Digital Ad Fraud Statistics 2026: Every Rate With Its Source

37 Digital Ad Fraud Statistics, Trends, and Data Points for 2026

There is no single ad fraud rate. Published 2026 figures run from 0.1% invalid traffic in the ANA’s Q2 2026 programmatic benchmark to 41% on mobile app inventory in Pixalate’s Q2 2026 global open-auction data. The gap comes from what each firm counts and what it divides by. On money, the most cited figure is a vendor forecast: Juniper Research modeled $84 billion, or 22% of online ad spend, lost to fraud in 2023, rising to $172 billion by 2028.

Below, every number sits next to who measured it, on what sample, when, and how. Figures older than 2023 are marked historical.

Ad fraud rate: why published numbers run from 0.1% to 41%

Read this table from top to bottom. Each row uses a different denominator, which is why the numbers look like they disagree. They mostly do not. They measure different things.

RateWhat it measures (the denominator)Who, sample, periodMethod
0.1%Invalid traffic as a share of tracked programmatic spend for large US advertisersANA Programmatic Transparency Benchmark with TAG TrustNet and Fiducia, Q2 2026Log-level data from advertisers and platforms
0.6%Fraud/SIVT violation rate on DoubleVerify-measured North American media, down 41% year over yearDoubleVerify 2026 Global Insights, 2025 data, as reported by PPC LandVendor measurement on client campaigns, mostly protected
0.7%Fraud rate in campaigns using anti-fraud tools (optimized)IAS Media Quality Report, 20th edition, 2024 data, 280B+ daily interactionsImpression-level vendor measurement
0.82%IVT in TAG Certified Channels, US (non-certified: 1.19%)TAG US Fraud Benchmark, Jan to Jun 2023The 614 Group, agency data from MRC-accredited vendors, no sampling
5.58%Share of paid ad clicks flagged as fraud (4.32% in full-year 2025)Spider AF, 4.006B paid clicks, H1 2026Vendor detection platform (click fraud, not impressions)
10.9%Fraud rate in campaigns with no fraud protection (non-optimized), a four-year highIAS, end of 2024Impression-level vendor measurement
20% / 26% / 41%IVT share of impressions on web / CTV / mobile appPixalate, 80B+ global impressions, Q2 2026Mostly buy-side open-auction traffic, MRC-accredited for SIVT
22%Share of all online ad spend lost to fraudJuniper Research, 2023 estimate, 78,700 datapoints, 45 countriesForecast model, released by vendor Fraud Blocker
53%Automated share of all web traffic, of which 40% is malicious bots. Not an ad metric.Imperva (Thales) 2026 Bad Bot Report, 2025 dataBot-protection network traffic

Three reasons vendors disagree

  1. Denominator. Share of spend, share of impressions, share of clicks and share of all web traffic are four different numbers. A 41% IVT rate on open-auction app impressions and a 0.1% IVT rate on audited spend from big marketers can both be true.
  2. Which supply is measured. IAS, DoubleVerify and the ANA benchmark mostly see campaigns already using verification and pre-bid filtering. Pixalate’s sample is “predominantly buy-side open auction” traffic. IAS shows the effect directly: unprotected campaigns ran 15x the fraud of protected ones in 2024.
  3. Definition. Some count all invalid traffic (IVT), some only sophisticated invalid traffic (SIVT) or “fraud” with intent. Pixalate uses “ad fraud” for IVT where traffic “suggests deliberate intent to mislead.” Since June 2020 the MRC’s IVT standards also require accredited measurers to disclose a “decision rate”: the percentage of impressions where the vendor had enough data to rule valid or invalid at all. Impressions with no decision are in neither column, so ask for that number.

GIVT vs SIVT, in the MRC’s words

The Media Rating Council split invalid traffic in two on October 27, 2015. General invalid traffic (GIVT) is caught by routine, list-based filtering: known bots, spiders and crawlers, non-browser user agents, pre-fetched or pre-rendered pages. Sophisticated invalid traffic (SIVT) needs “advanced analytics, multipoint corroboration, and human intervention”: hijacked devices, hijacked ad tags or creative, adware, malware, misappropriated content. The June 25, 2020 update added more specific SIVT areas, data center rules, OTT considerations and logged-in environments. When a vendor quotes a “fraud rate”, ask which of the two it covers.

