The AI Lover That Doesn't Exist: How Romance Scam Chatbots Are Fooling People for Months

Romance scams cost Americans $1.14 billion in 2023, according to the FTC. Now they're getting harder to detect, because the person you're talking to may not be a person at all.

A romance scam AI chatbot is a large language model deployed by criminal networks to build fake emotional relationships at scale. These systems run 24 hours a day, remember everything you share, and never break character. Research shows they outperform human scammers at gaining trust and extracting compliance.

This guide explains how these chatbots operate, what behavioral patterns give them away, and what to do if you think you've already been targeted.

Introduction: the relationship that was never real

Picture this. You matched with someone six weeks ago. They message you every morning without fail. They remember your sister's name, the job stress you mentioned on a Tuesday night, the trip you've always wanted to take. The conversation feels real. Because to you, it is.

Here's the part that's hard to sit with: the person on the other end may not exist at all.

What you're experiencing could be a large language model (LLM) running on a remote server, managing dozens of identical "relationships" at the same time. It doesn't sleep. It doesn't forget. And it's been trained to say exactly the right thing at exactly the right moment.

Romance scams have always preyed on loneliness. But AI chatbots have changed what these scams look like and how long they can run. The clunky, obviously scripted messages of the past have been replaced by emotionally adaptive conversations that can sustain themselves for months without a single human operator involved.

The financial damage is real. The FTC reported that Americans lost $1.14 billion to romance scams in 2023 alone, with a median individual loss of $2,000. That made romance scams the highest-loss fraud category that year. And those are only the cases people reported.

This guide breaks down exactly how romance scam AI chatbots work, what red flags to watch for, and the concrete steps you can take right now to verify whether someone you've met online is genuinely who they claim to be.

If something feels slightly off about an online relationship, trust that feeling.

What makes AI romance scam chatbots different from old-school scammers

The old version of this scam had a human problem. Somewhere in a compound in Cambodia or Myanmar, a trafficked worker sat at a desk for 14 hours managing a handful of fake romantic personas across WhatsApp and Instagram. They got tired. They made spelling mistakes. They forgot what they'd told you last Tuesday.

That model still exists. But it's being replaced by something that doesn't sleep, doesn't forget, and doesn't have a shift change at midnight.

The same organized crime syndicates that built the pig butchering crypto scam industry have now automated the most labor-intensive part of their operation: the conversation itself.

Large language models (LLMs), the same technology behind tools like ChatGPT, can generate emotionally convincing, contextually aware messages at scale. A single criminal operation can run hundreds of simultaneous "relationships" with consistent personas, remembered details, and responses that adapt to your emotional state in real time. What once required a trafficked workforce of thousands now runs at near-zero marginal cost per conversation.

The research backs this up. A peer-reviewed study published in December 2025, "Love, Lies, and Language Models" by Gressel et al. from Ben Gurion University and partner institutions, found that 87% of scam labor consists of systematized conversational tasks readily susceptible to LLM automation. The same study found that an LLM scam agent earned greater trust from participants than human operators, and achieved 46% compliance with financial requests compared to just 18% for humans.

The AI isn't just cheaper. It's more effective.

The language barrier that once tripped up scammers working from Southeast Asia is gone too. OpenAI confirmed in February 2025 that it banned a cluster of accounts originating in Cambodia that were using its models to generate fluent English, Japanese, and Chinese romance scam conversations in real time. The operators' primary language was Chinese. Their victims never knew.

The scale of the criminal infrastructure behind this is hard to overstate. The UNODC reported a 600% rise in deepfake-related criminal services advertised on platforms like Telegram between February and July 2024 alone. These aren't isolated bad actors. They're organized syndicates upgrading their toolset the same way any business would adopt new software.

Here's the kicker on economics: human scam operations are expensive, logistically complex, and increasingly exposed to law enforcement raids. An LLM-powered operation has almost no marginal cost per additional victim. Every new conversation is essentially free.

