AI Headshots and Team Photos: Consent, Accuracy, and When Not to Fake It

AI headshot generators crossed a threshold sometime in the last two years. The output from tools like Aragon, Secta Labs, and the newer HeyGen and ProPhoto features is now good enough that a candidate scrolling your team page usually can't tell. That's precisely why this is the moment to be careful, not casual.

There's a real, legitimate use case here. Getting thirty distributed employees into a studio for consistent headshots is expensive and logistically painful. Remote hires often have no company photo at all, so their tile is a gray silhouette or a cropped wedding photo. AI headshot tools solve a genuine problem. But they generate images of real, identifiable people, and that changes the rules compared to inventing a background scene.

Consent is the whole foundation

You cannot generate an AI headshot of an employee without their informed, specific, revocable consent. Not "they signed a general media release in onboarding." Specific consent for this: their likeness, processed through an AI model, to produce a synthetic image used on company materials.

What informed consent actually covers:

• What you'll do: which tool, that it's AI-generated from their uploaded selfies, and that it produces a new synthetic image rather than a retouched photo.
• Where it'll appear: careers page, LinkedIn, internal directory, job ads. Be specific. Consent for the internal directory is not consent for a paid recruiting campaign.
• How long, and how to revoke: people leave, people change how they look, people simply change their mind. You need a clean process to pull an image and, ideally, to delete the training uploads.
• What happens to their photos: most of these tools train a small personalized model on 10 to 20 selfies. Where do those go? Are they deleted after generation? Read the tool's data terms, and tell your employee the honest answer.

This is an area where data-protection law is genuinely relevant, and it varies a lot by region. Biometric and likeness rules under regimes like the EU's GDPR, Illinois's BIPA, and a growing patchwork of 2025 to 2026 state laws can treat facial data as sensitive. I'm a creative director, not a lawyer, so treat this as a prompt to talk to your privacy or employment counsel, not as legal advice.

Accuracy: the line between a headshot and a deepfake of your own staff

Here's the failure mode nobody warns you about. AI headshot tools don't just clean up lighting. They subtly change the person. They slim faces, lighten skin, straighten teeth, remove distinguishing features, and drift toward a generic conventionally attractive average. Run a diverse team through one of these and you can end up with thirty people who all look a little more alike, a little more Anglo, a little more airbrushed than they are.

That's not a cosmetic issue. It's an accuracy and a fairness issue:

• A candidate who meets a very different-looking person on day one feels quietly misled.
• Systematic lightening of skin or Westernizing of features is a real, documented tendency in image models, and it sends an ugly signal about who your company thinks is presentable.
• An employee may not even recognize the "improved" version of themselves, which is its own kind of disrespect.

My rule: the person must approve their own image and must feel it looks like them, not an idealized stranger. Give them veto power and multiple options. If the tool won't produce something that reads as actually them, don't ship it.

Practical accuracy safeguards

• Feed enough varied source selfies (different angles, real lighting, no heavy filters) so the model has honest material to work with.
• Turn off or dial down the "enhancement" and "beautify" settings where the tool allows it. You want a clean headshot, not a glamour shot.
• Generate a batch and let the employee pick. Never auto-publish the first result.
• Do a bias check across the whole team's set. If everyone's skin got lighter and every face got narrower, your tool is drifting and you need to correct or switch.

When not to fake it at all

Some situations call for real photography, full stop.

Leadership and public-facing profiles. Your CEO, your founders, anyone quoted in press or investor materials should have real photos. The reputational downside of "the CEO's headshot is AI-generated" is not worth the convenience.

Anything implying a specific event or achievement. Don't generate a "team at the offsite" or "volunteering day" photo of people who weren't there. That's not a headshot; it's a fabricated record of something that didn't happen.

Fabricated employees. Never generate photos of people who don't exist and present them as staff, testimonial-givers, or "a day in the life" characters. This is the bright line. It shows up on Glassdoor threads and Reddit within weeks, and it torches trust for every real thing you say afterward.

Inflating team size or diversity. If your team is four people, a wall of twelve smiling faces is a lie candidates will discover on the first call. Same for diversity you don't have. Fix the hiring, don't fake the photo.

A team-photo workflow I trust

1. Offer AI headshots as an opt-in convenience, especially for remote and new hires, alongside the option of submitting their own real photo or having none.
2. Get specific written consent covering tool, usage, duration, and deletion.
3. Use a reputable tool with clear data-deletion terms; favor ones that let you disable beautification.
4. Let each person choose and approve their own final image. Their face, their call.
5. Keep source uploads secure and delete them on a set schedule.
6. Maintain a record of which images are AI-generated, and pull them promptly when someone leaves or revokes.
7. Re-run a fairness pass on the whole set before it goes live.

The reframe

AI headshots are best understood not as a photography shortcut but as a small HR data process involving people's faces. That framing gets you to the right instincts automatically: consent, accuracy, control, deletion. Handled that way, they're a legitimately good tool for a distributed workforce. Handled carelessly, they're a trust liability wearing a friendly face, sometimes literally the wrong one.