Tunnel = a 1-mile stretch of road you own. Zone = a neighborhood-sized area of about 0.28 square miles that you own. Background = the whole member network, no geography limit. Full glossary
Definition
Impression fraud (Invalid Traffic / IVT): Impression fraud, formally Invalid Traffic (IVT), is any digital ad impression generated by something other than a legitimate human viewer: data-center bots, malware-driven hidden browsers, AI crawlers, ad-stacking, pixel-stuffing, domain spoofing, SDK spoofing, or abuse of server-side ad insertion. The Media Rating Council classifies IVT in two tiers, General Invalid Traffic (GIVT), which routine filtration catches, and Sophisticated Invalid Traffic (SIVT), which mimics human behavior and requires advanced analytics, multi-point corroboration and human investigation to detect.
Every CPM-priced campaign rests on one assumption: that the impression the advertiser paid for was rendered to a person. Impression fraud breaks that assumption mechanically. Something that is not a person loads an ad-bearing page or app screen, an impression event fires into the measurement chain, and the advertiser is billed exactly as if a human had been there.
The techniques have names, and they all attack the same weak point, which is that the buyer never observes the human directly. Domain spoofing sells worthless inventory under a premium publisher's name. Ad-stacking layers several ads in one slot so all of them bill while one is visible at best. Pixel-stuffing renders an ad into a one-pixel frame. SDK spoofing fabricates mobile-app impressions that no app ever served. Server-side ad insertion in connected TV lets impressions be generated upstream of any device the buyer can check.
What matters for a business deciding where to spend is the size, not the taxonomy, and the size is the part most articles on this subject get wrong.
Two sources measure this at scale and publish their method. DoubleVerify's 2025 Global Insights Report, covering measurement year 2024 across more than a trillion impressions, puts fraud and sophisticated invalid traffic at 1.0% in North America and 0.8% in the United States. The ANA's Q3 2025 transparency benchmark, built from impression-level logs matched across the buy and sell sides, puts invalid traffic at 0.3% of programmatic spend on web and mobile excluding connected TV.
Higher figures circulate widely, generally around a fifth of web impressions and a third of mobile-app impressions. Those come from vendor press releases distributed by newswire, with no stated sample and no stated test for what counted as invalid, and they differ from the measured benchmarks by roughly twenty times. This site quoted them until August 2026. It does not any more, for the same reason it does not quote agency blog CPLs: a number you cannot check is not evidence, however useful it would be to the argument.
The direction of travel is genuinely bad, and saying the level is one percent does not soften that. DoubleVerify measured bot fraud up 101% year over year in North America, driven largely by mobile-app video, and general invalid traffic up 86% in the second half of 2024, with 16% of that increase from AI crawler bots. Fast growth on a small base is the honest description, and it is also a reason to expect the level to stop being small if verification coverage slips.
Measurement year 2024, over 1 trillion impressions; United States 0.8%
DoubleVerify 2025 Global Insights Report, North AmericaQ3 2025, web and mobile excluding CTV, impression-level logs
ANA Programmatic Transparency Benchmark, Q3 2025Year over year; general invalid traffic up 86% in H2 2024
DoubleVerify 2025 Global Insights ReportIf bots take a cent of the dollar, the obvious question is what happens to the rest of it. The ANA's Q3 2025 benchmark answers that for web and mobile excluding connected TV. Transaction costs take 26.9%, of which SSP platform fees are 13.2% and DSP data fees 3.6%. Media-productivity loss takes 26.1%. 47.1% survives as TrueAdSpend, the share that became a benchmark-qualified impression: measurable, viewable, non-MFA and non-invalid. Whether a person then looked at it is a separate question the benchmark does not answer. The ANA builds the waterfall sequentially, so 26.9 plus 26.1 does not add to exactly 47.1. Q3 2025 was the last quarter the ANA itemized this cut; its current all-environments headline is 45.1% (Q2 2026).
Inside the productivity loss, the ranking settles the argument. Non-measurable impressions are 15.2% of the dollar and non-viewable impressions are 10.1%, against 0.5% for made-for-advertising inventory and 0.3% for invalid traffic. Impressions that no verification vendor could say anything about are fifty times the size of the fraud line. That is the leak, and it does not have a villain, which is probably why it gets a fraction of the coverage.
