One in twelve of your ad clicks is not a person: 8 ways to stop paying for them
Oct 18, 2026
Roughly one in twelve paid ad clicks is not a person.
That is the finding of the Global Invalid Traffic Report 2026, covered by the Association of National Advertisers in January: an 8.51% invalid traffic rate across global paid media, and about 63 billion dollars a year paying for it. The rate is not evenly spread. TikTok came in at 24.2%, LinkedIn at 19.88%, Google Video Partners at 20.62%, and Google Ads across its ecosystem at 7.57%.
Before anyone cancels their media plan, a caveat that belongs at the top rather than buried at the bottom. That report comes from Lunio, a company that sells invalid traffic protection. Vendors measuring the problem they solve tend to find a large problem. The figures are still worth taking seriously, and they are worth reading as the upper end of a range rather than a precise bill.
The ANA study is the more neutral of the two, since a trade body representing advertisers has no product to sell against the finding. It is also older, with fieldwork running from September 2022 to January 2023. Treat it as the best available map of the supply chain rather than this quarter's accounts.
Pull your analytics and look for the signature: sessions with a sub-second duration, a single page view, no scroll and no interaction, clustered at odd hours or from data centre networks. Compare that cohort against your conversion rate. If a placement sends volume and converts nothing, you have found something worth investigating whether or not it meets anyone's definition of fraud.
Open your placement report and sort by spend. Most advertisers find a long tail of apps, parked domains and auto-playing video properties they would never have selected deliberately. Exclude them. Then check again next month, because the tail regrows.
Sometimes they are right. Often they are the mechanism by which a tightly targeted campaign quietly becomes a loosely targeted one. Switch them off, establish what your genuinely intentional spend performs at, and reintroduce them one at a time with a measurement window attached.
Decide what a real outcome is: a qualified lead, a booked call, a closed sale. Then make that the optimisation target rather than the top-of-funnel proxy that happened to be easy to track.
Offline conversion imports close that loop. When a lead becomes qualified, push that event back with its value. This is integration work rather than a setting, which is why it is skipped so often and why the accounts that do it pull ahead.
Invisible bot detection rather than a puzzle that punishes real users, honeypot fields, submission rate limiting by address, and basic validation that rejects obvious nonsense before it reaches your sales team. The goal is not to block everything. It is to stop paying a salesperson to call a phone number that was never real.
Ask your agency or platform for a log-level report showing where impressions actually ran. Build an inclusion list of sites you are willing to appear on rather than relying on exclusion lists you update after the damage. Fewer, better placements consistently beat broad reach that nobody can account for. If nobody can produce the report, that is itself the finding.
Server-side tracking, a CRM that records the source of every enquiry, and a simple weekly reconciliation between what the platforms claim and what actually landed in your pipeline. When those two numbers diverge, the gap is the most valuable report in your business.
An hour of work, no software purchase, and at the end you will know whether your problem is bots or something closer to home.
That is the finding of the Global Invalid Traffic Report 2026, covered by the Association of National Advertisers in January: an 8.51% invalid traffic rate across global paid media, and about 63 billion dollars a year paying for it. The rate is not evenly spread. TikTok came in at 24.2%, LinkedIn at 19.88%, Google Video Partners at 20.62%, and Google Ads across its ecosystem at 7.57%.
Before anyone cancels their media plan, a caveat that belongs at the top rather than buried at the bottom. That report comes from Lunio, a company that sells invalid traffic protection. Vendors measuring the problem they solve tend to find a large problem. The figures are still worth taking seriously, and they are worth reading as the upper end of a range rather than a precise bill.
Where the money goes
8.51% of global paid ad traffic is invalid, costing around 63 billion dollars a year.
Roughly one click in twelve. The UK figure is 7.97%, or about 2.9 billion dollars.Lead generation campaigns see invalid traffic rates 32.07% higher than transactional ones.
If you buy clicks to a form rather than to a purchase, you are in the worse half of this distribution.Only 36 cents of every dollar entering a demand-side platform reaches a consumer.
From the ANA Programmatic Media Supply Chain Transparency Study. Around 29 cents goes to intermediary fees and 35 cents to low-quality media. This covers open web programmatic only, not search or the walled gardens.About 22 billion dollars of 88 billion in open web programmatic was judged wasteful or unproductive.
Roughly a quarter. The study tracked 123 million dollars of spend and 35.5 billion impressions across 21 marketers.Made-for-advertising sites grew from around 5% of web auctions in early 2020 to nearly 30% by mid-2023.
Pages built to carry ads rather than to be read, and your budget reaches them through automated buying you did not inspect.
The ANA study is the more neutral of the two, since a trade body representing advertisers has no product to sell against the finding. It is also older, with fieldwork running from September 2022 to January 2023. Treat it as the best available map of the supply chain rather than this quarter's accounts.
