LinkedIn generated the highest invalid traffic rate of any major ad platform measured in a nine-month study published this month by Lunio, an invalid traffic detection company, with the rate climbing from 13% in Q3 2025 to 17.62% in Q1 2026. The report, based on 64 million clicks tracked across Google, Bing, LinkedIn, and Meta, arrives at a moment when bots have crossed a threshold: automated traffic now accounts for the majority of all web requests, according to Cloudflare data published this week.

The Lunio report, titled The Invalid Traffic Impact Report: IT and Security 2026, draws on a representative sample of client accounts analyzed across three consecutive quarters: Q3 2025 (July 1 to September 30), Q4 2025 (October 1 to December 31), and Q1 2026 (January 1 to March 31). All data was collected from accounts operating in unprotected, monitor-only mode - Lunio tracked invalid traffic without actively blocking it - providing an unfiltered view of what passes through platform-level defences and reaches advertisers.

Invalid traffic (IVT) is any click or site visit that does not originate from a real person with genuine intent to engage. The category covers bots both benign and malicious, accidental clicks, clicks manufactured by competing advertisers to drain rivals' budgets, and sophisticated automated agents that generate plausible-looking browsing sessions. According to Lunio, the IVT that platforms do not catch tends to be the hardest kind to detect - automated activity that closely mimics real human behaviour and passes basic post-click engagement signals.

The timing gives the data unusual context. Cloudflare figures published this week show bots now account for 57.4% of web traffic to HTML content, with human visitors at 42.6%. HUMAN Security, another company working in fraud detection, reported in April 2026 that automation is growing eight times faster than human traffic. Covered by PPC Land at the time, that acceleration complicates every assumption advertisers make about who is actually clicking their ads.

LinkedIn: the worst platform in the study by a large margin

LinkedIn recorded an average IVT rate of 15.34% across the nine-month period - by far the highest of the four platforms analyzed. The rate moved consistently higher each quarter: 13.00% in Q3 2025, 15.40% in Q4 2025, and 17.62% in Q1 2026. That last figure is the single highest platform-level rate recorded across any quarter in the dataset.

The elevated rate on LinkedIn is driven by a specific combination of structural factors identified in the report. Lead Gen Forms - which allow users to submit contact details without leaving the platform - are actively exploited by bots that submit plausible-looking data, flooding advertiser CRMs with junk leads that are structurally indistinguishable from genuine submissions. Audience network expansion settings increase exposure to low-quality off-platform inventory. And LinkedIn's cost per click, which industry benchmarks place in the $10 to $15 range for B2B advertisers, means each invalid click is proportionally more expensive than on any other channel in the study.

A transparency gap compounds the problem. According to the report, LinkedIn lacks the IP-level reporting and automated click quality reports that are available on other platforms. Fraudulent click activity is harder to identify there, and refunds, when sought, are rare and difficult to obtain.

The platform's own transparency reporting has acknowledged bot and fake accounts as a persistent problem. LinkedIn partnered with HUMAN Security in June 2024 to strengthen invalid traffic protections, a move that reflected the scale of the challenge the platform faces. Despite that partnership, the Lunio data suggests the IVT rate on LinkedIn has continued rising quarter on quarter through Q1 2026.

That creates a tension for B2B advertisers. LinkedIn has simultaneously established itself as the dominant B2B advertising platform by spend share, delivering 121% ROAS in 2025 and attracting 41% of total B2B ad budgets. Those returns are real. But the Lunio data indicates a meaningful share of that spend is not reaching real buyers.

Bing: nearly three times Google's rate, slowly rising

Microsoft Ads (Bing) averaged an IVT rate of 11.63% across the analysis period - nearly three times the Google Ads average - with a gradual upward trajectory: 11.20% in Q3 2025, 11.66% in Q4 2025, and 12.04% in Q1 2026.