For structure rather than level, the ANA's December 2023 study is still the only fully itemized waterfall it has published: on $1,000 of open-web programmatic spend it traced $130 to SSP platform fees, $80 to DSP platform, $60 to DSP data and $20 to other DSP costs, then $150 to non-measurable impressions, $100 to made-for-advertising, $95 to non-viewable and $5 to invalid traffic, leaving $360 of working media, and it estimated $22 billion of available industry efficiency. We date-stamp it every time it appears, because the level has since improved to 47.1 cents and 36 cents is not today's number.
One point of vocabulary that gets abused in this area: the ANA's Q4 2025 findings report a higher-performing half at 56.7% of TrueAdSpend and a lower-performing half at 37.5%. That is a median split on the outcome, dividing the benchmark's participants into halves by result. It is not a comparison of advertisers who do and do not run a transparency program.
Q3 2025, web and mobile excluding CTV. Ninety times the invalid-traffic line
ANA Programmatic Transparency Benchmark, Q3 2025Of the ad dollar, Q3 2025
ANA Programmatic Transparency Benchmark, Q3 2025The ANA's 2023 study, structure only. Current level: 45.1% all environments (Q2 2026), 47.1% on the Q3 2025 ex-connected-TV cut
ANA Programmatic Media Supply Chain Transparency Study, December 2023| Line | Share of the ad dollar | What it is |
|---|---|---|
| Transaction costs | 26.9% | SSP platform 13.2%, DSP data 3.6%, DSP platform and other costs the balance |
| Non-measurable impressions | 15.2% | Served, but nothing about them could be verified |
| Non-viewable impressions | 10.1% | Measured, and failed the MRC viewability standard |
| Made-for-advertising inventory | 0.5% | Sites built to harvest ad spend rather than serve an audience |
| Invalid traffic (bots) | 0.3% | Impression fraud, the subject of this page |
| TrueAdSpend (benchmark-qualified) | 47.1% | Measurable, viewable, non-MFA, non-invalid. Not a measure of whether a person looked |
The MRC's Invalid Traffic Detection and Filtration Standards split IVT into two tiers. The split matters because a platform-level filter and third-party verification are not the same thing, and because the two tiers behave differently over time: GIVT grows with the crawler population, while SIVT grows with the money available to steal.
Two continuous seconds for video. A definition, never a rate
MRC and IAB Viewable Impression Measurement GuidelinesDisplay 73%, video 79%, measurement year 2024
DoubleVerify 2025 Global Insights Report| Category | What it is | Detection | Examples |
|---|---|---|---|
| GIVT: General Invalid Traffic | Non-human traffic that is straightforward to identify, either because it declares itself or because the behavior is obviously not a person. | Routine list-based and rules-based filtration. | Search-engine crawlers, AI crawlers (GPTBot, ClaudeBot, AppleBot), data-center IP ranges, declared spiders. |
| SIVT: Sophisticated Invalid Traffic | Bots and schemes that actively mimic human behavior (cursor jitter, scroll dwell, click cadence) to evade filtration. | Advanced analytics, multi-point corroboration and human investigation, per the MRC standard. | Hijacked devices running hidden browsers, ad-stacking, pixel-stuffing, domain spoofing, SDK spoofing in mobile apps, server-side ad insertion abuse in CTV. |
Google and Meta run large ad-traffic-quality operations and reject enormous volumes of invalid requests. They still do not catch everything, for three structural reasons that have nothing to do with effort.
The auction is built to fill. Every refused impression is revenue a platform does not book. Filters are therefore tuned to a tolerance rather than to zero, and the tolerance sits wherever platform revenue and advertiser patience meet.
SIVT is designed to evade. Sophisticated operators imitate cursor jitter, scroll dwell and click cadence closely enough that behavioral filtration alone misses them. The MRC standard is explicit that catching SIVT takes advanced analytics, multi-point corroboration and human investigation, which is not work an exchange performs inside a 100-millisecond auction.
