Detecting AI-Generated Photos & Deepfakes

The rise of AI image generation has created an entirely new category of risk for clients vetting providers. What used to be a simple problem — "Are these photos real and recent?" — has evolved into something more complex: "Does this person even exist?" AI-generated faces are now good enough to fool most people at first glance, and the tools to create them are free, fast, and require zero technical skill.

This guide arms you with the knowledge to spot AI-generated images, verify real providers, and protect yourself from an increasingly sophisticated category of scams. The techniques here are current as of early 2026, but AI capabilities evolve rapidly — treat this as a foundation and stay alert for new developments.

---

Why AI Fakes Are a Problem in This Industry

The Scam Economy

AI-generated photos enable scammers to create entirely fake provider profiles at scale. They generate a face, write an ad (often with AI as well), collect deposits or full payments, and disappear. Because the person in the photos doesn't exist, there's no real provider to find and no trail to follow. The barrier to entry for this scam has dropped to nearly zero.

Catfishing and Bait-and-Switch

Some operations use AI-enhanced or AI-generated photos as their advertising face, then send a different person to the actual booking. The "provider" in the ad was never real — they were generated to attract bookings. When the real person shows up and doesn't match, the client has already paid a deposit or traveled to the location.

Blackmail and Extortion

Deepfake technology can also be used against clients. Scammers can generate fake compromising images using your photos (pulled from social media or sent during screening) and threaten to distribute them. Understanding how deepfakes work helps you assess these threats rationally and respond effectively.

---

Visual Tells: What to Look For

AI image generators have improved dramatically, but they still leave artifacts — subtle visual inconsistencies that a trained eye can catch. Here are the most reliable tells as of 2026.

1. Hands and Fingers

AI still struggles with hands. Look for: too many or too few fingers, fingers that merge together, fingernails that are absent or oddly shaped, hands that seem to melt into objects they're holding, or wrists that transition unnaturally into arms. If a photo conveniently crops out hands or poses them out of frame, that's a yellow flag — not proof of AI, but worth noting alongside other signals.

2. Teeth and Mouth

AI-generated teeth often look too uniform, too perfect, or slightly blurred compared to the rest of the face. Check for: teeth that blend together without individual definition, asymmetrical lip edges that don't match natural anatomy, or a smile that seems "off" in a way that's hard to pinpoint. Real teeth have imperfections — slight color variations, minor misalignments, visible gum lines.

3. Ears and Hair Boundaries

The boundary where hair meets skin — especially around ears — is a common failure point. Look for hair that seems to merge with the background, earlobes that don't match (different shapes left vs. right), ear structures that look melted or oversimplified, or hair strands that abruptly appear or disappear rather than flowing naturally.

4. Jewelry and Accessories

Earrings, necklaces, and other accessories often reveal AI generation. Common issues: earrings that don't match left to right, necklaces that seem to float above or sink into skin, glasses frames that warp or disconnect, and accessories that partially phase through hair or clothing. If someone is wearing jewelry, examine it closely.

5. Background Inconsistencies

While the face might be polished, AI generators often produce backgrounds with subtle errors: text that's garbled or nonsensical (look at any visible signs, books, or screens), architectural elements that don't follow geometry (windows with warped frames, doors that don't align with walls), and objects that seem to blend into each other. Zooming in on the background often reveals more artifacts than examining the face.

6. Skin Texture

AI-generated skin tends to be unnaturally smooth and uniform — almost waxy. Real skin has pores, moles, freckles, subtle color variations, and texture that changes across the face (smoother on the forehead, more textured on the nose and cheeks). If every square inch of skin looks airbrushed, it could be AI generation rather than just Photoshop retouching. The distinction: retouching preserves some texture variation; AI generation often creates a completely uniform surface.

7. Eye Details

Check both eyes carefully. AI sometimes produces: different-sized pupils, iris patterns that don't match between eyes, reflections in the eyes that show different environments (one eye reflects a window, the other doesn't), or a gaze direction that seems inconsistent with the head angle. Real eyes have consistent reflections and matching iris detail.

8. Symmetry That's Too Perfect

Real human faces are asymmetrical — subtly, but meaningfully. One eye is slightly higher, the nose tilts slightly, one nostril is marginally bigger. AI tends to produce faces that are too symmetrical, giving them an uncanny-valley quality. If a face looks like it could be mirrored perfectly down the center, that's suspicious.

---

Tool-Based Detection

Visual inspection alone isn't sufficient against the best AI generators. These tools add a technical layer of verification.

Reverse Image Search

Start here. Run every provider photo through Google Images, TinEye, and Yandex Image Search. If the photos appear on stock photo sites, are associated with multiple different ads in different cities, or show up under different names, you've found a fake. AI-generated images won't appear in reverse searches (they're unique), but this step catches the large percentage of scams that still use stolen photos rather than AI.