1. Measure your own rate before you buy a solution
Industry averages tell you the weather, not whether your roof leaks. A local service business running branded search has almost nothing in common with a lead generation campaign on display.Pull your analytics and look for the signature: sessions with a sub-second duration, a single page view, no scroll and no interaction, clustered at odd hours or from data centre networks. Compare that cohort against your conversion rate. If a placement sends volume and converts nothing, you have found something worth investigating whether or not it meets anyone's definition of fraud.
2. Look at the placements you never chose
The worst numbers in the report are not search. They are video partners at 20.62% and display at 12.02%, which are the networks your campaigns reach into automatically when you leave the defaults alone.Open your placement report and sort by spend. Most advertisers find a long tail of apps, parked domains and auto-playing video properties they would never have selected deliberately. Exclude them. Then check again next month, because the tail regrows.
3. Turn off the expansions nobody asked for
Audience expansion, display network opt-ins on search campaigns, broad match without negatives, automatically created assets. Each one exists to spend your budget somewhere the platform thinks is promising.Sometimes they are right. Often they are the mechanism by which a tightly targeted campaign quietly becomes a loosely targeted one. Switch them off, establish what your genuinely intentional spend performs at, and reintroduce them one at a time with a measurement window attached.
4. Optimise to the conversion that pays you
This is the largest and least discussed source of waste, and it is entirely self-inflicted. If your campaign optimises to form submissions, the platform will faithfully find you people who submit forms. Some of them will be bots, more of them will be unqualified, and the algorithm cannot tell the difference because you never told it.Decide what a real outcome is: a qualified lead, a booked call, a closed sale. Then make that the optimisation target rather than the top-of-funnel proxy that happened to be easy to track.
5. Send the real outcomes back to the platform
Following from the above, the platforms cannot learn what they cannot see. If your sales qualification happens in a CRM and never returns to the ad account, you are asking an optimisation system to improve a number that has no relationship to your revenue.Offline conversion imports close that loop. When a lead becomes qualified, push that event back with its value. This is integration work rather than a setting, which is why it is skipped so often and why the accounts that do it pull ahead.
6. Defend the form, not just the click
Given that lead generation carries meaningfully higher invalid traffic, the landing page is your second line of defence and usually the neglected one.Invisible bot detection rather than a puzzle that punishes real users, honeypot fields, submission rate limiting by address, and basic validation that rejects obvious nonsense before it reaches your sales team. The goal is not to block everything. It is to stop paying a salesperson to call a phone number that was never real.
7. Inspect the supply chain if you buy the open web
If any of your budget goes through programmatic display, the ANA numbers apply to you and the remedy is unglamorous.Ask your agency or platform for a log-level report showing where impressions actually ran. Build an inclusion list of sites you are willing to appear on rather than relying on exclusion lists you update after the damage. Fewer, better placements consistently beat broad reach that nobody can account for. If nobody can produce the report, that is itself the finding.
8. Keep a measurement system the platforms do not own
Every ad platform reports favourably on its own performance, which is not a conspiracy so much as a conflict of interest operating exactly as you would expect.Server-side tracking, a CRM that records the source of every enquiry, and a simple weekly reconciliation between what the platforms claim and what actually landed in your pipeline. When those two numbers diverge, the gap is the most valuable report in your business.
The waste nobody sells a product for
Here is the honest version, which is less dramatic than the headline figures and more useful. For most small and mid-sized advertisers, invalid traffic is a real cost and not the main one.
The platforms do filter a great deal of invalid activity before you are billed, and they issue credits for some of what gets through. The figures in reports like these are measured before that filtering, and the methodologies vary enough that two vendors can produce very different numbers from the same campaign. Meanwhile the quieter waste sits in places nobody sells protection against: campaigns optimising to a form fill that generates no revenue, leads arriving at an inbox nobody checks until Tuesday, a landing page that loads slowly on the phones carrying most of the traffic, and a sales process that never reports back what converted.
That waste does not have a vendor or a conference talk, so it gets less attention than fraud does. It is also usually larger, and unlike the bots, it is entirely within your control.
The useful order of operations follows from that. Fix your measurement first, because without it every other decision is a guess. Then fix what you optimise towards, because that single setting directs everything the platform does on your behalf. Then exclude the placements doing nothing for you. Then, if your numbers still look wrong, go looking for bots. Most advertisers who work through that list in order find the answer before they reach the last step.
Three checks to run this week
Sort your placement report by spend and read the bottom half.
Ten minutes. Exclude anything you would be embarrassed to see your brand on, and anything that has spent money without producing a single conversion.Compare last month's platform-reported conversions against actual enquiries in your CRM.
If the two numbers are not close, stop optimising until you know why.Check what your campaigns are optimising towards.
Open the conversion settings and ask whether that action is something you would pay for individually. Many accounts are quietly buying form submissions at the price of a customer.
An hour of work, no software purchase, and at the end you will know whether your problem is bots or something closer to home.