Nobody owns the whole chain. The ANA's benchmark exists precisely because no single participant can see end to end: it takes impression-level logs from advertisers, matches them against sell-side logs through TAG TrustNet and Fiducia, and only then can anyone say where the dollar went. A platform can police its own hop. It cannot vouch for the hops on either side of it, which is also why 15.2% of the dollar lands on impressions nobody could measure at all.
Invalid traffic, non-measurable impressions and non-viewable impressions have one root cause in common: the advertiser pays for an inferred event, generated upstream, by infrastructure the advertiser does not control. Removing the inference removes all three at once, which is a bigger claim than removing fraud and a more useful one, because fraud was only a cent.
WilDi Maps prices on Cost Per Verified Delivery (CPVD): from $0.25 per verified delivery on the background tier, set by account tier (Enterprise $0.25 at $3,000 a month or more, Pro $0.32 at $1,000, Local $0.40 at $250, Starter $0.50), with tunnels and zones priced above the background rate for hyper-local precision. A delivery is not an impression and it is not a lead. It is one offer reaching one opted-in person whose location was verified, billed only when that person claims it.
We are not claiming this buys more impressions per dollar. On the published cross-channel CPM tables, social and digital display sit as low as or lower than out-of-home, and a delivery is not sold per thousand of anything. The claim is narrower and checkable: there is no SSP, DSP, exchange or data vendor between the advertiser and the phone, so the 26.9% transaction layer does not exist, and there is nothing unmeasurable to bill for because billing happens on the claim.
The product
WilDi Maps is not a single flat-rate product. You pick the tier that matches how local you need to be. All three are GPS-verified per claim, with no auction, no exchange rake, no Middleman Tax.
1-mile road strip
Premium
Hyper-local, just-in-time
Claim a one-mile stretch. When a member enters the strip, they get a just-in-time message, perfect for emergency services, on-route specials, and anything where being right there now beats brand awareness later.
Best for
0.28-square-mile area
Premium
Hyper-local, area-based
Claim a one-square-mile block, not tied to a single road. Catches the residential cluster, retail district, or industrial park where your work actually lives. Same just-in-time delivery as tunnels; different geometry.
Best for
City-wide rotation
From $0.25
per claim, tier-based
City-wide brand presence on rotation. Highest reach for the budget; best when familiarity beats precision. Per-delivery rate drops by tier (Enterprise: $0.25 / Pro: $0.32 / Local: $0.40 / Starter: $0.50). See /pricing for the live rate card.
Best for
What the member gets when an ad is claimed
If the member wants to act on the ad, the app navigates them straight to the advertiser's location.
Click-through to any URL: ordering page, brand site, blog post, lead form.
Open a specific page inside the WilDi app: promo details, daily specials, claim instructions.
See the full pricing breakdown on the pricing page.
Impression fraud is the generation of digital ad impressions that no real human ever saw, using bots, malware-controlled hidden browsers, ad-stacking, pixel-stuffing, domain spoofing, SDK spoofing or server-side ad insertion abuse, and billing the advertiser as though the impressions had reached real viewers. The industry term is Invalid Traffic (IVT), and the Media Rating Council splits it into General Invalid Traffic, which routine filtration catches, and Sophisticated Invalid Traffic, which mimics human behavior to evade filtration.
About one percent as measured. DoubleVerify's 2025 Global Insights Report, covering measurement year 2024 across more than a trillion impressions, put fraud and sophisticated invalid traffic at 1.0% in North America and 0.8% in the United States. The ANA's Q3 2025 benchmark put invalid traffic at 0.3% of programmatic spend on web and mobile excluding connected TV. Much higher figures circulate, but they come from vendor press releases without a stated method or sample, and we do not quote them.
No source at our sourcing bar publishes a credible global dollar total, so we do not quote one. The forecasts that circulate (tens or hundreds of billions lost to ad fraud) trace back to research whose accessible primary is years out of date, and one widely repeated figure is a forecast of ecommerce fraud rather than ad fraud altogether. What can be measured is the rate: 1.0% of North American impressions and 0.3% of programmatic spend. Multiply that by your own budget and you have a defensible estimate for your business, which is more useful than an industry total nobody can check.