The structural reasons for Bing's elevated baseline are, according to the report, well-understood. Microsoft's advertising network relies heavily on third-party network partnerships, including the Microsoft Audience Network, from which advertisers cannot fully opt out. This generates a disproportionate share of low-quality traffic. Broad-match and automated bidding strategies, common on Bing because of its smaller search volume, cast a wider net and pull in more irrelevant traffic proportionally. And lower CPCs on Bing reduce advertiser vigilance: placement exclusions and traffic quality monitoring receive less attention than they would on Google, where the same wasted click costs more.

Microsoft Clarity launched Bot Activity tracking in January 2026, giving website operators visibility into which automated systems are accessing their properties - a step that addresses the identification problem at the analytics layer, but does not touch invalid click filtering within the ad platform itself.

Meta: a smaller problem that nearly doubled

Meta averaged a 5.37% IVT rate across the analysis period - higher than Google but well below LinkedIn and Bing. The directional trend is what warrants attention: from 3.57% in Q3 2025, Meta's rate climbed to 5.90% in Q4 2025 and 6.64% in Q1 2026, coming close to doubling across two quarters.

Meta occupies a secondary position in most B2B media mixes, used selectively for retargeting, brand awareness, or top-of-funnel reach rather than direct lead generation. The financial exposure per invalid click is lower than on LinkedIn, given the difference in CPMs. But the volume of impressions that Meta campaigns generate - particularly those using Advantage+ - means the aggregate invalid spend accumulates quickly at scale.

Google: lower overall, but a rising search problem

Google averaged a blended 3.90% IVT rate across all campaign types, the lowest of the four platforms. Inside that overall figure, the data by campaign type tells a more complicated story.

Google Search saw its IVT rate rise from 3.95% in Q3 2025 to 4.78% in Q1 2026 - a 21% increase in two quarters. Search has historically been treated as the most fraud-resistant channel, valued for high commercial intent, well-understood user behaviour, and strong platform-level filtering. A 21% rise in IVT during the period of maximum spend - Q1, when budgets reset and campaigns scale aggressively - is a meaningful shift from that baseline assumption.

Google Display recorded the sharpest upward trajectory of any channel in the entire analysis. The IVT rate rose from 2.94% in Q3 2025 to 3.70% in Q4 2025 and then surged to 6.81% in Q1 2026 - a 132% increase from the Q3 starting point. The report attributes much of this to a seasonal pattern: Display budgets often contract in Q3 and Q4, resulting in narrower, more curated placements. When budgets reset in January and advertisers activate broader audience targeting and automated bidding simultaneously, IVT exposure increases sharply. The Q1 figure likely reflects what full-scale Display campaigns actually face in an unprotected state.

Google Demand Gen averaged 4.27% IVT but displayed uneven movement: 5.43% in Q3 2025, falling to 3.03% in Q4 before rising to 4.34% in Q1 2026. The volatility may reflect shifts in audience targeting, creative rotation, or the mix of Google properties served within Demand Gen inventory across different quarters.

Performance Max recorded the lowest average IVT rate of any Google campaign type at 3.39%, with stable quarter-on-quarter movement - 3.33%, 3.40%, and 3.44% across the three quarters. The report urges caution reading that figure. Performance Max distributes budget across Search, Display, YouTube, Discovery, Gmail, and Maps within a single campaign, so 3.39% is a blended average across all those inventory types. Channels with historically lower IVT rates pull the overall figure down. Placement-level IVT attribution within PMax remains, according to the report, "virtually impossible without independent traffic verification."

PPC Land has tracked Google's gradual expansion of PMax transparency: the Google Ads API v23 released in January 2026 added channel-level reporting, and a subsequent test in early May 2026 began allowing advertisers to individually toggle the Search Partner Network and Display Network on or off. Neither change provides placement-level IVT attribution inside PMax.

Nick Morley, CEO of Lunio, framed the broader problem directly in the report: "The era of AI-powered bots has fundamentally changed the fraud landscape. Invalid traffic is no longer just a tax on ad spend; it is actively degrading the quality of the data our algorithms learn from. Every fake click isn't just wasted budget, it's a corrupted data signal that pulls campaigns further from genuine buyers."