Global ad fraud losses: forecasts, not audits

No one audits global ad fraud. Every dollar total below is modeled, usually by a firm that sells or sponsors detection. Use them for scale, not precision.

StatisticSource and dateSample and methodType
$84B lost to fraud in 2023, 22% of online ad spend; 30% of mobile ad spendJuniper Research, released by Fraud Blocker, Sept 26, 202378,700+ datapoints, 45 countries, 8 regions, 2019 to 2028Vendor estimate
$172B lost by 2028; $23B a year recoverable with mitigation toolsSame reportSame modelForecast
$25.3B of ad spend “exposed” to fraud in H1 2026 ($50.6B annualized)Spider AF, Sept 2, 20265.58% click fraud rate applied to EMARKETER’s $907.0B 2026 global digital ad spend forecastVendor estimate (rate times spend)
About $1.8M lost per billion CTV impressions in unprotected campaignsDoubleVerify, May 7, 2026DV measurement data plus controlled tests without protectionVendor estimate
$10.8B saved by advertisers in 2023 through industry anti-fraud programsTAG, Oct 29, 2024Separate TAG analysis cited in the releaseTrade body estimate
Historical: $68B lost in 2022, up from $59B in 2021; US above $23B (35%); top 5 countries 60%Juniper Research, Feb 21, 2022Methodology not published in the releaseVendor forecast

Two cautions. The $68B, $84B and $172B figures come from the same Juniper model, so they confirm each other’s lineage, not each other. And Spider AF’s $25.3B is a click rate multiplied by all digital spend, which assumes every dollar faces the same fraud rate as its clients’ search and social clicks.

Ad fraud by channel: web, mobile app and CTV

Pixalate publishes the most regular public series by environment. Its sample skews to open-auction programmatic, so treat these as the risk on the open exchange, not on direct deals or walled gardens.

Bar chart of Pixalate's Q2 2026 global invalid traffic rates: mobile app 41%, CTV 26%, web 20%.
Source: Pixalate, 2026. Sample is mostly buy-side open-auction traffic, not direct deals or walled gardens.
QuarterWeb IVTMobile app IVTCTV IVTImpressions analyzed
Q3 202521%33%19%106B+
Q1 202620%39%25%82B+
Q2 202620%41%26%80B+

Country rates swing far more. In Pixalate’s Q1 2026 EMEA release the UK’s CTV IVT rate hit 46%; in Q2 2026, mobile app IVT was 36% in the UK and 37% in Germany.

CTV is where the newest schemes show up. DoubleVerify’s May 2026 streaming report found 140% more CTV fraud schemes and variants in Q1 2026 than in Q1 2025, 50+ distinct CTV bot attacks in 2025, and 10x more fraudulent CTV apps in 2025 than in 2024. Bots made up 82% of the violations it found in North America, while data center traffic dominated in APAC (98%), LATAM (91%) and EMEA (66%). It also found bots behind 34% of impressions in one direct consumer healthcare buy and 25% in a major CPG direct buy, so “direct” is not a fraud guarantee. Our connected TV advertising statistics cover the spend side of that channel.

Protected vs unprotected campaigns

The clearest finding across vendors: the same open web carries very different fraud rates depending on whether the buyer filters it.

Bar chart of IAS 2024 global fraud rates: 10.9% for non-optimized campaigns and 0.7% for optimized campaigns, a 15x gap.
Source: Integral Ad Science, 2025. Measured by a vendor that sells the protection.
ComparisonProtectedUnprotectedSource, period
Global fraud rate0.7% (optimized, down 9.8% YoY)10.9% (non-optimized, up 19.0% YoY)IAS, 2024
CTV fraud rateUnder 1% (DV-protected)Nearly 9%DoubleVerify, May 2026 report
US IVT, certified vs non-certified channels0.82%1.19% (46% higher)TAG, H1 2023
European IVT, certified vs non-certifiedUnder 1% for six years62% higher in non-certifiedTAG Europe, 2023 data, 308B+ impressions
Historical: European CTV IVT1.28% (certified)8.85% (non-certified)TAG Europe, 2022 data