This is why the romance scam AI chatbot isn't a niche threat. It's an industrial one.

The pig butchering connection: organized crime goes automated

The name comes from the Chinese phrase sha zhu pan, which translates roughly to "killing pig plate." The idea is deliberate: fatten the pig first, then slaughter it. In practice, that means building emotional trust over weeks or months before moving in for the financial kill.

This isn't the work of lone scammers. Pig butchering operations are run by organized criminal enterprises, many operating out of fortified compounds in Myanmar, Cambodia, Laos, and the Philippines. A significant number of the workers inside those compounds are themselves victims, trafficked and forced to run scams under threat of violence.

The structure follows three phases:

• Hook: Mass outreach across dating apps and social platforms to find emotionally available targets
• Line: Weeks of trust-building, manufactured intimacy, and personal disclosure to deepen the emotional bond
• Sinker: A pivot to a fake crypto investment platform, a direct financial request, or a malicious link disguised as something legitimate

What's changed is the automation of all three phases. A March 2026 report from the U.S.-China Economic and Security Review Commission (USCC) found that scam centers are now using LLMs and AI image generation software to scale their operations and create more convincing personas. Operations that once required shifts of human workers can now run continuously across hundreds of targets at once, with no drop in consistency.

American losses from these Chinese criminal group-operated scam centers topped $10 billion in 2024 alone. Automating pig butchering hasn't just made these scams cheaper to run. It's made them harder to detect.

Why AI chatbots are actually better at this than humans

Here's the part that should make you stop scrolling: in a controlled study, people trusted an AI scammer more than a real human one.

Researchers from four universities ran a blinded, seven-day conversation study in early 2025, published as Gressel et al. via USENIX Security 2026. Twenty-two participants chatted daily with two "people," one human, one an AI agent built on a commercial LLM. Neither was identified. At the end of the week, the AI scored significantly higher on trust (p=0.007), averaging 3.78 versus 3.31 for the human. When both partners made a request asking participants to download an app, the AI achieved 46% compliance. The human managed 18%.

That gap isn't a fluke. It reflects something real about how these systems work.

An AI never has a bad day. It doesn't get impatient at 2am, accidentally contradict something it said three weeks ago, or lose track of which "relationship" it's managing. It mirrors your vocabulary, matches your emotional tone, and remembers every detail you've shared: your dog's name, your difficult week at work, the trip you mentioned wanting to take. It's available whenever loneliness peaks, including 3am when you can't sleep.

These aren't bugs in the technology. They're features being deliberately exploited.

The kicker: the same study tested commercial safety filters from major AI providers against hundreds of simulated romance-baiting conversations. According to Help Net Security's coverage of the research, detection rates ranged from 0% to 18.8%, and none of the flagged conversations were correctly identified as scam content. The tools built to catch this missed it entirely.

How these chatbots build trust over weeks and months

You've been messaging someone for six weeks. They remember that your sister's wedding is coming up. They ask how your job interview went. They send a good morning text every single day. It feels real. And that's exactly the point.

This is how romance scam AI chatbots work. Not with a single dramatic lie, but with a slow, patient accumulation of warmth. The FTC reported that Americans lost $1.16 billion to romance scams in just the first nine months of 2025. Most of those victims didn't hand over money on day one. They were groomed over weeks, sometimes months, before a single financial ask appeared.

Here's how the arc typically unfolds.

Phase 1: The opening move

The first message arrives on a dating app, in an Instagram DM, or as a "wrong number" text that somehow turns into a conversation. The profile looks plausible. The opening message feels personal, like they actually read your profile. That's not coincidence. It's a tested opening optimized across thousands of previous interactions.

Phase 2: Love bombing

Within days, the pace picks up. Good morning texts. Questions about your day. Compliments that feel specific and genuine. The messages arrive at all hours because the bot never sleeps. According to the Global Cyber Alliance, AI removes many of the classic red flags people expect: the grammar mistakes, the awkward timing, the emotional inconsistency. What's left feels smooth, attentive, and real.