The worst individual placements

The report identifies specific placement domains generating the highest observed IVT rates among those receiving at least 1,000 clicks across the sample. The worst was www.lyngsat.com at 93.8% - more than nine in ten ad clicks from that domain were invalid. The second was podcastim.org.il at 89.3%.

Gaming websites and mobile applications dominate the list. Mobile gaming apps - World Survival at 81.2%, Island Craft at 67.5%, Craft Mini Game at 58.2%, and Craftsman: Go Craft at 52.6% - appear alongside browser-based gaming sites colonist.io at 66.4% and www.bridgebase.com at 64.9%. Gaming placements concentrate IVT for a specific mechanical reason: pages with multiple ad units and gameplay that requires rapid screen interaction produce high rates of accidental clicks, particularly on mobile touchscreens. The report notes that some of these layouts appear deliberately designed to increase the likelihood of those accidental interactions.

Made-for-advertising (MFA) sites also appear in the worst-performing placement list. MFA sites host low-quality, often AI-generated content and are designed primarily to generate ad revenue rather than serve readers. According to the Association of National Advertisers, MFA sites drain an estimated 15% of total programmatic ad spend. Freemagazines.top at 64.5% and en15.apktoget.com at 51.7% both exhibit characteristics the report associates with MFA classification. PPC Land has covered how MFA sites were identified and defined by IAB Australia and how DoubleVerify subsequently launched tiered MFA categories to help advertisers identify and avoid them.

Why platforms don't solve this on their own

A consistent argument running through the Lunio report is that platform-level defences address the straightforward end of the IVT spectrum and leave the sophisticated end largely intact. Basic cases - declared bots, duplicate clicks, simple behavioural patterns - are typically caught. Sophisticated invalid traffic generated by agents deliberately mimicking legitimate human behaviour is not.

The economic incentives are structural. Comprehensive real-time machine-learning-based traffic analysis is computationally expensive. According to the report, processing latency alone can increase costs tenfold, making full-scale implementation economically impractical for platforms whose revenue depends on click volume.

The Lunio report cites findings from the Google antitrust trial, which revealed that Google manipulated its ad auctions to meet revenue targets, at times artificially increasing ad prices by up to 5%. It also notes that Google's own documentation explicitly assigns responsibility for traffic quality monitoring to advertisers rather than retaining it at the platform. A March 2025 investigation covered by PPC Land found that major verification systems routinely fail to detect and block non-human traffic even when bots openly identify themselves - in documented cases, Google's systems served ads to declared bots running from Google's own data center IP addresses.

The cost of mounting sophisticated automated attacks has also collapsed. Large language models have, according to the report, made it possible to deploy an agent that runs a synthetic browsing session, generates a plausible click pattern, and passes basic post-click engagement signals for cents per session. The defence costs roughly the same as it always did. The offence is now orders of magnitude cheaper.

Lunio expanded its own product to cover affiliate fraud in May 2026, citing that 24% of all affiliate traffic is invalid and that $2.8 billion was lost to affiliate click fraud in the United States alone in 2025. The company's detection engine, operating at click level across Google, Bing, Meta, and LinkedIn, works by excluding invalid users from future targeting as soon as they are identified. For borderline interactions, a "suspicious" classification gathers additional data points before a determination is made.

What the data signals distortion means beyond wasted spend

The Lunio report makes a distinction between two categories of damage from invalid traffic: direct budget waste and the corrupted data signal problem. The first is visible on a cost report. The second is not.

When invalid traffic inflates click volumes and skews conversion data, the signals fed back into ad platform bidding algorithms are corrupted. Systems trained on that data misallocate budget, overvalue underperforming keywords and placements, and optimise toward fraudulent interactions rather than genuine buyers. The result is a compounding efficiency loss that extends beyond the immediate cost of the invalid clicks. Spam leads generated from invalid form submissions contaminate CRMs, consume sales team capacity, and distort the performance benchmarks that inform future budget decisions.

A TAG study released in October 2024 and covered by PPC Land found that industry anti-fraud programs saved advertisers an estimated $10.8 billion in US display and video channels in 2023. That figure represents what was caught and prevented. The Lunio data shows what continued to reach advertisers in unprotected accounts even after those programs were in place.

