How can I be sure the person selling me tickets online is real, not AI?

When a stranger offers concert seats at a price that feels almost too convenient, the oldest question in commerce takes a new form: is the seller even human? Confirming a real person online selling tickets used to mean checking a profile photo, reading reviews, and trusting your instincts. That playbook is now obsolete. Generative tools can spin up a believable face, a verified-looking history, and even a live video call where the "seller" smiles, nods, and answers questions in real time. For fraud teams running peer-to-peer marketplaces and ticket resale platforms, the consumer's anxiety is a direct signal of an operational gap: the identity layer was never designed to answer whether a face belongs to a living body.

The U.S. Federal Trade Commission reported consumers lost more than $12.5 billion to fraud in 2024, a 25 percent increase year over year, with online shopping among the largest complaint categories. Pindrop's 2025 analysis found AI-enabled fraud rose more than 1,200 percent over the prior year, far outpacing the growth of conventional fraud.

Verifying a real person online selling tickets in the age of synthetic identity

The core problem is that almost every consumer-facing trust signal is now forgeable. A scammer can clone a face from a single public photo, generate a unique non-existent person, or hijack a real seller's likeness for a video chat. The phrase "real person online selling" has quietly shifted from a description to a verification requirement. Buyers want assurance the account is operated by a live human, not an automated identity farm or a deepfake puppet, and platforms increasingly need that assurance at scale, on consumer-grade webcams and phone cameras, without adding friction that kills conversion.

Traditional defenses operate on the wrong layer. Document checks confirm a credential exists. Behavioral analytics flag suspicious patterns after the fact. Reputation systems reward longevity, which patient fraud rings simply buy or age. None of these answer the physiological question at the heart of buyer anxiety: is there a heartbeat behind the face on the screen?

That question is where remote photoplethysmography, or rPPG, enters marketplace security. rPPG reads the tiny color changes in skin caused by blood pumping through capillaries with each heartbeat. A genuine human face, captured on a standard camera, carries this pulse signal distributed naturally across the cheeks and forehead. A generated face, a static image, a replayed video, or a real-time face swap struggles to reproduce a coherent, physiologically plausible blood-flow pattern across the whole face.

How verification methods compare for marketplace sellers

Marketplace and resale fraud teams typically evaluate verification options against friction, cost, and resistance to synthetic media. The table below summarizes how the common approaches hold up when the adversary is an AI-generated or deepfaked identity.

Verification method	What it confirms	Friction for seller	Resistance to AI / deepfake sellers
Profile photo and reviews	A picture and a history exist	None	Very low - photos and reviews are trivially faked
Phone or email verification	A contactable channel exists	Low	Low - disposable numbers and inboxes are cheap
Document upload	A credential image exists	Medium	Low to medium - templates and synthetic IDs circulate widely
Active liveness (blink, turn head)	A face can follow prompts	Medium to high	Medium - advanced deepfakes can animate on cue
Passive rPPG liveness	A live human pulse is present	Very low	High - synthetic faces lack coherent blood-flow signals

The pattern is consistent: the methods buyers trust most are the easiest to fake, while the signal hardest to fake is the one most platforms do not yet measure. A layered stack closes that gap.

Key takeaways for fraud and trust-and-safety teams:

• Static signals (photos, reviews, account age) authenticate a record, not a person.
• Active challenge-response liveness adds friction and is increasingly bypassed by real-time animation.
• Passive physiological liveness runs silently during an existing video step, preserving conversion.
• No single layer is sufficient; physiological signals work best alongside document and behavioral checks.

Industry applications across peer-to-peer platforms

Ticket and event resale marketplaces

Ticket resale is a high-velocity, high-emotion category where buyers act fast before an event sells out. That urgency is exactly what fraud rings exploit. Embedding a passive liveness check at seller onboarding, or before a high-value listing goes live, lets platforms confirm a real person online selling the inventory rather than an automated identity created minutes earlier. Because rPPG can run during a brief selfie-video capture, it does not force the seller through a clumsy challenge sequence.

General peer-to-peer commerce

Classified and resale apps face industrial-scale account farming, where thousands of synthetic profiles are generated to post scam listings. Pulse-based liveness raises the cost of each fake account dramatically, because a generated face cannot simply be screenshotted into existence; the system expects a living physiological signal that synthetic media does not reliably carry.