Phase 3: Deepening intimacy

Over the following weeks, the "person" starts sharing personal struggles. A difficult childhood. A health scare they haven't told anyone else about. These manufactured vulnerabilities are designed to trigger your empathy and create a sense of shared trust.

The bot also remembers everything you've shared. Your job, your family situation, your anxieties. It weaves these details back into conversation naturally, creating the feeling of a shared history. That's not emotional intelligence. That's a context window doing exactly what it was built to do.

Phase 4: Isolation and dependency

Subtly, the dynamic shifts. The "person" becomes your primary emotional support. They might express mild hurt when you mention spending time with friends, or suggest that the people around you don't really understand you the way they do. You don't notice the drift. You just feel closer to them than to almost anyone else in your life.

Phase 5: The pivot

Then something changes. A crisis emerges. A business opportunity appears. A link gets shared. By this point, the emotional investment is so deep that the request feels reasonable. You've been talking for months. You trust this person.

This is the moment the scam was always building toward. Everything before it was infrastructure.

The most unsettling part? You don't have to be lonely or naive for this to work. These systems are built to exploit the very things that make us human: the desire to connect, to be seen, and to matter to someone.

The emotional manipulation playbook: tactics AI chatbots use

Let's be clear before we go any further: victims of romance scams aren't naive. They're targeted. These aren't random accidents. They're the result of systems built to exploit the most normal parts of human psychology: the desire to feel seen, loved, and needed.

AI makes each of these tactics more effective because it never gets tired, never loses patience, and never breaks character. Here's exactly what that looks like.

Tactic 1: Love bombing

The conversation starts fast and intense. Compliments, affection, and deep interest arrive in a flood. A real person couldn't sustain this pace for long. An AI chatbot can. It sends good morning messages every day, remembers what you said you were worried about last Tuesday, and checks in at exactly the right moment. The attention feels extraordinary because, in a real relationship, it would be.

Tactic 2: Mirroring

Scam chatbots are trained on vast volumes of real human conversation, which means they're good at picking up your tone and reflecting it back. If you're playful, the bot gets funny. If you're serious, it matches that register. If you mention loving hiking or hating your commute, those details get woven back into future messages. The result is a disarming sense of being truly understood. Global Cyber Alliance notes that this mirroring is engineered, tested, and optimized across thousands of interactions.

Tactic 3: Manufactured vulnerability

At some point, the 'person' opens up. A difficult childhood. A painful divorce. A fear they've never told anyone else. This manufactured intimacy does two things: it makes you feel trusted, and it makes you feel responsible for their emotional wellbeing. That's a heavy anchor. And it's entirely fabricated.

Tactic 4: Urgency and scarcity

Once emotional investment runs deep enough, a crisis appears. A medical emergency. A frozen bank account. A once-in-a-lifetime investment window closing in 48 hours. These scenarios are built to bypass rational thinking. When you care about someone, you act fast. That's the point.

Tactic 5: Isolation

The comments are subtle at first. 'I feel like you're the only one who really gets me.' 'I don't want to share what we have with anyone else.' Over time, this quietly discourages you from talking to friends or family about the relationship, exactly the people who might spot what's happening.

Tactic 6: The sunk cost trap

After weeks or months of daily messages, shared secrets, and genuine emotional investment, walking away feels like losing something real. According to Finextra, the sunk cost fallacy keeps victims engaged long after warning signs appear, because the emotional cost of leaving feels greater than the cost of staying.

That's by design. The longer the relationship runs, the harder it is to accept it was never real.

A human scammer gets tired. They make mistakes. They go off-script. An AI chatbot does none of those things. It's consistent, available around the clock, and perfectly calibrated to keep you engaged. That's what makes it so effective, and so hard to walk away from.

Behavioral red flags that suggest you are talking to an AI

Not every red flag means you're talking to a bot. But some patterns are hard to explain any other way.