The chain and the unmeasurable inventory. On the ANA's Q3 2025 benchmark for web and mobile excluding connected TV, transaction costs take 26.9% of the ad dollar and media-productivity loss takes 26.1%, leaving 47.1% as benchmark-qualified working media. That was the last quarter the ANA itemized the ex-connected-TV cut; its current all-environments headline is 45.1% (Q2 2026). Within the productivity loss, non-measurable impressions are 15.2% and non-viewable impressions are 10.1%, against 0.3% for invalid traffic. Fees and unmeasurable inventory are two orders of magnitude larger than the fraud line.
GIVT (General Invalid Traffic) is non-human traffic that is easy to identify: declared crawlers, AI crawlers such as GPTBot and ClaudeBot, data-center IP ranges, and behavior that is obviously not a person. The MRC says routine list-based and rules-based filtration catches it. SIVT (Sophisticated Invalid Traffic) is the harder tier, covering hijacked devices running hidden browsers, ad-stacking, pixel-stuffing, domain spoofing, SDK spoofing in mobile apps and server-side ad insertion abuse in connected TV. The MRC standard says SIVT detection requires advanced analytics, multi-point corroboration and human investigation.
It has grown fast from a small base. DoubleVerify measured bot fraud up 101% year over year in North America, driven largely by mobile-app video, and general invalid traffic up 86% in the second half of 2024, with 16% of that increase attributed to AI crawler bots. The measured level nonetheless stayed near one percent, because that figure is what remains after filtration on campaigns running verification. Growth rate and level are different claims, and conflating them is how the double-digit numbers get manufactured.
Three practical signals. First, impressions and clicks that do not match downstream behavior: bots fire events but do not scroll, dwell or convert. Second, a high share of traffic from data-center IP ranges or anonymizing proxies, which a third-party verification vendor can flag. Third, spikes in impressions or clicks with no matching change in calls, form fills or sales. Running third-party verification is the direct answer, though for a small local budget the arithmetic is worth doing first: at a measured 1%, verification has to cost less than the cent it protects.
Three structural reasons. The auction is built to fill, so filters are tuned to a tolerance rather than to zero, because every refused impression is revenue foregone. Sophisticated invalid traffic actively evades behavioral detection, and the MRC standard says catching it needs analytics and human investigation that no exchange performs inside a 100-millisecond auction. And no single party sees the whole chain: the ANA's benchmark exists because it takes impression-level logs from advertisers matched against sell-side logs before anyone can say where the dollar went. A platform can police its own hop, not the hops either side of it.
CPVD (Cost Per Verified Delivery) is WilDi Maps' pricing model: from $0.25 per verified delivery on the background tier, set by account tier, with tunnels and zones priced higher for hyper-local precision. The advertiser pays when an opted-in person, driving through the area the advertiser chose, claims the offer on their own phone. The location signal is device-side rather than inferred from a bid stream, and the event is logged by infrastructure WilDi controls end to end, so there is no exchange layer to insert anything into. That removes the fraud surface, but more importantly it removes the 26.9% of the dollar that transaction costs take and the 15.2% that lands on impressions nobody could measure.
One verified delivery of your offer to the phone of one real person who was physically inside your chosen geography at that moment, confirmed by GPS on the device itself. That person also taps to acknowledge the offer, so a delivery is never an invisible impression. Bots, background tabs, and off-screen impressions cannot generate one. You are billed only when a verified delivery happens.
The Starter tier opens with a $50 deposit, and that deposit becomes your ad budget. Background deliveries on Starter run $0.50 per verified delivery, so the first deposit buys 100 GPS-verified deliveries to the phones of real local people out on the road. There is no auction and no platform fee stacked on top.
Real local people who opted in to get offers in our app while they are out driving. Not bots, not a gig fleet. Location is reported by the phone itself, and each person earns a share of every offer they accept. There is no ad exchange in the loop, which is why bot exposure is 0%.
No. Every tier has a fixed, published rate per verified delivery. The price you see is the price you pay, whether it is game day or a Tuesday morning. Higher tiers carry lower per-delivery rates.
About this analysis
Written by Timm Ross, founder of WilDi Maps · Jacksonville-based · Veteran-owned. Sources are cited inline; we update the numbers when the underlying research updates.
Published · Last updated
More about Timm RossFixed from $0.25 per GPS-verified delivery. No auction, no bid stream, no Middleman Tax.