Methodology

The 64 million clicks in the dataset span Google, Bing, LinkedIn, and Meta, with all platforms required to have generated a minimum of 150,000 clicks across the sample period to appear in the channel breakdown. For the Google campaign-type analysis, each of the four campaign types - Search, Performance Max, Demand Gen, and Display - produced at least 3 million clicks across the nine months, with Search alone accounting for more than 26 million. The financial estimates in the report use a representative average cost per click of $8.00 and model losses against a $5 million annual spend figure with a conservative 3:1 return on ad spend benchmark.

Timeline

  • June 2024 - LinkedIn partners with HUMAN Security to bolster invalid traffic protections. Initial results cite less than 1% of impressions on LinkedIn's core platform as invalid. PPC Land coverage
  • June 2024 - IAB Australia publishes guidance on identifying and avoiding Made-for-Advertising sites. PPC Land coverage
  • July 2024 - Cloudflare data shows AI bots accessed approximately 39% of the top one million internet properties in June 2024. PPC Land coverage
  • October 2024 - A TAG and ANA study finds industry anti-fraud programs saved advertisers $10.8 billion in US display and video channels in 2023, equivalent to a 92% reduction in IVT-related losses. PPC Land coverage
  • March 2025 - An Adalytics investigation finds that leading verification systems routinely fail to detect non-human traffic; Google served ads to declared bots from its own data center IP addresses. PPC Land coverage
  • January 2026 - Google releases Ads API v23 with channel-level reporting for Performance Max campaigns. PPC Land coverage
  • January 2026 - Microsoft Clarity launches Bot Activity tracking, giving website operators visibility into which AI systems crawl their properties. PPC Land coverage
  • May 2026 - Lunio launches affiliate fraud detection, citing $2.8 billion in US affiliate click fraud losses in 2025 and a 24% invalid traffic rate across affiliate channels. PPC Land coverage
  • May 2026 - Google begins testing placement controls inside Performance Max, allowing advertisers to individually toggle the Search Partner Network and Display Network. PPC Land coverage
  • June 2026 - Cloudflare Radar data shows bots now account for 57.4% of web traffic to HTML content, with human visitors at 42.6%. PPC Land coverage
  • June 2026 - Lunio publishes its Invalid Traffic Impact Report, based on 64 million clicks tracked from July 2025 through March 2026 across Google, Bing, LinkedIn, and Meta.

Summary

Who: Lunio, an invalid traffic detection and prevention platform based in the United Kingdom, published the report. The findings cover any advertiser running paid campaigns across LinkedIn, Microsoft Ads (Bing), Meta, and Google Ads.

What: A nine-month analysis of 64 million clicks found LinkedIn recording the highest IVT rate of any major ad platform at 15.34% average and 17.62% in Q1 2026. Bing averaged 11.63%. Meta's rate nearly doubled from 3.57% to 6.64% across two quarters. Within Google, Display IVT rose 132% from Q3 2025 to Q1 2026, and Search IVT rose 21%. Individual placement domains recorded IVT rates as high as 93.8%. Some single campaigns recorded IVT rates as high as 77%.

When: The data covers Q3 2025 (July to September), Q4 2025 (October to December), and Q1 2026 (January to March). The report was published in June 2026.

Where: The analysis covers paid advertising activity across LinkedIn, Microsoft Ads (Bing), Meta, and Google Ads. The worst-performing placement domains identified include gaming sites, mobile gaming applications, and made-for-advertising sites operating across multiple countries.

Why: The report matters because invalid traffic does more than waste budget directly. It corrupts the conversion signals that feed back into automated bidding algorithms, causing platforms to misallocate spend toward fraudulent interactions rather than genuine buyers. The problem has accelerated as AI has collapsed the cost of generating sophisticated fake traffic, while platform-level defences have not kept pace. Advertisers relying on platform reporting alone have no structural mechanism to detect sophisticated IVT that bypasses native filters and reaches their accounts.