at distinguishing AI-generated faces from real ones. The full picture is more nuanced, and understanding the details matters if you're making decisions about your professional image.

What the Science Actually Shows

A 2025 study published in Royal Society Open Science by Gray et al. examined whether even "super-recognizers" could detect AI-generated faces. These are people with exceptional face recognition abilities.

The findings were striking.

Untrained participants performed barely above chance when trying to identify AI-generated faces. The technology has reached a point where the old tells have largely been resolved. Weird ears, too-smooth skin, asymmetric jewelry? Modern AI face generation produces results that pass casual inspection almost every time.

The study did find that brief training improved detection accuracy. As little as five minutes of guided practice made a significant difference. The training helped participants look for specific technical artifacts, not the kind of thing a LinkedIn connection or hiring manager would ever think to check.

The Detection Gap Is Growing

Earlier AI portrait tools produced obvious artifacts. Fingers that didn't quite work. Backgrounds that melted into clothing. Teeth that looked like they belonged to two different people.

Those days are over.

Current generation AI headshot tools produce results that are technically indistinguishable from professional studio work at normal viewing sizes. This is particularly true for those trained specifically on portrait photography rather than general image generation. The pixel-level analysis that researchers use to detect AI artifacts requires zooming in far beyond what any reasonable person would do when glancing at your LinkedIn profile or company website.

A separate dataset study compiled over a million AI-generated face images specifically to build better detection tools. The AI-Face dataset from CVPR 2025 tells you something about how good the technology has become. Researchers need million-scale datasets and specialized algorithms just to detect AI faces.

What People Actually Notice

When someone looks at your professional headshot, they're evaluating a few things in about two seconds:

1. Does this person look professional?
2. Does the photo quality suggest they take their career seriously?
3. Does anything feel "off" enough to distract from the content?

They're not analyzing pupil reflections or checking whether the bokeh pattern is physically consistent with a real lens. They're making a gut-level judgment about credibility based on composition, lighting, and expression. Technical origin is irrelevant.

Research in first impressions established that people form trait judgments from faces in as little as 100 milliseconds. At that speed, whether a camera was involved doesn't matter. What matters is whether the photo communicates competence and trustworthiness.

Where AI Headshots Still Fall Short

Honesty matters here. Let's address the limitations directly.

Consistency across contexts. If someone meets you in person and you look significantly different from your headshot, that creates a credibility gap. Whether it's AI-generated or heavily retouched by a photographer makes no difference. The best AI headshot tools produce results that look like you on a good day in good lighting, not like a different person entirely.

Extreme close-up scrutiny. If someone downloads your headshot and examines it at full resolution looking for artifacts, they might find them. This is not normal behavior, but it happens in specific contexts like journalism or academic profiles where authenticity is professionally relevant.

Group consistency. When an entire team page uses AI headshots generated from different tools or with different settings, the inconsistency between photos can be more noticeable than any individual photo's AI origin. Purpose-built tools like Narkis have an advantage here. Generating an entire team's headshots through one platform keeps the style consistent.

The Real Question You Should Be Asking

Instead of "can people tell?", ask this: "does this photo serve my professional goals better than what I currently have?"

If your current headshot is a cropped vacation photo, a five-year-old image from a different hairstyle, or simply missing entirely, an AI headshot is a significant upgrade regardless of whether anyone can detect its origin.

Professional headshot photographers charge anywhere from $150 to $500 for a single session. AI headshot generators like Narkis start at $27 for 200 photos with different styles, backgrounds, and compositions.

The math is simple. For most professionals, the AI option delivers a better photo at a fraction of the cost, and the detection question is largely academic.

When Detection Matters

It genuinely doesn't matter for most professional contexts:

• LinkedIn profiles
• Company team pages and directories
• Email signatures
• Business cards
• Social media professional accounts
• Speaker bios and conference materials

It might matter in a few specific situations:

• Roles where visual authenticity is part of the job: broadcast journalism, modeling
• Academic or research contexts with specific image policies
• Industries with regulatory requirements around identity verification

For the vast majority of professionals, no one is running your headshot through a detection algorithm. They're deciding in two seconds whether you look credible. Then they're reading your content.

What Makes AI Headshots Undetectable

The difference between an obvious AI headshot and an undetectable one comes down to the tool and the input photos.

General-purpose AI image generators create beautiful images of people who don't exist. Midjourney, DALL-E, ChatGPT's image feature all work this way. That's fundamentally different from what a dedicated AI headshot generator does.

Dedicated tools take your real photos and produce professional portraits that look like you, just with better lighting, composition, and backgrounds.

The dedicated approach is far harder to detect because the underlying face is real. The AI is handling the photography part. Lighting, composition, background, color grading. It's not inventing a face from scratch. This is closer to what a skilled photographer and retoucher do than what a generative art tool does.

Research Summary

The research is clear. Casual observers cannot reliably distinguish modern AI headshots from traditional photography. Detection requires specialized training, magnification, or algorithmic analysis that no one applies to professional headshots in normal contexts.

If the only thing stopping you from upgrading your professional headshot is the fear that someone will "know," the science suggests that fear is unfounded.

What people will notice is whether your photo looks professional. How it was created is, for practical purposes, invisible.

Related Guides

• anti-AI headshot movement
• what recruiters check
• symmetry in professional photos

FAQ

Can LinkedIn detect if my headshot is AI-generated?

LinkedIn does not currently scan profile photos for AI generation. The platform's image policies focus on requiring a real photo of yourself, which AI headshots of your actual face satisfy. There is no automated detection system checking the technical origin of profile photos.

Are AI-generated headshots considered "fake"?

An AI headshot generated from your real photos is a professional portrait of you. It's your face, your features, your expression. The AI handles the photography variables the same way a photographer and retoucher would. Lighting, background, composition. Calling it "fake" would be like calling any retouched professional photo fake.

How accurate are AI detection tools?

Current detection tools vary widely. Research-grade detectors can achieve high accuracy under controlled conditions, but publicly available tools produce inconsistent results. This is especially true with dedicated headshot generators that work from real source photos rather than generating faces from scratch.

Will AI headshot detection improve over time?

Detection methods and generation methods are in an ongoing arms race. As detection improves, generation does too. For professional headshots specifically, the trend favors undetectability because the tools are refining existing photos rather than creating synthetic faces.

Should I tell people my headshot is AI-generated?

There's no professional obligation to disclose how your headshot was created, just as you wouldn't disclose your photographer's retouching process. If someone asks directly, honesty is always the best policy. Most people never ask.