AI Is Enabling the Largest Wave of Click Fraud in Marketing History

Table of Contents

Introduction: The Invisible Tax on Every Digital Marketer

Picture this. You wake up on a Tuesday morning, grab your coffee, and open your ad dashboard with optimism. The numbers look good. Your click-through rate is up 18 percent from last week. Your impressions are climbing. Your cost per click is exactly where you want it. You feel a quiet sense of pride. Your campaigns are working.

But here is the uncomfortable truth that nobody in the ad tech ecosystem wants you to fully understand. Up to 40 percent of the traffic you just paid for never came from a human being. Those clicks you are celebrating? A significant portion of them were generated by artificial intelligence. They came from bots that mimic human behavior so convincingly that even sophisticated fraud detection systems struggle to flag them. And while you are sipping that coffee feeling good about your performance metrics, somewhere in a server farm or a ghost click farm operation, your budget is being systematically drained by actors who have mastered the art of algorithmic deception.

This is not a fringe problem affecting careless advertisers. This is a systemic crisis that touches every corner of the digital advertising ecosystem. In 2025, when global digital ad spend surpassed $750 billion for the first time, advertisers lost a conservative estimate of $165 billion to invalid traffic. Other estimates place the figure even higher. Spider Labs’ 2026 Ad Fraud White Paper, which analyzed more than 6 billion clicks and $6.2 billion in ad spend across 242 countries, calculated global ad fraud losses at $32.6 billion for 2025 alone. TrafficGuard puts the global figure at approximately $250 billion. The IAB estimated $84 billion in losses for 2025. Juniper Research projects that losses will exceed $100 billion in 2026 and reach a staggering $172 billion by 2028.

These numbers are not abstract. They represent real money. Your money. Money that was supposed to acquire customers, build brand awareness, and drive growth. Instead, it is flowing into the pockets of fraudsters who have weaponized the very technology that was supposed to make advertising more efficient.

This is the silent heist of the digital age. And if you are a marketer, a business owner, or anyone who spends money on digital advertising, you are already a victim. The only question is whether you know it yet.

In this article, we will pull back the curtain on the AI-powered click fraud epidemic. We will examine how fraudsters are using generative AI, machine learning, and massive bot networks to steal billions from advertisers. We will look at the real-world schemes that have infected tens of millions of devices. We will explore why traditional fraud detection is failing. And most importantly, we will give you a practical roadmap for protecting your ad budgets in an era where the line between human and machine has become dangerously blurred.

The Scale of the Problem — How Big Is Click Fraud Really?

Let us start with the numbers that should make every marketer sit up and pay attention. Because until you understand the scale of this problem, you cannot begin to comprehend the urgency of addressing it.

The Billion-Dollar Drain

The digital advertising industry is in the midst of an unprecedented growth spurt. Global digital ad spend surpassed $750 billion in 2025. But with that growth has come an explosion in fraud. According to Anura’s June 2026 executive brief, fraud rates held consistently between 25 percent and 28 percent throughout 2025. By June 2026, global invalid traffic rates had surged to 40 percent. That means for every dollar you spend on digital advertising, as much as 40 cents could be going to fraudulent traffic.

The numbers vary depending on who you ask and how they measure, but the trend is unmistakable:

Spider Labs estimates global ad fraud damage at $32.6 billion in 2025, with fraud rates in AI-optimized campaigns representing up to a 2x higher risk compared to average campaigns.
Lunio’s 2026 Global Invalid Traffic Report found that 8.51 percent of all paid traffic is invalid, meaning nearly one in every twelve clicks does not come from a real user with genuine purchase intent. That translates to $63 billion in wasted global ad spend.
Juniper Research projects that global ad fraud losses will exceed $100 billion in 2026 and reach $172 billion by 2028.
TrafficGuard estimates that 20 percent of digital traffic is non-human, and 22 percent of global digital ad spend is lost to ad fraud. Their analysis suggests that for every $3 spent on marketing, $1 is lost to fraud.
Statista projects that ad fraud losses could reach $100 billion globally in 2026, almost five times the estimated cost in 2018.

These are not rounding errors. These are catastrophic losses that directly impact your return on ad spend, your customer acquisition costs, and ultimately, your bottom line.

The Industry-by-Industry Breakdown

Click fraud does not affect all industries equally. Some sectors are absolute feeding grounds for fraudsters:

Finance and Fintech: A staggering 43 percent of digital ad budgets in the fintech industry are lost to fraudulent clicks. Finance and legal verticals consistently exceed 42 percent invalid traffic rates.
Sports Betting: The sports betting industry loses an estimated $10.5 billion to click fraud annually, with an average invalid traffic rate of 17 percent.
E-commerce: Mid-sized e-commerce online stores lose an average of 21 percent of their monthly budget to invalid traffic. During the holiday season, 57 percent of e-commerce traffic is bots.
Travel: A shocking 80 percent of traffic for travel advertisers is made up of bots.
Affiliate Marketing: An estimated 25 percent of leads generated through affiliate marketing campaigns can be fake. Affiliate fraudulent activity cost digital advertisers an estimated $3.4 billion. Cookie stuffing schemes affect around 5 percent to 10 percent of affiliate marketing transactions.
Gaming: Invalid traffic rates in the gaming industry reached 18.49 percent, according to Lunio’s 2026 report.

For mid-sized advertisers, this represents what industry experts have called a “fraud tax” of 14 percent to 22 percent on every dollar spent.

The Geographic Dimension

Click fraud is a global problem, but certain regions are hotspots. Pixalate’s Q1 2026 benchmarks showed 20 percent invalid traffic on web, 39 percent on mobile apps, and 25 percent on connected TV. The US had a 24 percent IVT rate on mobile app traffic in Q2 2025. In EMEA countries, desktop fraud rates reached 30 percent in Q2 2025.

The Genisys scheme, which we will examine in detail later, infected devices globally, with APAC accounting for approximately 33 percent of activity, covering India, the Philippines, Indonesia, South Korea, Malaysia, Japan, Thailand, Australia, Vietnam, and Singapore.

What these numbers tell us is simple: no advertiser, no industry, and no geography is safe. If you are running digital ads, you are being targeted.

The AI Revolution in Fraud — How Generative AI Supercharged Click Fraud

The fraud landscape has changed fundamentally. The era of the obvious bot that clicked a link every 60 seconds from a single IP address is over. Today’s fraud is powered by a sophisticated triad of technologies: Generative AI, Residential Proxies, and Hybrid Click Farms.

