Blog

Fake App Installs: 7 Signs Your Network Cheats

Lakshith Dinesh

Lakshith Dinesh

Head of Growth, Linkrunner

Last updated on 7 min read

Fake App Installs: 7 Signs Your Network Cheats

Short answer: a network is likely selling you fake installs when several of these show up together: installs that open seconds after the click, millions of clicks with almost no installs, near-zero day-one retention, a strange device or OS mix, installs from data centres or at odd hours, identical event patterns, and a sudden jump in volume while your organic installs fall. No single sign proves fraud. Three or more on the same network is a reason to pause payment and investigate.

Why does one network deliver installs at half the cost of every other channel, while none of those users ever buy anything? Sometimes it has found a cheap audience. Sometimes you are paying for installs a script, a device farm or a paid-to-install scheme produced.

AppsFlyer's State of Fraud for Marketers, 2026 edition, built on more than 100 billion installs across over 240k apps, says fraud rates held largely flat through 2025, and puts the fraud gap between affiliates and self-reporting networks at 36x.

Below: seven signs, why each happens and how to check it. For the detection rules behind each fraud type, read how mobile install fraud is detected.

What counts as a fake app install?

A fake app install is an install you are charged for that did not come from a real person choosing your app because of that network's ad.

That covers three different things:

  • Installs that never happened: a script sends fake install signals to your measurement tool (SDK spoofing).
  • Installs by devices, not people: bots, emulators and device farms installing and opening the app at scale.
  • Real installs with stolen credit: a genuine user who would have installed anyway (often organically), with a fake click fired at the last moment so the network gets paid.

The third type is the hardest to spot: the users are real, but you already had them.

Sign 1: Do installs open seconds after the click?

What it looks like: a big share of a network's installs open within a few seconds of the click.

Why it happens: a real user has to see the ad, read the store page and wait for the download. Seconds is too fast. On Android, a malicious app on the phone can detect a download starting and fire a click just in time to claim it (click injection).

How to check:

  • Plot click-to-install time for the suspect network against your other sources. Across Linkrunner data, the median real click-to-install time was about 2 minutes on Android and just under 40 minutes on iOS (2026 timing benchmarks).
  • Check whether the spike is Android only. Click injection does not work on iOS.

Do not over-read it. On one fintech app in 2026, about three quarters of a single affiliate's installs opened within 10 seconds of the click. Yet about half of the flagged installs went on to sign up, which points to users paid to install rather than fake devices. The same speed rule also caught WhatsApp and referral links opened by people who already had the app, so set speed limits per channel, not one number for everything.

Sign 2: Millions of clicks, almost no installs?

What it looks like: a network reports enormous click volume, but its click-to-install rate is a fraction of your other channels. Installs credited to it arrive spread evenly over days, with no peak soon after the click.

Why it happens: this is click spamming. The network fires clicks for devices that never saw an ad, hoping one lands inside the attribution window of someone who installs later for another reason.

How to check:

  • Compare clicks per install by network and by sub-publisher.
  • Look at the shape of the click-to-install curve. Real traffic peaks early and tails off; spammed traffic is flat.

What it looked like in Linkrunner data: two apps on the same affiliate network each received between 1M and 1.7M clicks over five weeks in autumn 2026, and only about 0.5% became installs. A few hundred thousand of those clicks came from a small group of cloud server IPs that hit both apps.

Sign 3: Do the installs never come back?

What it looks like: day-one retention near zero, and no sign-ups, purchases or other events after install.

Why it happens: bots and device farms install and open the app to trigger the payout, then stop. Incentivised users install for a reward and leave.

How to check: compare D1 and D7 retention and events per install by network and sub-publisher, not as a blended number. See how to measure app retention by channel. A source that is cheap on installs and empty on events is a red flag on its own.

Sign 4: Does the device mix look wrong?

What it looks like: a network's installs cluster on one or two device models, old OS versions, tablets in a phone-heavy market, or device languages and time zones that do not match the country you targeted.

Why it happens: device farms buy phones in bulk and emulators reuse the same profile, so the mix is far narrower than real users.

How to check: split installs by device model, OS version and country for the suspect network, then put the same split for your organic installs next to it. Real users look messy. Farmed traffic looks tidy.

Sign 5: Installs at 3am, or from data centres?

What it looks like: install volume that peaks at night local time, steady round-the-clock volume with no daily rhythm, or many installs from the same IP address.

Why it happens: scripts run on cloud servers and do not sleep. Real people do.

How to check:

  • Chart installs by hour for the network against your organic curve.
  • Count installs per IP. Heavy repetition from a few addresses is a flag.
  • Watch the click side too. In one case, a single cloud server sent about 4 clicks a second to a tracking link, a rate no person could produce.

Sign 6: Do events fire in identical patterns?