IAS also breaks out optimized US fraud by format for 2024: desktop display 1.6% (from 2.2%), desktop video 1.1% (from 1.6%), mobile web display 0.7% (from 0.8%) and mobile web video 0.5% (flat). TAG held US certified-channel IVT under 1% for a fourth straight year in H1 2024, across 745 billion impressions, up from 6.5 billion in its first study. Historical: the 2022 US figure was 0.98% for January to June 2022.

One honest caveat. Certified-channel and vendor-protected rates are measured by the same industry that sells the protection, on supply that has already been filtered. They show filtering works on what it touches. They do not tell you the rate on inventory you buy without it.

Imperva Bad Bot Report: automated traffic percentage, 2024 and 2025

Imperva’s annual report is the source most people mean when they say “half the internet is bots.” It measures all web traffic across Imperva’s network, not ad impressions, so it is context for ad fraud, not an ad fraud rate.

Metric202320242025
Automated share of all web trafficn/a51%Over 53%
Bad (malicious) bots, share of all traffic32%37%40%
Simple bad bot trafficJust under 40%45%n/a

Sources: Imperva 2025 Bad Bot Report (published April 15, 2025, 2024 data) and the 2026 Bad Bot Report announcement (May 5, 2026, 2025 data), which also reports AI-driven bot attacks up 12.5x. Bots that never render ads are a different problem from ad blocking, where a real human sees no ad at all; both shrink the audience you think you bought.

Made-for-advertising sites and programmatic waste

MFA sites are pages built to resell ad slots. Their impressions may reach humans, so they are waste rather than fraud, but they sit on the same exchanges. The ANA’s numbers below match our programmatic advertising statistics.

  • ANA Programmatic Media Supply Chain Transparency Study (Dec 5, 2023): only 36% of every dollar that enters a DSP “effectively reaches the consumer.” 35% of each dollar went to non-viewable, IVT, non-measurable and MFA traffic combined. $22B of efficiency gains possible in an $88B open web market. Buyers used 40,000-plus websites. Sample: 21 marketers, 12 supply chain companies (3 DSPs, 6 SSPs, 3 verification firms).
  • ANA first look, as reported by Digiday (June 19, 2023): MFA took 15% of spend and 21% of impressions; the average campaign ran on 44,000 websites.
  • ANA benchmark, Q1 2026 (May 27, 2026): TrueAdSpend 43.3%, MFA exposure 1.1% (up from 0.4% to 0.6% through 2025), 86 participating marketers. Q2 2026 (Aug 6, 2026): TrueAdSpend 45.1%, invalid traffic 0.1%.

The jump from 15% MFA in 2023 to about 1% in 2026 reflects a different, self-selected panel with exclusion lists in place, not proof the open web cleaned up. Context-based buys run through the same exchanges, so contextual advertising still needs the same IVT checks.

Ad fraud cases with court and takedown records

Rates are estimates. These are documented schemes with dates you can check.

SchemeHow it workedScaleOutcome and source
Methbot (historical, Sept 2014 to Dec 2016)Datacenter servers posed as humans on spoofed publisher domains, mainly video ads2,000+ servers in Dallas and Amsterdam, 6,000+ publisher domains spoofed (incl. The New York Times), 765,000+ leased IPs; victims paid $7M+Aleksandr Zhukov convicted May 2021, sentenced Nov 10, 2021 to 10 years plus $3,827,493 forfeiture (DOJ, EDNY)
3ve (historical, Dec 2015 to Oct 2018)Kovter botnet loaded ads in hidden browsers on infected PCs1.7M+ infected computers; businesses paid $29M+13-count indictment of 8 defendants unsealed Nov 27, 2018; FBI sinkholed 31 domains and searched 89 servers; DOJ credited White Ops (now HUMAN) and Google (DOJ, EDNY)
BADBOX 2.0 (disclosed March 5, 2025)Backdoored off-brand Android (AOSP) devices plus “evil twin” apps1M+ infected devices (vs 74,000 in the 2023 BADBOX); 24 evil twin apps; 5B fraudulent bid requests a week at peak; ~1,000 ad-heavy gaming sites for cashoutDisclosed and partly disrupted by HUMAN’s Satori team with partners (HUMAN)

How to check your own invalid traffic

Industry rates tell you the weather. Your own accounts tell you whether you are getting wet. Work through these in order.