How AI Changed the Game

Before AI, click fraud was relatively easy to detect. Bots followed predictable patterns. They clicked at regular intervals. They came from data center IP addresses that could be blacklisted. They did not scroll, hover, or behave like real humans.

AI changed everything.

Fraudsters now use Large Language Models (LLMs) and behavioral AI to create what the industry calls “human-mimicry bots”. These bots do not just click. They simulate natural mouse movements. They scroll at variable speeds. They spend realistic amounts of time on landing pages. They navigate through websites the way a real person would. They can even solve CAPTCHAs and bypass standard analytics.

Matt Sutton, Chief Commercial Officer at TrafficGuard, put it bluntly: “Fraudsters have become highly skilled at using AI to create bots that closely mimic human behaviour, making them difficult to detect. These bots are able to generate large volumes of ad impressions and clicks that mirror and mimic human interactions in real time. This creates the illusion of high volume, genuine traffic performing in human ways, deceiving advertisers into believing their campaigns are performing well.”

Sutton adds that the sophistication of ad fraud has reached unprecedented levels, with fraudsters leveraging advanced AI tools to exploit vulnerabilities. “Everyone is aware of the phenomena of click farms, this can now all be automated and orchestrated using generative AI that evolves where, when and how automated scripted traffic is being generated and how it behaves.”

The Three Pillars of Modern Click Fraud

Let us break down how modern click fraud actually works.

Pillar One: The AI-Enhanced Botnet

These are not your grandfather’s bots. AI-powered bots can now:

Navigate websites and fill out forms
Stay on-site for realistic durations
Add items to shopping carts before clicking ads
Generate fake engagement at scale
Adapt in real-time to evade detection

“In 2025, we’re seeing a rise in AI-generated bots that adapt in real-time, pass CAPTCHAs, spoof device IDs, and mimic human browsing behaviour down to the millisecond,” according to TrafficGuard’s analysis. “These bots don’t just click.”

Pillar Two: The Residential Proxy Shield

Traditional blacklists are often bypassed through the use of Residential Proxy Networks. By routing fraudulent traffic through the IP addresses of legitimate home internet users, attackers appear to be coming from a local neighborhood rather than a suspicious data center. This makes geographic targeting, a staple of local advertising, highly vulnerable to spoofing.

Pillar Three: Sophisticated Click Farms

While automation is king, the human touch still exists in click farms. These are physical operations, often in regions with low labor costs, where workers use rows of smartphones to manually engage with ads. In 2026, these farms have become hybrid, using software to automate the volume while humans step in to perform complex Cost-Per-Install actions or fill out lead forms with randomized but plausible data.

The Sophisticated Invalid Traffic (SIVT) Problem

The industry divides invalid traffic into two categories. General Invalid Traffic (GIVT) covers declared bots, datacenter IPs, and non-human signatures you can identify from lists. This is the easy stuff. Most pre-bid filters catch GIVT.

Sophisticated Invalid Traffic (SIVT) is the real problem. SIVT covers residential proxies, human-mimicking automation, hijacked devices, and in 2026, autonomous AI agents. SIVT is what passes through pre-bid filters and infects your conversion training data.

Unlike GIVT, SIVT schemes are intentionally designed to evade detection. Fraudsters use SIVT to disguise themselves as human visitors, making tactics more damaging for advertisers. For example, SIVT can be caused by scrapers that conceal their nature by hiding behind seemingly benign user credentials.

The result is a flood of impressions that are technically viewable yet deliver zero actual human attention.

Anatomy of an AI Fraud Scheme — The Genisys Operation

To understand just how sophisticated AI-powered click fraud has become, we need to look at real-world operations. The Genisys scheme, uncovered by Google and Integral Ad Science (IAS), is one of the most revealing examples.

The Scale of the Operation

Genisys was not a simple bot network. It was what IAS senior manager of engineering threat intelligence Hadi Shiravi called “a coordinated ecosystem designed to simulate legitimate ad supply at massive scale”.

The scheme operated through nearly 500 AI-generated publisher sites and 115 Android applications that appeared harmless—utility tools, PDF readers, and casual games. These apps were downloaded millions of times. Some individual apps had as many as five million installations in APAC alone.

At its peak, Genisys had infected more than 25 million Android devices globally. APAC accounted for about 33 percent of the activity.

How It Worked

The Genisys operation had multiple layers:

Synthetic Publisher Sites: Fraudsters used generative AI to mass-produce blog and news-style websites that were never built for real audiences. These sites existed solely to receive and legitimize fake traffic.
Compromised Apps: The malicious process ran in the background of infected devices without users’ knowledge, diverting processing power and network resources to generate traffic.
Bundle ID Spoofing: The scheme massively spoofed app bundle IDs, disguising bot traffic as inventory from legitimate, widely used applications.
Traffic Laundering: The fake impressions were effectively “washed” through the programmatic ecosystem.

Why It Was Different

What set Genisys apart was the use of generative AI to fabricate domains from scratch. In an AI-driven environment, IAS noted that domain names, bundle IDs, and install counts can be faked or synthetically scaled, but behavioral signals are harder to manipulate convincingly.

“This is not just a regular bot network,” Shiravi said. “It’s a coordinated ecosystem designed to simulate legitimate ad supply at massive scale, from synthetic publisher environments to sophisticated traffic misattribution tactics.”

Google’s internal systems flagged anomalies based on suspicious patterns in user agents, IP addresses, and engagement data. Detection involved behavioral analysis of speed, repetition patterns, and cross-network inconsistencies.

Since the takedown, bid request volume associated with Genisys dropped by more than 95 percent. Google Play Protect now automatically disables affiliated apps, including versions installed from outside the official store.

The IAS Threat Lab also identified developer accounts that repeatedly violated rules, including one account that had published 13 malicious apps. Even after apps were removed, new low-quality utilities kept emerging from the same profiles, demonstrating the resilience of the operation.

The SlopAds Epidemic — When 38 Million Devices Become Fraud Nodes

If Genisys was sophisticated, SlopAds was an order of magnitude larger and more brazen.

The Discovery

In September 2025, HUMAN’s Satori Threat Intelligence and Research Team uncovered and disrupted a sophisticated ad fraud and click fraud operation dubbed SlopAds. The threat actors behind SlopAds operated a collection of 224 apps and growing, collectively downloaded from Google Play more than 38 million times across 228 countries and territories.