AI Detection Tools

Several tools specifically detect AI-generated images:

• Hive Moderation AI Detector: A browser-based tool that analyzes images and provides a probability score for AI generation. Effective against most current generators.
• Illuminarty: Detects AI-generated content and identifies which model likely created it (Stable Diffusion, Midjourney, DALL-E, etc.).
• AI or Not: A straightforward yes/no detector with confidence scores. Good for quick checks.
• FotoForensics: While not AI-specific, this tool performs error-level analysis (ELA) that can reveal digital manipulation. Areas that have been altered show up differently under ELA than untouched areas.

No single tool is 100% accurate. Run suspicious images through multiple detectors and consider the aggregate result. A photo that passes one detector might fail another. False positives happen (real photos flagged as AI), so don't reject a provider based solely on one tool's output.

Metadata Analysis

Real photos taken with real cameras contain EXIF metadata — information about the camera model, date, location, settings, and software used. AI-generated images typically lack this metadata entirely or contain generic placeholder data. You can check metadata using online tools like Jeffrey's EXIF Viewer or by examining file properties on your computer. The absence of metadata isn't proof of AI (many platforms strip metadata on upload), but its presence is a positive signal.

---

Verification Techniques That Beat AI

The most reliable defense against AI-generated profiles isn't detecting the AI — it's verifying the real person behind the profile. These techniques confirm that a living human matches the advertised photos.

1. Request a Verification Selfie

Ask the provider to send a selfie holding a piece of paper with your name (or a specific word you choose) and today's date. This is nearly impossible to fake convincingly with current AI — the combination of a specific face, handwritten text, and real-time generation is beyond what readily available tools can do. Legitimate providers are often willing to comply, especially if you explain your concern respectfully.

2. Video Verification

A short video call or a brief video clip with a specific gesture you request (wave, hold up a certain number of fingers, say a specific phrase) is extremely difficult to fake in real time with current deepfake technology. Real-time deepfakes exist but require specialized hardware and produce visible artifacts — lag, blurring when moving quickly, inconsistent lighting. Even a 10-second video clip significantly increases confidence.

3. Platform Verification Badges

Many advertising platforms now offer verification services where the platform itself confirms the provider's identity and photo accuracy. Tryst, Slixa, and others have various verification levels. While no system is foolproof, a platform-verified provider has passed at least one layer of human or automated identity checking. Prioritize verified profiles.

4. Consistent Photo Sets

Look for providers who have multiple photos that are clearly from different sessions — different outfits, different locations, different lighting, different times. AI generators can produce variations of the same face, but maintaining perfect consistency across 10+ images in varied settings is much harder than generating a single convincing face. Providers with extensive, varied photo galleries are much less likely to be AI fabrications.

5. Social Media Presence Over Time

Check if the provider has a social media account (Twitter/X is most common in the industry) with posts spanning months or years. An active account with candid photos, stories, interactions with other providers, and a visible history is very difficult to fake at scale. A profile created last week with professional photos and no interaction history is a yellow flag.

---

Red Flags Specific to AI Scam Profiles

When assessing a provider profile, watch for these patterns that correlate strongly with AI-generated fake profiles:

1. Photos are flawless but the ad has errors. AI generates perfect faces but scammers still write sloppy ad copy — broken English, inconsistent details, generic descriptions that could apply to anyone.
2. No candid or casual photos. Every image looks like a magazine shoot. Real providers typically include at least some casual selfies or behind-the-scenes shots alongside professional photos.
3. The face never appears in different lighting. All photos have the same soft, diffused lighting. Real photo sets taken at different times and places naturally have varying light conditions.
4. No reviews or only brand-new reviews. AI profiles can't accumulate a genuine review history. Reviews that all appeared within a short time span may also be fabricated.
5. Prices are significantly below market. Scam profiles using AI photos often price low to maximize the number of marks who send deposits.
6. Aggressive push for upfront payment. Insistence on full payment before meeting, especially via untraceable methods (gift cards, crypto), combined with any other red flag on this list, is a strong scam indicator.
7. Refusal to verify. A legitimate provider has no reason to refuse a verification selfie or a brief video call. If someone refuses all forms of live verification, they're either not real or not willing to prove they are — both are dealbreakers.

---

What Providers Should Know

AI deepfakes aren't just a client problem. Providers face unique and serious risks from this technology:

Your Photos Can Be Stolen for Scam Profiles

Scammers routinely steal photos from legitimate provider profiles and use them in fake ads on different platforms or in different cities. This damages your reputation if the scam is associated with your face, and it can divert potential clients to the scammer. Protect yourself by watermarking your images with a subtle, consistent mark, running reverse image searches on your own photos periodically to check for unauthorized use, and reporting stolen photos to platforms immediately. Some providers add a small, unique identifying mark to each platform's photos so they can trace which platform the theft originated from.