It helps to split these warning signs into two groups: the ones that apply to any online scammer, human or AI, and the ones specific to AI-driven personas. Knowing the difference helps you respond more accurately to what you're actually dealing with.

Red flags that apply to any online scammer

These are the classic signs that something is off, regardless of whether there's a human or a machine on the other end:

• Never willing to meet in person, despite weeks or months of conversation
• Always has a reason to cancel video calls: a broken camera, bad connection, work emergency
• Claims to be working overseas in a role that conveniently prevents visits: military deployment, oil rig, remote medical posting
• Asks for money, gift cards, wire transfers, or cryptocurrency, often framed as a temporary emergency
• The relationship escalates emotionally far faster than normal: declarations of love within days, talk of a future together within weeks

If you're seeing two or more of these, that's already enough to slow down and ask questions.

Red flags specific to AI chatbots

AI-driven personas have tells that human scammers don't, because the technology has its own limitations and patterns.

• Responses arrive at oddly consistent speeds, day or night, with no variation for time zones or sleep schedules
• Grammar and vocabulary are unusually polished for someone who claims limited education or says English isn't their first language. As the Global Cyber Alliance notes, AI removes many of the classic linguistic red flags people have traditionally relied on
• Deflection on hyper-specific local questions: ask them to name a restaurant near their home, describe their neighborhood, or recall a local news story. Vague or generic answers are a signal
• Profile photos look slightly too perfect: symmetrical lighting, flawless skin, no candid shots, no tagged photos from friends. Check for unnatural backgrounds, oddly rendered hands, or accessories that shift between images
• Photos pass a reverse image search but still feel off: AI-generated faces often do. Graphika's 2026 research found that AI-generated dating profile images are convincing at a glance but reveal detectable artifacts on closer inspection
• Emotional intensity keeps climbing without any real-world meetup attempts: the connection deepens on paper, but nothing ever moves forward
• Off-script questions get generic or slightly delayed responses: try asking something unexpected and watch how the reply lands
• The conversation pivots to financial topics or a shared link after a period of emotional buildup
• They can send pre-recorded video messages but claim they can't do a spontaneous live call

Here's the kicker: even a live video call isn't the safety net it used to be. AI deepfake tools now allow real-time face-swapping during video calls, meaning the face you see may not be the person you're talking to. A video call confirms there's a camera. It doesn't confirm there's a real person behind it.

Red flags in the conversation itself

Real conversations are messy. People trail off, send half-formed thoughts, misread questions, and occasionally reply with nothing but "lol." AI doesn't do any of that.

Here's what to watch for in the chat itself:

• Responses feel over-complete. You ask a casual question and get a three-paragraph answer. Real texting doesn't work that way. An AI is built to be helpful, and that helpfulness is a tell.
• The reply addresses a slightly different question than you asked. It's subtle, but the bot is pattern-matching to your message, not actually listening. The answer sits adjacent to what you meant, not quite on it.
• Emotional declarations arrive too fast. "I've never connected with anyone like this" in week two isn't romantic. It's a script. Genuine feelings take time and awkwardness to develop.
• They never have a bad day. No short replies. No distracted responses. No "sorry, crazy day at work." A real person is inconsistent. An AI is relentlessly present and even-keeled.
• Empathy that feels slightly generic. When you share something painful, the response is warm but oddly smooth. There's none of the fumbling that comes when a real person tries to find the right words.
• The conversation always has somewhere to go. There's no small talk that just fizzles. Every exchange nudges the relationship forward. That's not chemistry. That's a narrative being managed.

A peer-reviewed study from Ben-Gurion University found that LLM agents generated higher trust scores than human scam operators in long-term conversation tests. Trust your gut when something feels slightly off. That instinct is worth listening to.

How to verify whether someone online is genuinely who they claim to be

Wanting to verify someone you've met online isn't paranoid. It's smart. Here's a practical toolkit you can use right now, before any emotional or financial damage is done.