The operation was named “SlopAds” because the apps associated with the threat had the veneer of being mass-produced, a la “AI slop,” and as a reference to a collection of AI-themed applications and services hosted on the threat actors’ command-and-control server.

The Technical Sophistication

SlopAds stood out for its novel use of attribution and measurement tools as an obfuscation tactic. Only downloads that could be traced to a threat actor-run ad campaign attempted downstream ad and click fraud attacks. Downloads that were unconnected to the ad campaign did not attempt fraud. This made the scheme incredibly difficult to detect because the apps appeared to function normally for most users.

The apps delivered their fraud payload using steganography and created hidden WebViews to navigate to threat actor-owned cashout sites, generating fraudulent ad impressions and clicks. Steganography is the practice of hiding malicious code within innocent-looking files—in this case, within the apps themselves.

At its peak, SlopAds accounted for 2.3 billion bid requests per day. Traffic came from all over the world but was heaviest in the United States (30 percent), India (10 percent), and Brazil (7 percent).

The fraudsters operated more than 200 AI-themed apps, including productivity tools and text-to-image creators, to capitalize on consumer interest and generate downloads. These were cheap app interfaces designed to look legitimate on Google Play—much like mass-produced “AI slop”—but technically performed as advertised.

The Deception Layer

Only users who discovered the app via a threat actor-run ad campaign, rather than finding it organically, would trigger a fraud payload in the background. These apps quietly launched a malicious module and an invisible browser to load “cashout” sites filled with ads. By simulating human behaviors like scrolling, clicking, and viewing, the malware generated billions of fraudulent ad impressions with users none the wiser.

The goal was money, pure and simple. Fraudsters directed the ghost click farms to games and news websites they controlled, with every fake impression resulting in micro-payments.

The Aftermath

Google removed all identified apps from Google Play. Google Play Protect now warns users and blocks apps known to exhibit SlopAds-associated behavior at install time on certified Android devices, even when apps come from sources outside of Play. All users who have these identified apps installed on their device receive a warning and are prompted to uninstall them.

But the threat actors behind SlopAds continued to adapt and attempt to stage new apps to Google’s Play Store even during the assembly of the report. The extensiveness of the network of command-and-control servers and traffic-driving domains suggested the threat actors had plans for expanding the operation.

SlopAds represents the new normal. These are not amateur operations. These are well-funded, technically sophisticated criminal enterprises that are constantly evolving.

The Ghost Click Farm Phenomenon

The term “click farm” used to conjure images of warehouses filled with racks of smartphones, each one staffed by a low-wage worker manually tapping on screens. That model still exists, but it has been superseded by something far more insidious: the ghost click farm.

What Are Ghost Click Farms?

Unlike traditional click farms, where rows of smartphones imitate customers, ghost click farms leverage personal devices to invisibly and remotely carry out the fraud, unlocking newfound scale and obfuscating detection. Instead of operating their own hardware, fraudsters compromise your devices and use them to generate fraudulent ad traffic without your knowledge.

“Hundreds of fake apps and millions of downloads contributing to billions in ad fraud—welcome to the brave new world of ghost click farms,” wrote Mike Schrobo, CEO and Founder of Fraud Blocker.

The SlopAds operation is a perfect example. It secretly co-opted user smartphones into a massive distributed network. The users had no idea their devices were being used to commit ad fraud. They just thought they had downloaded a harmless utility app.

The Scale of the Threat

Akamai reported a dramatic 300 percent year-over-year surge in AI-driven bot traffic targeting websites across industries. According to Akamai, AI bots now account for nearly 1 percent of total traffic across its global platform, generating billions of automated requests that disrupt normal site operations.

DoubleVerify’s Fraud Lab revealed that AI bots are responsible for up to 15 percent of all clicks on unprotected media. The Imperva Bad Bot Report 2025 found that 51 percent of web activity is automated, with 37 percent generated by malicious bots.

The implication is staggering. More than half of all web traffic is now automated. Over a third of it is malicious. And a significant portion of that malicious traffic is directed at digital advertising.

The Impact on Advertisers

Ghost click farms drain ad budgets in ways that are almost impossible for the average advertiser to detect. The traffic looks legitimate. It comes from real IP addresses. It exhibits human-like behavior. The clicks register in your analytics. The impressions count toward your viewability metrics.

But none of it is real. None of it will convert. None of it will become a customer.

“AI-powered ad fraud exacerbates an already expensive problem,” Schrobo notes. “Global ad fraud drains more than 20% of digital marketing spend and scams like this exacerbate an expensive problem.”

Why Traditional Fraud Detection Is Failing

If AI is making fraud more sophisticated, the natural question is: why can’t AI just as easily detect it? The answer is complicated, and it reveals a fundamental asymmetry in the fraud ecosystem.

The Cat-and-Mouse Game

“Detecting ad fraud is a constant game of cat-and-mouse,” says Matt Sutton of TrafficGuard. “As bad actors increasingly leverage AI to evade detection, in turn our machine learning and generative AI models continue to evolve to identify new anomalies in advertising traffic that become new, known ad fraud techniques.”

The problem is that fraudsters have a structural advantage. They can iterate faster. They can test their techniques against detection systems. They can deploy new variants in minutes. Detection vendors, by contrast, must validate their models, avoid false positives, and protect legitimate traffic.

The Failure of Blacklists and Rules

Most outdated click fraud protection methods rely on blacklists and rule-based detection. The problem is that fraudsters adapt.

Blacklists: Once an IP or device is flagged, fraudsters simply switch to a new one. This cat-and-mouse game makes blacklists ineffective at stopping persistent, well-funded fraud operations.
Static rules: Fraud detection based on set rules is rigid. Fraud tactics evolve, but static rules don’t. By the time a new fraud type is detected and a rule is created, ad budgets have already been drained.

The Verification Vendor Problem

Perhaps the most troubling development in the fraud detection space is the growing evidence that even major verification vendors are failing to catch sophisticated fraud.

In March 2025, a 240-page Adalytics report covered by the Wall Street Journal found that Integral Ad Science labeled known URLScan.io bot traffic as human 77 percent of the time across a 2019 to 2024 dataset. DoubleVerify missed the same bots 21 percent of the time. DoubleVerify’s stock dropped roughly 70 percent. A securities class action followed in July 2025.

“The foundational assumption of the IVT detection category collapsed,” the report concluded. “MRC accreditation does not guarantee bots get blocked.”