What it looks like: sign-ups landing exactly the same number of seconds after install, purchases with no session before them, or dozens of events stamped in the same second.

Why it happens: scripts that fake the SDK, or replay one recorded session, produce the same sequence every time. If you pay a network on sign-ups or purchases, fake in-app events are where the money goes.

How to check: pull raw event timestamps for a sample of installs from the network and look at the gaps. Real users vary. Also check whether the install and its click came from unrelated networks or countries.

Sign 7: A sudden surge while organic installs dip?

What it looks like: a network doubles its installs overnight at the same price, and your organic installs drop in the same week.

Why it happens: stealing credit for organic users is the easiest fraud to hide. AppsFlyer's 2026 report says organic traffic now accounts for 52% of all fraudulent installs, which is why poached organics need watching.

How to check: chart the network's daily installs on top of your organic installs. If one rises as the other falls, with no change to your budget, creatives or store listing, treat it as suspect. Also compare with the numbers your self-attributing networks report, since a gap there is not always fraud (see why attribution numbers differ).

The 7 signs at a glance

SignLikely causeFirst check
Opens seconds after clickClick injection, paid-to-installClick-to-install time by network
Huge clicks, tiny install rateClick spammingClicks per install by sub-publisher
No retention, no eventsBots, device farms, incentivised trafficD1 retention and events by source
Narrow device or OS mixDevice farms, emulatorsDevice and OS split vs organic
Odd hours, data-centre IPsScripts on serversInstalls by hour, installs per IP
Identical event timingSDK spoofing, fake eventsRaw event timestamps
Surge while organic dipsOrganic poachingNetwork vs organic daily trend

What should you ask the network?

Ask questions a clean network can answer easily:

  • Sub-publisher breakdown: which apps or sites did the traffic come from? Refusing to share sub-publisher IDs is a warning on its own.
  • Traffic type: is any of it incentivised traffic? Rewarded installs are not fraud if both sides agreed, but you should not pay a non-incentivised rate for them.
  • Click source: are clicks sent only on a real tap, or also on impressions and auto-redirects?
  • Their own filtering: what do they block before sending traffic, and do they remove flagged sub-publishers?
  • Contract terms: will they accept withheld postbacks as non-billable, and is there a clawback window for installs found to be fraudulent later?

Agree these terms before the next campaign. For dispute steps, see the click injection guide.

How does Linkrunner flag these signs?

Linkrunner, a mobile and web attribution platform, includes Fraud Control on every plan. It flags suspect traffic in real time using 8 rule templates:

  • Click injection
  • Install hijacking
  • SDK spoofing
  • SDK spoofing with IP mismatch
  • Click flooding by IP
  • Click flooding by device
  • Installs from fraudulent clicks
  • Reinstalls from fraudulent clicks

Each rule runs as Tag only, so you can study what it catches, or Tag and withhold postbacks, so the network does not receive the conversion for flagged installs. Linkrunner does not block installs before attribution; it flags them as they arrive and lets you decide what the network is told.

How the templates map to the signs:

  • Signs 1 and 2: click injection, install hijacking, click flooding by IP and by device, and the two "from fraudulent clicks" rules.
  • Signs 5 and 6: click flooding by IP covers repeat-IP clicks; the two SDK spoofing rules cover faked SDK traffic.
  • Signs 3, 4 and 7: these are quality and trend signals. Read them in your cohort retention, device and organic reports by network.

Start new rules on Tag only, then move to withholding one rule at a time. To compare tools, see the best mobile ad fraud prevention tools.

FAQ

How do I check whether I'm paying for fake installs?

Compare the suspect network with your other sources on click-to-install time, clicks per install, D1 retention, events and device mix. If it stands out on three or more, pause and ask for sub-publisher data.

What is click spamming and how do MMPs detect it?

Click spamming is firing masses of clicks for devices that never saw an ad, so one lands in a real install's window. MMPs detect it by huge click volume per IP or device with almost no installs, and a flat click-to-install curve.

Is a fast click-to-install time always fraud?

No. Paid-to-install users and people who already have the app (referral or WhatsApp links) also open fast. Treat it as a reason to investigate and set limits per channel.

How do I stop paying affiliates for bot and device-farm installs?

Flag the traffic, withhold postbacks for rules you trust, share evidence by sub-publisher, and agree in the contract that withheld installs are not billable.

The takeaway

One odd number is noise. One network failing several of these checks is a pattern worth stopping payment over. Run the checklist on your cheapest network this week, starting with click-to-install time and D1 retention. If you want Fraud Control running on your own traffic, request a demo from Linkrunner or read the Fraud Protection page.

Start measuring the installs your team cares about

Bring attribution, deep links, SKAN, cohorts, and campaign intelligence into one workflow your growth team can trust.