Step 1: Reverse image search the profile photo

Go to Google Images or TinEye. Right-click the profile photo and select "Search image." If that face shows up on stock photo sites, other dating profiles, or unrelated websites, it's almost certainly stolen. The FTC recommends this as a first-line check for anyone you've met online.

Step 2: Check whether the photo is AI-generated

A clean reverse image search doesn't mean the photo is real. It may have been generated from scratch. Look for telltale signs: slightly asymmetrical facial features, blurred or distorted edges around hair and ears, odd lighting that doesn't match the background, or hands with the wrong number of fingers. AI still struggles with hands.

Tools like Hive Moderation's AI image detector can scan a photo and flag synthetic content. Google's SynthID Detector identifies images created by Google's own AI tools. Neither is foolproof, but both add a useful layer of scrutiny.

Step 3: Run the photo through FaceCheck.ID

FaceCheck.ID is a facial recognition search engine that scans the web for matching faces. Upload the profile photo and see if that face appears elsewhere under a different name. It's particularly useful for catching scammers who reuse stolen photos across multiple fake personas.

Step 4: Request a spontaneous live video call with a specific action

This is one of the most reliable tests available. Ask them to wave with their left hand, hold up three fingers, or turn their head fully to one side. Do it without warning. Deepfake tools struggle with unexpected movement and unusual angles. A refusal, a technical excuse, or a camera that's suddenly "broken" is a significant red flag.

Note: even a live video call isn't guaranteed proof anymore. Scammers now use real-time deepfake overlays that can mimic a face during a call. The spontaneous action test disrupts most of these tools, but treat video as one signal, not final confirmation.

Step 5: Ask hyper-local questions

Ask them to name a specific street near where they claim to live, describe a local landmark, or recommend a restaurant in their neighborhood. AI personas built on generic backstories can't answer these convincingly. Vague, deflecting answers to simple local questions are a warning sign.

Step 6: Cross-reference their social media history

Check when their accounts were created. Look at the depth of their friend network. Real people have years of posts, tagged photos, and genuine interactions with other real accounts. A profile that's six months old, has 40 connections, and features only polished photos with no candid moments deserves scrutiny.

Step 7: Never click a link without checking it first

This is where a romance scam often crosses from emotional manipulation into active financial threat. If someone you've met online shares a link to a "verification page," photo album, or investment platform, don't click it directly. Copy the URL and run it through a link scanner first.

Better still, use a browser security tool that does this automatically. Guardio's browser extension scans URLs in real time and blocks access to known phishing and scam sites before any page loads, before you enter any credentials or personal information. If a scammer sends a malicious link, Guardio catches it at the moment it matters most.

The link moment: when a romance scam becomes a digital threat

Everything up to this point has been emotional. But there's a moment when the scam shifts from manipulation to active attack. The 'person' sends a link.

It might be framed as a photo album they want to share, an identity verification page the dating platform supposedly requires, a cryptocurrency investment dashboard they're excited about, or a document that needs your signature. It looks plausible. You're already emotionally invested. You click without a second thought.

That click is where the scam becomes a digital threat.

Those links lead to phishing pages built to harvest your login credentials, financial details, or personal data. Some install malware silently in the background. The page looks real because it's designed to. And by the time you realize something is wrong, the damage is already done.

This is exactly where a browser security tool earns its place.

Guardio's browser extension scans every URL in real time, checking it against a continuously updated database of known malicious sites. That includes newly registered phishing domains, which traditional antivirus tools frequently miss because they haven't been flagged yet. If a link is dangerous, Guardio blocks the page before it loads. You see a warning instead of a credential-harvesting form.

The detection engine is backed by Guardio Labs, our in-house research team, which continuously tracks emerging threat patterns, including the AI-generated scam infrastructure that powers romance scam chatbots. Those findings feed directly back into what the product detects.

In independent testing by PCMag, Guardio achieved a perfect 100% phishing detection rate. For context, PCMag's review of Aura recorded an 80% detection rate in the same evaluation category.

That gap matters. A link you click in a moment of trust shouldn't be the moment you lose control of your accounts.