Independent cybersecurity and ad fraud researcher Dr. Augustine Fou goes even further, arguing that ad fraudsters are not using AI, or AI agents that can buy stuff for you, to enhance SIVT fraud schemes, because it is entirely unnecessary. “Simple bots are already enough to bypass the fraud detection of the largest two legacy verification vendors,” Fou claims.

Fou adds that ad fraudsters were using simple Python code to remix plagiarized content for their WordPress-templated sites more than 15 years ago. “We just didn’t call it AI back then. That’s how they could create 100s of thousands of fake websites so easily.”

Whether or not fraudsters are using “true” AI, the effect is the same. The detection systems that advertisers rely on are not keeping up.

How Click Fraud Destroys Your Campaign Performance

The immediate cost of click fraud is obvious. You pay for clicks that never convert. But the damage goes much deeper than the direct cost of fraudulent clicks.

Algorithmic Misalignment

This is perhaps the most insidious effect of click fraud, and the one that most advertisers fail to understand.

Modern ad platforms like Google and Meta rely heavily on machine learning to optimize bidding. When fraudulent conversion data trains these algorithms, the platform actively purchases more fraudulent traffic because the fraud generates signals that look like success.

“If fraud enters the system as a learning signal, advertisers are not only losing budget, but they are teaching campaign algorithms to optimize toward the wrong outcomes as well,” said Satoko Ohtsuki, CEO of Spider Labs.

This is what the industry calls “inverse optimization” —where AI shifts budget toward fraudulent placements because it misinterprets invalid traffic as high-performing.

One audited demand-side platform (DSP) reached a 90 percent fraud rate. Ninety percent. That means for every ten dollars spent through that DSP, nine went to fraud. The algorithm was actively optimized to purchase fraudulent traffic because the fraud generated signals that looked like success.

Inflated Acquisition Costs

Advertisers face artificially high media costs due to bot competition in ad auctions. This lowers actual return on ad spend (ROAS) and drives up Customer Acquisition Costs (CAC), as marketing capital is diverted away from genuine prospects into fake leads and ghost impressions.

When bots compete in ad auctions, they drive up prices. You end up paying more for every click, whether it is real or fake. And because the fraudsters can generate unlimited fake engagement, they can bid on your keywords indefinitely, exhausting your daily budget before a single genuine customer sees your ad.

Competitor Sabotage

In high-stakes niches like law, insurance, and SaaS, competitors may hire Click Fraud-as-a-Service providers. These operations target specific keywords early in the morning, exhausting your daily budget by 9:00 AM and removing your ads from the auction for the rest of the day.

This is not theoretical. It is happening right now. Your competitors could be systematically draining your budget while you sleep.

Polluted Analytics

When automated bots inflate impressions, clicks, and session counts, the performance metrics used to evaluate campaign health become unreliable. You cannot optimize what you cannot measure accurately. Every decision you make based on polluted data is a decision that moves you further from your actual goals.

The Programmatic Advertising Vulnerability

Programmatic advertising; the automated buying and selling of online advertising—has been a transformative force in digital marketing. It has also become a primary vector for click fraud.

The Numbers Are Alarming

mFilterIt’s Ad Fraud Intelligence Report 2025 revealed that programmatic campaigns saw between 30 and 45 percent of supposedly valid traffic fail deeper checks. That means nearly half of what was being counted as “valid” traffic in programmatic campaigns was actually fraudulent.

A deeper audit revealed that 30 to 45 percent of programmatic traffic labeled “valid” failed secondary validation checks. 43 percent of invalid traffic originated from affiliate networks.

Walled Gardens (the closed ecosystems of major platforms like Google and Meta), long considered the industry’s gated sanctuaries, showed 9 to 18 percent of activity with signs of behavioral manipulation. This figure becomes even more damaging because these environments run on premium CPMs and CPCs.

The MFA Explosion

Made-for-Advertising (MFA) sites—low-quality websites built primarily to generate ad revenue—have exploded in number. Spider AF detected placements on MFA sites increased by a staggering 14 times from the previous year. The MFA site count reached 1,409 percent year-over-year growth.

MFA sites now represent roughly 21 percent of programmatic impressions, many buried inside Performance Max where the buyer cannot inspect placement. These sites stack or stuff multiple placements, pass viewability benchmarks with ease, and still offer no meaningful exposure.

The Viewability Myth

The report dismantles the myth that viewable impressions are genuinely seen by humans. AI-driven bots, operating across multiple channels, now mimic real browsing behavior so convincingly that they complete scroll gestures, replicate dwell times, and interact with content at human-like intervals.

The result is a flood of impressions that are technically viewable yet deliver zero actual human attention. Across audits, mFilterIt found numerous cases where ads achieved perfect viewability scores while human engagement was non-existent.

“Reliance on outdated metrics like viewability and clicks distorts optimization efforts,” the mFilterIt report concluded.

What the Platforms Are (and Aren’t) Doing

The major ad platforms are not ignoring the problem. But their solutions are incomplete, and in some cases, they may be part of the problem.

Google’s Response

Google has been quietly deploying its Gemini AI, a sophisticated system of multimodal large language models, across web and mobile apps to detect invalid traffic. In August 2025, Google announced that thanks to new AI defenses, it had slashed invalid ad traffic from deceptive practices by 40 percent.

Google says it is expanding its use of AI to fight fake traffic on its ads, which should lead to better actual results and data. Advances in large language models are changing the way Google detects and fights against invalid traffic, with the Ad Traffic Quality team working with Google Research and Google DeepMind.

But Google’s solution is not a complete solution. It doesn’t stop repeat clicks and lacks cross-platform scope. TrafficGuard’s 2026 analysis estimates up to 25 percent of Performance Max budget is wasted on click fraud.

Meta’s Response

Meta uses automated AI systems to detect and remove obvious invalid activity: bot farms spamming clicks, accounts flagged as spam, or users repeatedly violating policy. But these systems primarily catch the easy stuff.

The harder problem is sophisticated AI bots that mimic humans, corrupting ad performance and conversion data. These are precisely the types of activity that slip through Meta’s filters.

A bombshell Reuters investigation found that Meta was using AI to generate and test thousands of deceptive ad variations and steer them toward users who previously clicked on scam ads. The investigation found that Meta made a whopping $7 billion a year from scam ads alone. Generative AI has only made scamming easier for cyber criminals, who can now generate convincingly real images and videos in an attempt to legitimize their fake shops.

