Reward fraud and how publishers stop it

Reward fraud is not a rounding error, because it attacks the exact number the business is calculated from: the verified completion. An install that never happened, a survey answered twice, an offer claimed on a device that never completed it - each one credits a reward and removes revenue that will never arrive. In 2026 the volume is large enough that a publisher who is not measuring it is guessing.

What fraud looks like in a rewards app

The economics are the reason it persists. A reward app pays a person for a verified completion, so the fraud is a machine that converts a fake completion into a real payout. If the payout is generous and the verification is thin, the machine runs, and it runs at scale because the marginal cost of one more fake install is close to zero.

The other reason it persists is that fraud is invisible in the short term. A fraudulent install looks like a successful one in the dashboard. It shows up later, as a completion rate that does not match the revenue, as an advertiser who does not renew, or as a network partner whose fraud rate nobody has checked.

Where fraudulent installs come from

The AppsFlyer State of Fraud for Marketers 2026 report, covering 106.4 billion installs across 246,000 apps between Q1 2025 and Q1 2026, gives the clearest public picture of the channels.

Organic traffic accounted for 52% of all fraudulent installs, the single largest fraud channel. Affiliates and organic together accounted for nine in ten fraudulent installs, and affiliates alone carried nearly 40% of all fraud. The gap between the fraud rates affiliates report and the rates they actually deliver reached 36 times in Q1 2026 - which is to say that the self-reported figure is not a measurement.

That distribution has a direct consequence for a publisher. The channels most likely to carry fraud are the ones a publisher is most inclined to trust, because they are self-reported, familiar, and billing against a real contract. The channel that looks safest - organic, where nobody is claiming credit - is the one carrying the largest single share.

The numbers moved, in both directions

Fraud rates are not stable, which is the argument for measuring them continuously rather than writing a policy once. AppsFlyer found that owned media fraud rose 221% year on year, from 3.4% to 11%, and DSP fraud rose 59%, from 5.6% to 8.9%.

On iOS, fraud fell 33% year on year, from 17.5% to 11.7%. In Finance, Android advertisers saw Real Users Lift locked between 50% and 53% for five straight quarters - roughly half of what they thought they were buying was not there. In Gambling, Android fraud moved from 49% to 59% and peaked at 64% in Q4 2025, with a Real Users Lift of 175% in that quarter.

The vertical detail matters because it changes what a publisher should do. A network whose fraud rate is roughly flat is a different problem from one whose fraud rate tripled in a year, and the second kind is usually a change in how that network sources traffic rather than a change in the publisher's own controls.

What a publisher can actually control

The device cannot be the authority on a completion. A claim is what the player says happened; a completion is a fact somebody else can confirm. That means validation has to happen server-side, against evidence the player does not control - the store's install receipt, the survey platform's response record, the advertiser's own attribution.

The third control is rate and pattern monitoring on the claim stream rather than on the app. Completion rate, credit velocity and the ratio of reversals to credits are the signals that move before the revenue does. A jump in reversals is an incident; a jump in fraud rate is a report.

Working with the people who send the traffic

Contracts help, but only if they contain the right clause. A commitment to a specific fraud rate, with the right to audit and to withhold payment for traffic above it, changes the incentive. A commitment to a general quality standard does not.

The honest position is that a publisher cannot eliminate fraud and should not try to make that the goal. The goal is to know the number, keep it inside what the economics tolerate, and make sure it is not moving quietly while everybody looks at the revenue instead.

Related reading

• Reward apps: what users get and how the economics work — The pillar this article sits under - where the revenue comes from and why verification decides it.
• Server-side reward validation for reward apps — Duplicate claims, silent reversals and what a defensible validation layer looks like.
• Rewarded surveys for publishers — Where survey walls fail, and why completion quality varies so much by source.
• Offer wall or interstitial for a rewards app? — The difference between a format the app chose and one the player chose, and what that costs in trust.
• Talk to us about reward validation — Tell us the platform and roughly how many daily actives you have, and we will say what fits.

Questions

← All articles