What to do if you think you have been targeted or already sent money

Finding out the person you've been confiding in for months isn't real is genuinely painful. It's not embarrassing. It's not a sign you were careless. These systems are built by organized crime specifically to be convincing, and they work on people at every level of tech-savviness.

If something feels wrong, or you already know it is, here's what to do.

Step 1: Stop contact immediately

Block the account across every platform. Don't try to confront the scammer or negotiate. Don't send more money in exchange for a promise to return what you've already lost. That's a second layer of the same scam.

Step 2: Don't send more money, no matter what

Scammers often follow an initial payment with escalating pressure, sometimes threatening to share private photos or personal information if you don't comply. This is called sextortion, and it's a tactic, not a consequence you can prevent by paying. Report instead.

Step 3: Report to the FTC

Go to reportfraud.ftc.gov. This is the primary US consumer fraud reporting channel. Your report helps the FTC track patterns, warn others, and build cases against scam networks. According to FTC data from April 2026, nearly 60% of people who reported losing money to a romance scam in 2025 said it started on a social media platform.

Step 4: Report to the FBI's IC3

Go to ic3.gov. The Internet Crime Complaint Center handles cybercrime reports, including romance scam cases. The FBI's 2025 Internet Crime Report recorded over $19 million in losses tied to confidence and romance scams with an AI connection, and that figure almost certainly undercounts the real total.

Step 5: Report to the dating platform

Use the platform's built-in reporting tools to flag the profile. This protects other users who may be in the same conversation right now.

Step 6: Contact your bank or payment provider fast

If money was sent via wire transfer, call your bank's fraud department immediately. Transfers caught within 24 to 72 hours can sometimes be reversed, though the window is narrow. Credit card payments carry stronger protections. Cryptocurrency recovery is unlikely, but reporting to the platform and to IC3 is still worth doing.

Step 7: Monitor for identity breaches

If you shared personal details, clicked any links, or entered credentials at any point during the relationship, your data may already be exposed. Scammers frequently use those moments to harvest information for identity fraud. Guardio's Identity Breach Monitoring alerts you in real time if your personal information surfaces in a breach, so you can act before the damage compounds.

You don't have to process all of this alone. The FTC's romance scam resource page includes guidance on next steps and emotional support resources. What happened to you was a crime. Treat it like one.

How Guardio protects you when a scam conversation turns dangerous

Picture the moment it happens. You've been talking to someone for six weeks. You trust them. Then a link arrives: a photo-sharing page, a "verification" form, or an investment platform they want you to join. You click it. The page looks real. You type in your email and password. The damage is done before you realize something was wrong.

That's the moment romance scams stop being emotional and start being digital. It's the moment Guardio is built for.

With Guardio's browser extension installed, that scenario plays out differently. The extension scans the URL in real time, before the page loads. If it matches known phishing infrastructure, a newly registered suspicious domain, or a scam pattern flagged by the Guardio Research Team, the page is blocked. You see a clear warning. You don't enter anything.

This matters specifically for AI-generated threats. The fake pages that romance scam chatbots link to aren't old, recycled phishing sites. They're freshly built, convincingly designed, and often invisible to legacy threat databases. Guardio's detection engine analyzes behavioral patterns and structural signals, not just known bad URLs, which means it catches threats that haven't been seen before.

The scale behind that protection is real. According to Guardio's own data, the platform protects over 1.5 million people and blocked more than 964 million malicious sites in 2025, about 30 sites every second. Guardio also flagged 267 million phishing and scam emails, filtered 116 million scam and spam texts, and blocked 244,000 malicious downloads and 54,000 malicious browser extensions that same year.

More than 13,000 users have reviewed Guardio on Trustpilot, and the reviews consistently describe the same thing: a threat was caught before any damage occurred.

If you're in an online relationship and something feels slightly off, the free scan takes seconds. It shows you exactly where you're exposed right now, before a link arrives.

Get started with a free scan.