  1. Google Ads: add the invalid click columns. Campaigns menu, Campaigns, Columns icon, search “Invalid clicks”, Apply (Google Ads Help). Google defines the related invalid click rate metric as clicks filtered out divided by filtered plus non-filtered clicks (Google Ads API reference). You are not charged for these. Clicks caught after billing come back as credits: Billing, Summary, “Adjustments” in the Month card, “Invalid activity.”
  2. Do the math once. Illustrative example: a campaign shows 2,300 clicks and 200 invalid clicks. Invalid click rate = 200 ÷ (2,300 + 200) = 8.0%. Those 200 were already removed. The question is what got through, which is the next step.
  3. Reconcile billed clicks against server logs. Count landing page hits carrying a gclid for the same dates, for example grep -c "gclid=" access.log, and compare with billed clicks. Then list the top sources: grep "gclid=" access.log | awk '{print $1}' | sort | uniq -c | sort -nr | head -20. A gap is normal (people bounce before the page loads), so look for patterns: one IP or subnet with dozens of gclids, bursts at 3 a.m., identical user agents.
  4. Check who the heavy IPs are. Run host <IP>. A hosting provider rather than a consumer ISP is a data center signal. Do not exclude Google’s own crawlers: verify them with a reverse then forward DNS lookup to googlebot.com, google.com or googleusercontent.com, or against Google’s published IP lists, where AdsBot sits in special-crawlers.json (Google Search Central).
  5. Exclude and report. You can exclude up to 500 IP addresses per campaign, but not in Video, App, Hotel, Performance Max or Smart Display campaigns (Google Ads Help). For activity in the last 60 days, file the Click Quality form with your customer ID, dates, campaigns, server logs with IPs and GCLIDs, and for junk leads a spreadsheet of GCLIDs (Google Ads Help). Google says reviews typically take several business days and it will not confirm individual clicks.
  6. Know what GA4 does not tell you. GA4 automatically excludes known bots using Google research and the IAB’s International Spiders and Bots List. You cannot turn that off or see how much was excluded (Analytics Help). That is list-based, GIVT-style filtering. Bots that pass as humans land in your reports as sessions, so a clean GA4 view is not proof of clean traffic. Our GA4 setup walkthrough covers the configuration side.
  7. Protect the form, not just the click. Google itself recommends reCAPTCHA, server-side validation and double opt-in on lead forms, plus optimizing to qualified or converted leads rather than raw form fills (Google Ads Help).

For programmatic and CTV buys, ask your DSP or verification vendor for three numbers per line item: post-bid IVT rate, GIVT vs SIVT split, and decision rate. Compare the IVT rate with the open-market benchmark that matches the environment: Pixalate’s 20% web, 41% app or 26% CTV for open auction, and IAS’s 10.9% for unprotected campaigns. My view: if a vendor cannot give you the decision rate, its fraud rate is incomplete. If you would rather have someone audit this across your accounts, you can book a consultation.