Identity verification and KYC vendors

For the vendors and fraud teams selling verification as a service, blood-flow liveness is becoming a differentiator. It addresses the specific failure mode their clients fear most: a deepfake that passes a document check and a face match but has no body behind it. Adding a physiological layer answers the buyer-protection question marketplaces are now being held accountable for.

Current research and evidence

The research base for physiological liveness has matured quickly. A 2024 comprehensive review of deepfake detection using remote photoplethysmography, published through researchers associated with Torrens University Australia and indexed on IEEE Xplore, catalogued how rPPG-based methods exploit the absence or inconsistency of pulse signals in fabricated video. The review also documented the arms race candidly: detection accuracy is affected by lighting, camera quality, motion, and video compression.

Importantly, the field is not standing still on either side. A 2025 study highlighted by Frontiers and EurekAlert, summarized under the title "high-quality deepfakes have a heart," found that the most advanced deepfakes can now inherit a detectable heartbeat from their source "driver" video. The researchers showed that a naive pulse-presence test is no longer enough, because a sophisticated fake may show a valid global heart rate.

The response from the research community has been to move from detecting whether a pulse exists to analyzing how blood flow is spatially distributed across the face. Genuine human physiology produces a coherent pattern of perfusion across distinct facial regions; copied or stitched signals tend to break down spatially even when a global rate looks plausible. Work such as the BioVerify approach on ResearchGate and related 2024 to 2025 studies combine rPPG with texture and motion cues to harden detection against this newer generation of fakes. The practical lesson for fraud teams is that physiological liveness is powerful but must be implemented as spatial, multi-region analysis rather than a single heart-rate readout.

The future of real person online selling verification

Three trends will shape marketplace identity over the next several cycles. First, regulatory and consumer pressure will push platforms toward demonstrable proof of human sellers, not just terms-of-service promises. With the FBI's Internet Crime Complaint Center tracking AI-enabled fraud losses into the hundreds of millions and rising, "we checked their ID" will not satisfy regulators or defrauded buyers.

Second, verification will become continuous rather than a one-time gate. A seller may verify once, but the account, the device, and the high-value listing each represent a moment where re-confirming a live human becomes worthwhile. Passive methods make repeated checks tolerable because they add little friction.

Third, the contest will keep escalating. As deepfakes learn to mimic a heartbeat, detection will lean on signals that are harder to synthesize, such as the spatial coherence of blood flow, micro-variations in perfusion timing, and fusion with device and behavioral signals. The platforms that win buyer trust will be those treating "is this a real person?" as a measurable, physiological question rather than a matter of profile aesthetics.

Frequently asked questions

How can a marketplace actually tell a seller is a real person and not AI?

The most reliable signal is physiological liveness. Techniques like remote photoplethysmography read the subtle skin-color changes caused by blood flow during a short video capture. A real living face carries a coherent, full-face pulse pattern that synthetic images, replayed clips, and most real-time face swaps cannot reproduce, especially when the system analyzes how blood flow is distributed across multiple facial regions rather than just checking for a single heartbeat.

Can deepfakes fake a heartbeat to beat liveness detection?

Recent research, including a 2025 study summarized by Frontiers and EurekAlert, shows advanced deepfakes can inherit a believable global heart rate from their source video. That is why modern detection no longer relies on pulse presence alone. It examines the spatial coherence of blood flow across the face and fuses physiological signals with texture, motion, and device cues, which copied signals struggle to satisfy simultaneously.

Does adding a real-person check hurt conversion for legitimate sellers?

It does not have to. Passive liveness methods run silently during a selfie-video step that many platforms already use, with no blink-and-turn choreography. Because the check happens in the background of an existing flow, legitimate sellers experience little added friction while fraudulent automated identities face a barrier they cannot cheaply clear.

Why are document checks not enough on their own?

A document check confirms that a credential image exists, not that a living human is present. Synthetic and template-based IDs circulate widely, and a deepfake can pass a face-to-document match while having no body behind the screen. Pairing document verification with physiological liveness closes that gap by requiring evidence of an actual living person.

Circadify is addressing this exact space, building rPPG-based liveness that reads real blood flow to separate living humans from synthetic and deepfaked identities in marketplaces and verification pipelines. Fraud and trust-and-safety teams evaluating how to confirm a real person online selling on their platform can explore an enterprise security demo to see blood-flow liveness in action.