Deepfakes Can Be Created of You

If someone obtains enough photos of your face — from your ads, your social media, or even a single session if they photograph you without consent — they can generate deepfake content including explicit content you never produced or consented to. This content could be used for extortion, revenge, harassment, or simply posted online without your knowledge. Understand this risk and factor it into your decisions about how many face photos you publish and on which platforms. Consider whether partial-face photos adequately serve your marketing needs while limiting deepfake source material.

Proactive Verification Helps Your Business

In an environment where AI fakes are proliferating, providers who proactively demonstrate their authenticity stand out. Offering platform verification badges, posting casual candid content alongside professional photos, maintaining an active social media presence with real-time posts, and being willing to do quick verification selfies for serious clients all differentiate you from the growing number of AI scam profiles. Verification isn't just safety — it's competitive advantage.

Advocate for Better Platform Standards

Support and prioritize platforms that implement strong verification standards. Encourage your peers to do the same. The more the industry normalizes thorough verification, the harder it becomes for AI scam profiles to operate. This benefits every legitimate provider by increasing client trust in the platforms where real people advertise.

---

Step-by-Step Vetting Process

Here's a practical sequence you can follow for every new provider you're considering. Each step adds confidence, and the whole process takes 10–15 minutes.

1. Reverse image search all profile photos through Google Images, TinEye, and Yandex. Check for matches on stock sites, other ads, or different identities.
2. Visual inspection of photos for AI artifacts: check hands, teeth, hair edges, jewelry, background details, and skin texture as described above.
3. Run through an AI detector — use at least two different tools (Hive Moderation, Illuminarty, or AI or Not) and consider the aggregate result.
4. Check metadata on downloadable images using an EXIF viewer. Look for camera information that indicates a real photo.
5. Review their online presence — social media history, website age, number and quality of reviews, interaction with other providers. Depth of presence is hard to fake.
6. Check platform verification — is the profile verified by the platform? What does that specific platform's verification process actually entail?
7. Request a verification selfie with a specific word and today's date written on paper, or a short video with a specific gesture.
8. Assess the aggregate — no single check is definitive, but a provider who passes all seven steps is almost certainly real. Red flags at multiple steps suggest significant risk.

---

---

The AI Arms Race: What's Coming Next

AI image generation is advancing at an extraordinary pace. Every few months, new models reduce the number of artifacts and make detection harder. Here's what to watch for in the near future:

Real-Time Video Deepfakes

As of early 2026, real-time video deepfakes exist but require significant computing power and produce noticeable artifacts — lag when moving quickly, blurring around face edges, lighting inconsistencies. These limitations make video verification still effective as a defense. However, the technology is improving rapidly. Within the next 1–2 years, real-time video deepfakes may become convincing enough to fool casual observers. This means video verification will need to become more sophisticated — requesting specific, unpredictable actions (hold up a specific number of fingers, write a specific word on paper) will remain effective longer than simple "wave at the camera" requests.

AI Voice Cloning

Voice cloning technology can now replicate a person's voice from just a few seconds of sample audio. This means phone verification alone may become unreliable if a scammer has access to audio samples from a real provider. The combination of video and voice verification provides more protection than either alone, but be aware that even this dual approach has a shelf life.

Blockchain and Cryptographic Verification

Some platforms are exploring cryptographic verification systems where providers submit verified photos with embedded cryptographic signatures that can't be replicated or transferred. This technology is still in its early stages but may become a standard part of the platform verification ecosystem. Supporting platforms that invest in these technologies helps accelerate their adoption across the industry.

Community Intelligence

As AI tools become more sophisticated, community-based verification may become the most reliable defense. Provider review communities, client forums, and trusted networks of verified individuals create a web of social proof that's extremely difficult to fabricate at scale. Engaging with these communities — contributing reviews, participating in discussions, and building relationships — gives you access to collective intelligence that no individual tool can match.

---

The Bottom Line

AI-generated images will only get better. The visual tells described in this guide will become less reliable as generators improve. The enduring defense is verification — confirming that a real person exists behind every profile through live interaction, platform verification, and consistent social proof. No amount of pixel analysis substitutes for seeing a real person move, speak, and respond in real time.

Make verification a standard part of your vetting process, not an exception. The few minutes it takes to request a verification selfie or check a platform badge is insignificant compared to the cost of falling for a scam or putting yourself in a dangerous situation with someone who isn't who they claim to be. As the technology evolves, your vetting practices need to evolve with it — but the fundamental principle remains the same: trust evidence, not images.

For related reading, see our Vetting & Verification chapter, Photo Verification guide, and Scam Detection guide.