You're Paying for Clicks. Invalid Traffic Is Eating the Rest.

Open your ad dashboard and look at the clicks you paid for last month. A real slice of them was invalid traffic, meaning nobody was actually on the other end.

A 2026 report from ad-verification company Lunio, independently covered by MediaPost and the Association of National Advertisers, analyzed 2.7 billion ad clicks across 6 platforms and 10 countries and found an overall invalid traffic rate of 8.51%, worth $63 billion in wasted global ad spend. Even Meta, the cleanest of the major social platforms, still runs an 8.20% invalid traffic rate. That's roughly one in 12 clicks on the exact platform your creator campaigns live on, and none of them were ever going to buy anything.

Pay-per-click pricing has no way to catch that at the moment the charge fires.

What Invalid Traffic Costs Your Affiliate Program

That $63 billion in wasted spend globally is spread across every account running paid clicks, affiliate and creator campaigns included, even though those get treated as somehow immune because they're "performance based."

They aren't immune. Invalid traffic doesn't check whether a click came from a display ad or a creator's link before it fires. Lunio's report found lead-generation businesses see invalid traffic rates 32.07% higher than ecommerce brands, and a lot of affiliate activity, the click that starts a journey rather than finishes one, behaves more like lead gen than like a completed sale.

Picture the actual mechanics of one campaign. A creator posts, a hundred people click through, and your dashboard reports a hundred clicks against your budget. Somewhere between eight and 24 of those, depending on the platform, weren't a person at all. You're billed the same either way, and the campaign report that lands in front of you next week treats all of them as equally real.

There's one thing worth saying. Lunio sells invalid traffic detection, so the report comes from a company with a commercial reason to make the number sound alarming. What makes it usable anyway is that two independent trade outlets with no stake in Lunio's product, MediaPost and the ANA, picked the figures up and reported them on their own terms.

Performance marketing has priced three different units of value over its history, moving from the impression to the click to the completed sale, and each shift happened because the previous unit stopped explaining enough about what actually worked. Invalid traffic is the reason the click is now failing that same test. A unit that can be faked for pennies was never going to hold up forever as the thing brands pay for.

Why Even Meta Isn't Clean

Facebook and Instagram, the platforms your DM campaigns actually run on, post the best numbers of any major social network at 8.20% invalid traffic. TikTok sits at 24.2%. LinkedIn comes in at 19.88%. X lands at 12.79%.

Meta being the cleanest option doesn't make it clean. It means one in 12 clicks you're paying for on your best-performing platform still weren't real. Years of litigation pressure pushed Meta to invest harder in bot detection than most competitors, and the number still sits at 8.20%.

There's no platform where invalid traffic hits zero, and there isn't going to be one soon. That matters more for affiliate and creator campaigns than it does for a generic display buy, because those campaigns are usually judged on a tighter margin. A brand running broad awareness ads can absorb some invalid traffic inside a bigger budget without anyone noticing. An affiliate program justifying its existence line by line to a finance team notices immediately, and invalid traffic is exactly the kind of quiet, unexplained gap between clicks and results that makes a working channel look worse than it is.

What a Click Based Model Can Never Filter Out

Timing is the deeper problem here, more than the raw percentage.

A click-based system charges you the instant the click registers, before anyone knows whether a real person made it. By the time invalid traffic gets identified and refunded, weeks later if it happens at all, that data has already trained your bidding algorithms, already shaped which audiences your campaign chases next, already distorted the numbers a budget decision got made on.

A separate Lunio survey of 131 senior marketers found 75.6% lose more than 5% of their monthly performance budget to invalid traffic, and 85.6% said they're specifically concerned about agentic AI making it worse. Automated bidding systems learn from whatever conversion signals they're fed, and a bot posing as a converting customer teaches the algorithm to chase more bots that look exactly like it. The system gets worse at finding real buyers precisely because it's optimizing against fake ones, and it does this quietly, with no alert telling you it happened.

Linka built cost per DM around a different idea entirely. You price the event a bot can't cheaply fake, instead of the event that's easiest to measure. A click costs almost nothing to fake at scale. A qualified, two-way conversation about a real product question doesn't, because faking one requires the kind of specific, contextual reply that bot infrastructure isn't built to produce cheaply across thousands of attempts.

That's the real case for pricing around the conversation instead of the click, and it's why treating the DM as its own priced event is more important than it sounds like it should be on first read.

How Invalid Traffic Can't Fake a Qualified DM

A bot answering a comment can fire off a generic reply cheaply. A bot carrying an actual back and forth about sizing, shipping timing, or whether a product suits a specific need, at the volume needed to make faking it profitable, is a far harder and more expensive thing to build than a click farm.

That's the structural reason cost per DM sidesteps the invalid traffic problem instead of just managing around it. Linka's model treats the qualified DM as a compensable event in its own right, tracked from the moment a shopper comments a keyword through the conversation to the dynamic discount conversion page built for that campaign. Every part of that path is a real exchange, not a single click a script can reproduce a million times over.

The same resistance holds between campaigns, too. Approved products sit inside creators' AI Shopping Agents, answering shopper questions against real product information rather than a generic script, which is exactly the kind of exchange that stays expensive to fake no matter how the surrounding technology improves.

Brands rebuilding their partner mix around what actually converts rather than what's easiest to bill run into this same pattern from a different angle. The cleanest number in a report isn't always the most honest one.

Stop Paying for Invalid Traffic and Start Paying for Real Conversations

Invalid traffic isn't going away, and no platform, Meta included, is close to eliminating it. What's actually available is a different question. You can keep paying for events a bot can fake for pennies, or start paying for the one kind of event it structurally can't.

Ready to stop paying for invalid traffic? Book a call with Linka or explore Linka's affiliate marketing platform for brands to see what cost per DM looks like against your own numbers.