Which ad fraud statistic to cite

If your claim is about…Cite thisDo not use it for
Total money lost worldwideJuniper: $84B (22%) in 2023, $172B forecast for 2028, labeled as a vendor modelYour own campaign’s loss rate
Open-exchange risk by channelPixalate Q2 2026: web 20%, mobile app 41%, CTV 26% IVTDirect deals, walled gardens, or “share of spend”
Value of fraud protectionIAS 2024: 0.7% optimized vs 10.9% non-optimized, 15x gapA claim that fraud is solved
Certified supply chainsTAG: US certified channels under 1% IVT, H1 2024, 745B impressionsAll US inventory
How much of a programmatic dollar worksANA 2023: 36 cents reach the consumer; ANA Q2 2026 TrueAdSpend 45.1%A fraud rate (it mixes viewability, fees, MFA and IVT)
CTV fraud trendsDoubleVerify May 2026: 140% more CTV schemes Q1 2026 vs Q1 2025; ~$1.8M per billion unprotected impressionsMeasured loss across all CTV
Paid click fraudSpider AF H1 2026: 5.58% of 4.006B paid clicks, labeled as vendor dataImpression-based display or CTV
Bots on the internetImperva: over 53% automated, 40% bad bots in 2025Any ad fraud rate
Proof fraud is criminal, with datesDOJ: Methbot $7M+ and 3ve $29M+, indicted Nov 27, 2018; 10-year sentence Nov 10, 2021Current-year rates (historical)

If you plan budgets around these numbers, our media buying guide shows where fraud controls fit in the buying process.

Frequently asked questions

What is the average ad fraud rate?

There is no single average. In 2026 data, the ANA benchmark found 0.1% invalid traffic in audited programmatic spend, IAS found 0.7% fraud in protected campaigns and 10.9% in unprotected ones (2024), and Pixalate found 20% to 41% invalid traffic on open-auction web, CTV and app impressions in Q2 2026. The denominator decides the number.

How much money is lost to digital ad fraud each year?

The most cited estimate is Juniper Research's model: $84 billion, or 22% of online ad spend, lost in 2023, forecast to reach $172 billion by 2028. It was released by Fraud Blocker, a fraud software vendor. Spider AF separately estimated $25.3 billion exposed in the first half of 2026. Both are estimates, not audited losses.

Which channel has the highest ad fraud rate?

On open-auction programmatic, mobile apps. Pixalate measured 41% invalid traffic on mobile app impressions in Q2 2026, versus 26% for CTV and 20% for web, across more than 80 billion impressions. Rates on direct deals and verified buys are far lower, and DoubleVerify reports CTV fraud schemes up 140% in Q1 2026 year over year.

What percentage of web traffic is bots according to Imperva?

Imperva's 2025 Bad Bot Report put automated traffic at 51% of all web traffic in 2024, with bad bots at 37%, up from 32% in 2023. The 2026 report says bots made up over 53% of web traffic in 2025, with 40% malicious. These figures cover all web traffic, not ad impressions, so they are not an ad fraud rate.

What is the difference between GIVT and SIVT?

The Media Rating Council defines general invalid traffic (GIVT) as traffic caught by routine, list-based filtering, such as known bots, spiders and non-browser user agents. Sophisticated invalid traffic (SIVT) needs advanced analytics and human review to catch, such as hijacked devices, adware, malware and hidden or stacked ads. Ask any vendor which one its fraud rate covers.

How do I check for invalid clicks in Google Ads?

Add the Invalid clicks column: Campaigns, Columns icon, search Invalid clicks, Apply. Google defines invalid click rate as filtered clicks divided by filtered plus non-filtered clicks, and does not charge for them. Compare billed clicks with gclid hits in your server logs, and file the Click Quality form within 60 days if you see suspicious patterns.



About the author

Christoph Olivier Christoph Olivier is the founder of CO Consulting and a fractional CMO who has managed millions of dollars in ad spend and built a combined audience of over a million followers across social platforms. He works with 7- and 8-figure businesses, primarily in tax, M&A, consulting, real estate investing, capital raising, and financial services. His edge is a practitioner’s command of every major marketing channel, theory and execution, backed by the original marketing data reports he publishes here on CO Consulting.

Follow: YouTube · Instagram · LinkedIn

Work with CO Consulting

Get a free strategy call for your business

Share where you want to grow and what is in the way. We will review it and follow up to book a call with Christoph Olivier.

No guarantees, no hard sell. Your details stay private.

Free strategy call

Want marketing that brings in qualified clients?

Tell us about your business and goals. We will follow up to schedule a call with Christoph Olivier.

No guarantees, no hard sell. Just an honest look at what would move the needle.