Santa Clara County sued Meta in California, alleging that the tech giant profited from Facebook and Instagram scam ads. The lawsuit also alleged that the company’s artificial intelligence tools allow it to target vulnerable consumers.

The Platform Incentive Problem

Here is the uncomfortable truth that nobody in the industry wants to discuss. The platforms have a financial incentive to be imperfect at detecting fraud.

When fraudsters spend money on ads, the platforms collect their fees just as they would from legitimate advertisers. Every fraudulent click that is not detected is revenue for the platform. Every dollar spent on fraud is a dollar that flows through the platform’s ecosystem.

This is not to say that the platforms are complicit in fraud. They are not. But the structural incentives are misaligned. The platforms profit from ad spend regardless of whether that spend is legitimate. And their fraud detection efforts, while real, will always be limited by the fact that over-aggressive detection risks blocking legitimate traffic and reducing their revenue.

The Regulatory Response

Governments and regulators are beginning to take notice of the ad fraud epidemic, but the response has been slow and fragmented.

Federal Trade Commission Actions

In April 2025, the FTC issued a proposed order requiring Workado, LLC to stop advertising the accuracy of its artificial intelligence detection products unless it maintains competent and reliable evidence showing those products are as accurate as claimed.

The FTC also acted to stop “Click Profit,” an online business opportunity that had cost consumers at least $14 million. The FTC alleged that Click Profit and its owners deceived consumers by promising they could make large sums in “passive income” using a proprietary system powered by artificial intelligence. “Click Profit misled consumers by falsely promising them guaranteed passive income using cutting-edge AI technology and exclusive brand partnerships,” the FTC stated.

The FTC also required Cox Media Group, MindSift, and 1010 Digital Works to pay a total of $930,000 to settle allegations they deceived customers by falsely claiming to offer an AI-powered service that could target localized ads based on conversations captured from consumers’ smart devices.

Congressional Action

The Safeguarding Consumers from Advertising Misconduct Act (119 HR 7548 IH) recognizes that “online platforms have become a primary conduit for online scams or other digital advertising-related fraud, including fake giveaways, animal sales, deal advertisements tied to nonexistent products, government impersonations, health scams, and impersonations using AI-cloned voices and stolen images targeting legitimate businesses”.

What’s Still Missing

While these actions are positive steps, they are piecemeal. There is no comprehensive regulatory framework for addressing digital ad fraud. There are no mandatory disclosure requirements for fraud rates. There are no industry-wide standards for fraud detection. And enforcement resources are limited.

The regulatory response is years behind the problem. By the time regulators catch up to today’s fraud techniques, the fraudsters will have moved on to something new.

How to Protect Your Ad Budget from AI-Powered Click Fraud

The situation is dire, but it is not hopeless. There are concrete steps you can take to protect your ad budgets from AI-powered click fraud.

Step 1: Look Beyond Vanity Metrics

To protect budgets from AI-driven fraud, look beyond basic metrics like clicks and web traffic, which bots can easily fake. Instead, focus optimization on real business outcomes, such as bottom-of-funnel sales and verified revenues.

Clicks can be faked. Impressions can be faked. Viewability can be faked. What cannot be faked is a paying customer. Track your campaigns to the point of actual conversion—not just a form fill, not just a lead, but a verifiable sale.

Step 2: Run a Traffic Quality Audit

To maintain baseline data integrity and prevent budget waste, marketing teams should run a traffic quality audit and invest in third-party, independent ad tech that verifies traffic in real time.

An audit will reveal the true extent of the fraud in your campaigns. It will show you which channels, which placements, and which keywords are being targeted by fraudsters. And it will give you the data you need to make informed decisions about where to spend your money.

Step 3: Use Independent Verification

Ensure your chosen solution meets industry standards, such as a TAG Certified Against Fraud seal or MRC Accreditation. But remember that MRC accreditation does not guarantee bots get blocked. You need to go beyond the seal and actually test the effectiveness of your verification tools.

Independent verification is crucial because the platforms’ own fraud detection is incomplete. A third-party solution that operates independently of the ad platforms can catch what the platforms miss.

Step 4: Harden Your Campaign Settings

The first line of defense is often found in the settings we overlook.

Exclude Interest-Based Locations: In Google Ads, ensure your location settings are set to “Presence: People in or regularly in your targeted locations” rather than the default “Presence or Interest.” This prevents users in click-farm-heavy regions from seeing your ads just because they searched for your city.
Audit Search Partners: Many fraudulent placements occur on search partner networks. Review your search partner performance regularly and consider excluding them if they are generating suspicious traffic.
Monitor Your IP Exclusions: Maintain a dynamic list of suspicious IP addresses and update it regularly.

Step 5: Invest in AI-Powered Click Fraud Protection

Just as fraudsters are using AI to create more convincing and scalable fraud schemes, verification companies are leveraging AI to identify the constantly evolving landscape of bots.

Machine learning is no longer a buzzword; it’s a critical tool in the fight against fake clicks, invalid traffic, and wasted ad spend. Machine learning models process vast amounts of ad traffic data in real time, detecting anomalies that indicate click fraud. Instead of reacting after the damage is done, real-time click fraud prevention stops wasted spend before fraudsters get paid.

Unlike rigid rules, machine learning understands the context of every interaction. It evaluates factors like device behavior, click frequency, and user engagement patterns to distinguish between real users and bots.

Key features to look for in a click fraud protection solution:

Real-time monitoring and blocking
Contextual detection that evaluates behavioral patterns
Scalability to handle large volumes of traffic
Coverage across the entire funnel (impression, click, and conversion)
Continuous learning to adapt to new fraud tactics

Step 6: Understand the Different Layers of Protection

Most comparison lists rank IVT tools by feature count. The useful question is different: which layer does this tool actually cover?

Pre-bid IVT: Blocks fraudulent traffic before you bid on it.
Click-time IVT: Blocks fraudulent clicks at the moment they occur.
Conversion-time IVT: Identifies fraudulent conversions that have already happened and prevents them from training your optimization algorithms.

Most buyers solve the first or second problem and think they are done. The damage from the third is the one that compounds inside Smart Bidding and Meta’s Andromeda algorithm. If you are not protecting against conversion-layer IVT, your algorithms are still being trained on fraudulent data.