Protecting elderly relatives and less tech-savvy family members

If you have a parent, grandparent, or older relative who's active on Facebook or a dating app, this section is for you.

According to the FTC's 2025 annual report to Congress, total fraud losses among people aged 60 and over reached approximately $2.4 billion in 2024, a fourfold increase from 2020. Older adults don't get caught more often than younger people. But when they do, they lose far more. Retirement savings and decades of accumulated assets are simply available to move in a way that a younger person's account often isn't.

The emotional manipulation is identical to what any victim experiences. The vulnerability is often higher because someone who is recently widowed, newly divorced, or using social platforms for the first time is less likely to have seen these tactics before.

How to have the conversation without making it an interrogation

The worst thing you can do is swoop in with warnings and suspicion. That approach makes people feel surveilled, not protected, and it often pushes them to be less open with you.

Lead with curiosity instead. Ask about the person they've been talking to. Show genuine interest. Then, gently, suggest doing something together: "That's exciting. Want to look them up together? I do this with everyone I meet online."

Framing verification as a shared activity, rather than an intervention, keeps the relationship intact and gets the job done.

Questions to suggest they ask any new online contact:

• "Can we do a quick video call? I'd love to see your face."
• "What's the name of the street you grew up on?" (Follow up with a Google Maps check.)
• "Can you send me a photo holding today's newspaper?"

These aren't interrogation tactics. They're normal things people ask when they're genuinely interested in someone.

How Guardio's family plan helps here

The honest challenge with less tech-savvy relatives is that they can't always manage their own security settings. They click links. They fill in forms. They don't always know what a phishing page looks like.

Guardio's family plan covers multiple household members under one subscription. That means the same real-time link protection and identity breach monitoring that protects you also protects them, without requiring them to configure anything themselves. If a romance scam chatbot sends your relative a link to a fake "verification" page, Guardio flags it before the page loads.

You set it up once. It runs quietly in the background. That's the kind of protection that actually works for people who aren't thinking about security every day.

Key takeaways: what to remember

Here's what this guide comes down to:

• AI chatbots now outperform human scammers at building emotional trust. A peer-reviewed study by Gressel et al., published at USENIX Security 2026, found that AI agents achieved 46% compliance from targets, compared to just 18% for human operators. These aren't clunky bots. They're genuinely convincing.
• The financial damage is real and growing. The FTC reported $1.14 billion in romance scam losses in 2023 alone, with a median individual loss of $2,000. That's reported losses. The actual figure is almost certainly higher.
• The financial ask comes late. Weeks or months of emotional investment come first. By the time money enters the conversation, victims are deeply attached.
• The link moment is when things turn dangerous. Emotional manipulation becomes active digital theft the second you click a link to a fake verification page, investment platform, or photo-sharing site.
• Verification tools are available right now. Reverse image search, AI image detectors, spontaneous video calls with a specific action request, and hyper-local questions can all expose a fake persona before any damage is done.
• If you've been targeted, report it. File a report with the FTC at reportfraud.ftc.gov and the FBI at ic3.gov. Then stop all contact.
• A browser security tool is your last line of defense. When everything else fails, real-time URL blocking catches the moment a scam conversation turns into a credential or financial theft attempt.

Being targeted by one of these scams says nothing about your intelligence or character. These are sophisticated systems built by organized crime, engineered to exploit human connection. Awareness is your first defense. The right tools are your second.

Conclusion

AI-powered romance scams are more convincing than most people expect. The chatbots don't slip up the way human scammers do. They're patient, consistent, and built to exploit trust. Knowing the red flags, verifying identities before any money or personal data changes hands, and acting quickly if something feels wrong are your best defenses. The scam often turns dangerous the moment a link appears in the conversation.

Don't Let a Scam Link Be the Last Click You Regret

When a romance scam escalates from conversation to a link, Guardio's browser extension blocks it in real time before any credentials or financial information are entered. Run a free scan now and see exactly where you're exposed.

Get Started with a Free Scan