Pay-per-view has one weak point, and everyone in the creator-marketing industry quietly knows it: the view count. When you pay creators per thousand views, the number on the screen becomes the number on the invoice — which makes it the single most attractive thing in the whole system to fake. The uncomfortable truth is that most platforms simply trust whatever figure a social network reports and pay against it. Bots love that.

We built Content Rewards the other way around. This post explains, openly, how we separate real human views from manufactured ones — because if you are funding a campaign, you deserve to know exactly what is protecting your budget. Most vendors treat this as a black box. We would rather show you the thinking.

Why the view count is the entire game

In a flat-fee creator deal, a fake view is a vanity problem. In a pay-per-view campaign, a fake view is a billing problem — it spends real money. That single difference is why fraud follows performance marketing everywhere it goes. If a view can be manufactured cheaply and it pays out at full price, someone will manufacture it at scale.

So the integrity of one number decides whether pay-per-view is a smart, efficient channel or an expensive leak. Everything below exists to protect that number.

We measure the shape, not just the size

Here is the core idea: a view count is a snapshot, and snapshots are easy to fake. What is far harder to fake is the shape of how a clip earns its views over time — its view velocity.

Real content grows in recognizable ways. It rides discovery: a clip gets surfaced, views arrive in bursts tied to how the algorithm is pushing it, engagement roughly tracks reach, and interest decays in familiar curves. Manufactured views tend to arrive detached from all of that — appearing in patterns that do not line up with how real human attention actually moves. You cannot see this in a single number. You can only see it by watching the clip over time.

Why every clip is measured on a fixed 75-minute cadence

Two words matter here: fixed and uniform. Every clip in the system is measured on the same 75-minute cadence, from the first campaign to the largest. We do this on purpose, and we do not make it adaptive.

Uniformity is fairness. Because every clip is observed the same way, on the same clock, one creator's views are directly comparable to another's — and no one can dodge scrutiny by being measured less often than the next person. A fixed cadence also builds a clean, evenly-spaced time-series for every clip, which is exactly what a growth-shape analysis needs to be trustworthy. Adaptive, irregular sampling would introduce blind spots; a steady heartbeat does not.

This cadence is one of the platform's core invariants. We scale it by adding measurement capacity, never by quietly measuring popular clips less often.

Turning signals into a score

The view-velocity time-series feeds a model that produces two things: a per-clip bot score and a per-creator trust score. The clip score asks, "does this clip's growth look like real attention?" The creator score asks, "does this person have a track record of clips that hold up?" Together they let genuinely trustworthy creators move fast while suspicious activity gets held back for a closer look.

We deliberately do not publish the exact signals, weights, or thresholds behind these scores. Doing so would simply hand a playbook to the people we are trying to screen out. But the philosophy is public and it is simple: the burden of proof sits on the views, not on the brand.

Flagged views do not get paid — before, not after

This is the most important design decision we made, so we will state it plainly: filtering happens before payout, not as a clawback afterward. Views identified as inflated or botted are removed from what earns money in the first place. A brand should not have to notice fraud, dispute it, and chase a refund. The default should be that fraudulent views never converted to a bill at all.

The layers around measurement

Real-view measurement is the foundation, but it does not stand alone. Around it we run:

  • Dispute crosscheck — conflicting claims about clips and views get resolved with evidence, not guesswork.
  • Chargeback protection — a funded campaign cannot be quietly undermined after the fact.
  • KYC and tax handling — you know who is being paid, and payouts stay compliant across countries.
  • Escrow flow — funds are held and released against verified performance, which protects the brand and the creator at the same time.

Layered together, these make a pay-per-view campaign safe to run at real scale instead of being a gamble on the honesty of strangers.

What we will not do — and where we are honest with you

We are early, and we would rather earn trust than overclaim it. No detection system on earth is perfect; adversaries adapt, and so we tune ours continuously. We will get individual calls wrong in both directions from time to time, and when we do, our disputes and review process is there to correct them.

What we will never do is pay creators on raw, unverified view counts and hope for the best. That is the one shortcut the entire industry takes, and it is the one we refuse. Refusing it is the whole point of the platform.

What this means for you

For brands and agencies, this is simple: your budget follows real human attention, not inflated dashboards. You can run performance creator campaigns without funding a fraudster's side income.

For honest creators, it means something just as important — you are not quietly undercut by people gaming the numbers, because gamed numbers do not pay. That is the marketplace we want to build: one where doing it the real way is also the way that wins.

FAQ

How do you actually detect fake views? We measure every clip on a fixed 75-minute cadence and analyze its view-velocity — the shape of how views accumulate over time — rather than trusting a single reported number. That time-series feeds a bot-detection and creator-trust model, and views that look manufactured are filtered out before any payout.

Why 75 minutes, and why keep it fixed? A fixed, uniform cadence makes every clip directly comparable and builds a clean, evenly-spaced record of how each clip grows. It also means no one can escape scrutiny by being measured less often. We scale by adding capacity, never by measuring less.

Do brands get charged for bot views and refunded later? No. Filtering happens before payout, not as a clawback. Views flagged as inflated never earn money in the first place, so you are not chasing refunds for fraud you should not have been billed for.

Isn't publishing this a gift to botters? We share the philosophy, not the exact signals, weights, or thresholds — those stay private precisely so they cannot be gamed. Transparency about how we think is good for brands; a step-by-step evasion guide would not be.