Step 7: Identify the Warning Signs

Advertisers who know how to read these signals can detect fraudulent activity early and take protective action before significant budget is lost.

Look for:

Unusual click patterns: Sudden, unexplained spikes in clicks, especially during off-peak hours, weekends, or outside your typical business hours. Legitimate demand follows your audience’s schedule. Fraudulent clicks often do not.
Geographic anomalies: A surge in clicks from a region that does not match your target market is a classic indicator of click farm activity.
Low conversion rates despite high click volume: A sudden increase in clicks that does not translate into a proportional rise in conversions is one of the strongest signals of click fraud. Genuine clicks from interested users convert at roughly consistent rates. When your click volume jumps but your conversions stay flat or fall, the new clicks are almost certainly coming from non-converting sources.

Segment your click data by hour, day, and geography. Any pattern that diverges sharply from your historical baseline warrants investigation.

The Economic Impact on Small and Medium Businesses

While enterprise advertisers have the resources to invest in sophisticated fraud protection, small and medium businesses (SMBs) are the most vulnerable victims of click fraud.

The SMB Disadvantage

SMBs typically have smaller marketing budgets, less sophisticated analytics, and fewer resources to dedicate to fraud detection. They are the perfect targets for fraudsters because:

They are less likely to notice the fraud in their data
They are less likely to have third-party verification tools
The fraud is more damaging as a percentage of their total budget

AI-powered click fraud is costing SMEs more than ad spend because they lack the resources to detect and prevent it. For a business spending $10,000 per month on digital advertising, a 14 to 22 percent fraud tax means $1,400 to $2,200 is wasted every month. Over a year, that is $16,800 to $26,400—money that could have been used to acquire real customers.

The Conversion-Layer Blind Spot

For most SMBs, the fraud they experience at the conversion layer goes completely undetected. They see a lead come in, they count it as a conversion, and they optimize their campaigns toward more of the same. But if that lead was generated by a bot or a click farm worker with no purchase intent, it will never become a sale.

The damage compounds. The algorithms learn that these “conversions” are easy to generate, so they bid more aggressively on the traffic sources that produce them. The fraudsters get paid. The SMB gets nothing. And the algorithm keeps spending money on the same fraudulent sources.

The Solution for SMBs

SMBs cannot afford enterprise-grade fraud protection, but they can take practical steps:

Use free or low-cost fraud detection tools. Open-source solutions like AdTruth can help identify which advertising platforms waste budget on fraudulent traffic.
Monitor your conversion rates religiously. If your clicks are going up but your sales are flat, something is wrong.
Focus on bottom-of-funnel metrics. Track actual sales, not just leads or form fills.
Consider third-party verification. Even a basic verification tool can catch a significant portion of the fraud that the platforms miss.

The Future of Click Fraud — What’s Coming Next

The current wave of AI-powered click fraud is not the end of the story. It is just the beginning.

Autonomous AI Agents

Juniper Research projects that ad fraud losses will scale to $133 billion by 2028, driven by AI botnets and autonomous agents. Autonomous AI agents are self-directed AI systems that can make decisions and take actions without human intervention. In the context of ad fraud, these agents could:

Automatically identify vulnerable campaigns
Generate and test different fraud techniques
Adapt their behavior to evade detection in real-time
Operate at scales that are impossible for human-managed operations

The Agentic Audience

HUMAN has tracked a 6,900 percent increase in agentic AI activity over the past year. 64 percent of consumers plan to use AI for holiday shopping, up from just 11 percent in 2024.

As AI agents become more prevalent in consumer behavior, the line between legitimate AI-assisted activity and fraudulent AI activity will blur further. How do you distinguish between a consumer using an AI shopping assistant and a fraudster using an AI bot to generate fake clicks? The distinction will become increasingly difficult.

AI-Generated Content Environments

The threat is not just fake clicks. It is fake environments. AI-generated “MFA” sites have surged by 717 percent, with suspicious domains ballooning from approximately 17,000 in mid-2024 to over 108,000 by May 2025.

These sites are not just hosting ads. They are generating content—blog posts, news articles, product reviews—all created by AI, all designed to look legitimate, and all existing solely to generate ad revenue. When your ad appears on one of these sites, it is being seen by no one. There are no human readers. There is no engagement. There is just an AI-generated content farm designed to collect your ad dollars.

“This challenge represents the next evolution of click fraud,” wrote an op-ed contributor for MediaPost. “While advertisers have long battled basic bot networks generating fake impressions, AI-powered deception takes this threat to an unprecedented level.”

The Detection Arms Race

As fraud becomes more sophisticated, detection will have to evolve. Researchers are already developing advanced techniques:

A Multi-Expert Transformer Model for Click Fraud Detection highlights the potential of adaptive, privacy-aware click fraud detection for real-time advertisement bidding systems.
Hybrid Adaptive Ensemble models combining LightGBM, Recurrent Neural Networks, and AutoEncoder-based anomaly detection offer comprehensive fraud intelligence.
Deep Neural Networks and behavioral analysis systems can detect patterns and anomalies that static rules cannot.

But the arms race will continue. Every advance in detection will be met with an advance in evasion. The fraudsters will keep iterating. The vendors will keep iterating. And advertisers will keep paying the price.

A Call to Action for the Industry

The click fraud epidemic is not just a technical problem. It is an industry-wide crisis that requires industry-wide solutions.

Transparency

Advertisers need more transparency from the platforms about the true extent of fraud in their ecosystems. They need to know what percentage of their spend is going to invalid traffic. They need to know which placements are most vulnerable. And they need the tools to verify this information independently.

The current system is a black box. Advertisers put money in. They get numbers out. But they have no way of knowing whether those numbers are real. This has to change.

Standards

The industry needs better standards for fraud detection and reporting. The MRC accreditation process is a start, but as the Adalytics report showed, it is not enough. Standards need to be more rigorous. They need to be updated more frequently. And they need to be enforced.

Collaboration

Fraudsters are collaborating across borders and across platforms. The industry needs to collaborate too. Information sharing about fraud techniques, compromised devices, and threat actors is essential. No single company can solve this problem alone.

Education

Marketers need to be educated about the threat of click fraud. They need to understand how it works, how to detect it, and how to protect against it. Most advertisers still believe that the platforms are handling this problem for them. They are not.

Conclusion: The Choice Is Yours

The numbers are stark. Up to 40 percent of digital ad traffic is fraudulent. $165 billion was lost to ad fraud in 2025. Losses are projected to reach $172 billion by 2028. And the problem is getting worse, not better.

The fraudsters are using the most advanced technology available. They are well-funded, technically sophisticated, and constantly evolving. They are operating at scales that were unimaginable just a few years ago. And they are targeting you.

You have a choice.

You can continue to rely on the platforms’ fraud detection, hoping that they will catch the fraud before it drains your budget. You can continue to optimize based on vanity metrics that bots can easily fake. You can continue to be a victim of the silent heist.

Or you can fight back.

You can invest in independent verification. You can look beyond clicks and impressions to real business outcomes. You can audit your traffic. You can harden your campaign settings. You can educate yourself and your team about the threat.

The fraudsters are counting on your complacency. They are counting on you not noticing. They are counting on you assuming that the platforms have this handled. They are counting on you being too busy with other priorities to look closely at your data.

Do not give them what they want.

The AI-powered click fraud epidemic is the defining challenge of digital advertising in the 2020s. It is draining billions from the economy. It is distorting the data that marketers rely on. It is undermining the very foundation of performance-based advertising.

But it is not inevitable. With awareness, with vigilance, and with the right tools, you can protect your budgets and ensure that your ad dollars actually reach real humans.

The technology that is enabling this fraud—artificial intelligence—is the same technology that can stop it. The question is whether you will use it.

The choice is yours. Choose wisely.

This article was researched and written using data from industry reports, academic research, and verified sources including Anura, Spider Labs, TrafficGuard, HUMAN Security, mFilterIt, Pixalate, Juniper Research, the Federal Trade Commission, and other authoritative sources. All statistics and data points are cited with their original sources for verification.

Complete References List

Anura

1. Anura Executive Brief – “AI-Powered Ad Fraud & Marketing Data”
https://www.anura.io/executive-brief

2. Anura – “New Executive Brief Warns CMOs: Marketing Budgets Are Being Stolen” (via TMCnet)
https://www.tmcnet.com
(Coverage of Anura’s June 2026 executive brief findings)

3. Anura – “New Executive Brief Warns CMOs: Marketing Budgets Are Being Stolen” (via Metro Atlanta CEO)
https://metroatlantaceo.com

4. Anura – Click Fraud 101 Resource
https://www.anura.io

Spider Labs

5. Spider Labs – “2026 Ad Fraud White Paper Report” (Annual Edition)
https://spideraf.com/adfraud-report-whitepaper-2026

6. Spider Labs – Full 2026 Ad Fraud White Paper Report (Direct Download Page)
https://spideraf.com/2026-annual-ad-fraud-white-paper

7. Spider Labs – Press Release via Wedbush Investor
https://investor.wedbush.com/wedbush/article/accwirecq-2026-4-2-as-ai-ad-buying-expands-global-fraud-losses-hit-326-billion

8. Spider Labs – “A Complete Guide to Click Fraud & How to Prevent It (2026)”
https://spideraf.com

Lunio

9. Lunio – “2026 Global Invalid Traffic Report” (Coverage via VARIndia)
https://varindia.com/news/63bn-lost-to-ad-fraud-invalid-traffic-surges

10. Lunio – “2026 Global Invalid Traffic Report” (Coverage via MediaPost)
https://www.mediapost.com

11. Lunio – Click Fraud Calculator
https://www.lunio.ai/click-fraud-calculator

12. Lunio – “2026 Global Invalid Traffic Report” (Direct Report Page)
https://www.lunio.ai/2026-global-invalid-traffic-report

Juniper Research

13. Juniper Research – Ad Fraud Forecast (via CO Consulting)
https://christopholivierconsulting.com

14. Juniper Research – Ad Fraud Forecast (via BuiltIn)
https://builtin.com

15. Juniper Research – “The Global Cost of Ad Fraud: 2023-2028”
https://doubleverify.com

Statista

17. Statista – Ad Fraud Projections (via ClickPatrol)
https://clickpatrol.com

18. Statista – Ad Fraud Projections (via ClickGuardian)
https://clickguardian.ai

19. Statista – Ad Fraud Projections (via Global Risk Community)
https://globalriskcommunity.com

Pixalate

20. Pixalate – “Q1 2026 Invalid Traffic (IVT) & Ad Fraud Benchmark Reports”
https://www.pixalate.com/blog/q1-2026-ad-fraud-benchmarks-report-for-emea

21. Pixalate – “Global IVT Benchmarks Report”
https://www.pixalate.com/invalid-traffic-benchmarks-report-global

22. Pixalate – Q1 2026 IVT Benchmarks (Coverage via LinkedIn/ClickSambo)
https://www.linkedin.com

23. Pixalate – “Q1 2026 Web Seller Trust Index 2.0”
https://www.globenewswire.com

mFilterIt

24. mFilterIt – “Beyond the Linear Lens: Ad Fraud in 2025” (via Times of India)
https://timesofindia.indiatimes.com

25. mFilterIt – “Beyond the Linear Lens: Ad Fraud in 2025” (via Business Standard)
https://www.business-standard.com

26. mFilterIt – “Beyond the Linear Lens: Ad Fraud in 2025” (via India Television)
https://indiantelevision.com

27. mFilterIt – “AI-driven Ad Fraud Causes 12% Marketing Spend Leakage” (via News18)
https://one.news18.com

28. mFilterIt – “AI-driven Ad Fraud Causes 12% Marketing Spend Leakage” (via Storyboard18)
https://www.storyboard18.com

29. mFilterIt – FICCI Media & Entertainment Report 2026
https://www.mfilterit.com

HUMAN Security / Satori

30. HUMAN Security – “Satori Threat Intelligence Alert: SlopAds”
https://www.humansecurity.com

31. HUMAN Security – “HUMAN Disrupts Sophisticated Mobile App Fraud Scheme SlopAds”
https://www.humansecurity.com

32. HUMAN Security – “SlopAds’ Highly Obfuscated Android Malware Scheme”
https://www.humansecurity.com

33. HUMAN Security – “SlopAds Fraud Ring Exploits 224 Android Apps” (via The Hacker News)
https://thehackernews.com

34. HUMAN Security – “Agentic Visibility: How to See AI Agents in Your Traffic”
https://www.humansecurity.com

35. HUMAN Security – “State of Agentic Traffic” Reports
https://www.humansecurity.com

36. HUMAN Security – LinkedIn Post on 6,900% Agentic AI Growth
https://www.linkedin.com

TrafficGuard

37. TrafficGuard – “Mid-Audit Check-In: What Is Really Hiding in Your Search Data”
https://www.trafficguard.ai

38. TrafficGuard – “Why Residential Proxies Are the New Blind Spot in Invalid Traffic”
https://www.trafficguard.ai

39. TrafficGuard – “Navigating The Sophisticated Landscape Of Click Fraud In 2026” (via Martech Zone)
https://martech.zone

40. TrafficGuard – “Ghost Conversions in Fintech”
https://www.trafficguard.ai

41. TrafficGuard – “Performance Max Ad Fraud and Invalid Traffic Protection”
https://www.trafficguard.ai

42. TrafficGuard – “Combat Click Fraud in PMax for Smarter ROI”
https://www.trafficguard.ai

43. TrafficGuard – Click Fraud & Ad Fraud Resources
https://www.trafficguard.ai

Akamai

44. Akamai – “AI Pulse: How AI Bots and Agents Will Shape 2026”
https://www.akamai.com

45. Akamai – “Akamai Reports 300% Surge in AI Bot Traffic” (via Security Info Watch)
https://www.securityinfowatch.com

46. Akamai – “Akamai Research: AI Bots Threaten the Foundation of Web-Based Business Models”
https://www.ir.akamai.com

DoubleVerify

47. DoubleVerify – “Swat That Bot? New DV Tools Keep the Wrong Kind of AI Away From Your Ad Spend”
https://doubleverify.com

48. DoubleVerify – “Report: AI bots behind 15% of all click metrics” (via Mi3)
https://www.mi-3.com.au

49. DoubleVerify – “What the ‘Bot Fingers’ Era Means for Advertising”
https://advertisingweek.com

50. DoubleVerify – “DoubleVerify’s Response to Adalytics’ March 28 GIVT Report”
https://doubleverify.com

51. DoubleVerify – “Statement on Adalytics’ GIVT Report”
https://doubleverify.com

Imperva (Thales)

52. Imperva – “2025 Bad Bot Report”
https://www.imperva.com

53. Imperva – “AI-Driven Bots Surpass Human Traffic – Bad Bot Report 2025”
https://cpl.thalesgroup.com

54. Imperva – Bad Bot Report Coverage (via ABES)
https://abes.org.br

Adalytics

55. Adalytics – Report on IAS and DoubleVerify (via MediaCat UK)
https://mediacat.uk

56. Adalytics – Report Coverage (via B&T)
https://www.bandt.com.au

57. Adalytics – Report Coverage (via MediaPost)
https://www.mediapost.com

IAB (Interactive Advertising Bureau)

58. IAB – $84 Billion Ad Fraud Estimate (via Tracio)
https://tracio.ai

59. IAB – Ad Fraud Estimate (via Martech Series)
https://martechseries.com

Google

60. Google – “Google deploys Gemini AI to combat ad fraud with 40% reduction” (via PPC Land)
https://ppc.land

61. Google – “Google Has Been Quietly Using Gemini AI to Weed Out Ad Fraud” (via Yahoo Tech)
https://tech.yahoo.com

62. Google – “Google Quietly Deploys Gemini AI To Combat Ad Fraud” (via Marketing Mind)
https://marketingmind.in

63. Google – “Google’s Use of Large Language Models to Detect Invalid Ad Traffic” (via Seoteric)
https://www.seoteric.com

Reuters

64. Reuters – “Meta Earned $7 Billion Annually From Scam Ads” (via News18)
https://www.news18.com

65. Reuters – “Meta Earned $7 Billion A Year From Scam Ads” (via NDTV)
https://www.ndtv.com

66. Reuters – Reuters Investigation Cited in Santa Clara County Lawsuit
https://www.theguardian.com

Santa Clara County

67. Santa Clara County – “County Counsel Files Landmark Civil Prosecution Taking on Meta’s Role in Massive Consumer Fraud”
https://news.santaclaracounty.gov

68. Santa Clara County – Lawsuit Coverage (via Al Jazeera)
https://www.aljazeera.com

69. Santa Clara County – Lawsuit Coverage (via Bloomberg)
https://www.bloomberg.com

Federal Trade Commission (FTC)

70. FTC – “FTC Acts to Stop ‘Click Profit’ Online Business Opportunity”
https://search.ftc.gov

71. FTC – “FTC Sues Click Profit, Alleges Passive Income Amazon AI Scam” (via Entrepreneur)
https://www.entrepreneur.com

72. FTC – “FTC to Require Cox Media Group, Two Other Firms to Pay Nearly $1 Million”
https://www.ftc.gov

73. FTC – “CMG Media Corporation, In the Matter of”
https://search.ftc.gov

74. FTC – “FTC Approves Final Order against Workado, LLC”
https://www.ftc.gov

75. FTC – “FTC Order Requires Workado to Back Up Artificial Intelligence Detection Claims”
https://www.ftc.gov

76. FTC – “FTC Issues Proposed Order Requiring Workado, LLC”
https://www.ftc.gov

U.S. Congress

77. Safeguarding Consumers from Advertising Misconduct Act (SCAM Act) – 119 HR 7548 IH
https://www.govinfo.gov

Fraud Blocker

78. Fraud Blocker – Mike Schrobo Commentary on Ghost Click Farms
https://fraudblocker.com (Referenced in article; site domain noted)

Media Rating Council (MRC)

79. MRC – Invalid Traffic Definition (via Pixalate)
https://www.pixalate.com

Additional Industry Sources

80. Spider Labs – “Uncovering MFA Sites and How to Protect Your Business”
https://spideraf.com

81. Spider Labs – “Spider Labs Releases 2025 Ad Fraud Report”
https://www.einpresswire.com

82. Spider Labs – “Spider AF Releases 2024 Annual Ad Fraud White Paper”
https://global.spideraf.com

83. TrafficGuard – “TrafficGuard launches PMax Solution” (via Financial Express)
https://www.financialexpress.com

84. DoubleVerify – “DV Fraud Verification: Protect Your Brand Against Bad Actors”
https://doubleverify.com

85. HUMAN Security – “HUMAN Security’s 2026 State of AI Traffic & Cyberthreat Benchmark Report”
https://www.humansecurity.com

admin

I manage ClickBaton.com and ROIpad.com our product positioning intelligence platform. I am always open to new partnerships, collaboration and speaking directly with founders who are looking to test our products. Please feel free to connect with me on linkedin.