Delivery logs are not attention measurement
An ad server or player can record that a video event fired. That is important technical evidence, but it is not a complete statement about actual perception. Depending on the device architecture, playback events and the real screen or usage context can diverge.
Completion rates need context
Very high completion rates can be plausible in CTV because formats may not be skippable and sessions can be long. They are therefore neither automatic proof of quality nor automatic evidence of fraud. They become meaningful only in combination with session duration, device, app, player and timing patterns.
CTV is not only a bot problem
Quality loss can occur without classic bots: technically valid delivery without meaningful attention, device misclassification, supply-chain opacity or reduced observability caused by server-side delivery and proprietary environments.
Forensics documents measurement limits
Independent client-side signals are often more restricted in CTV than in browsers. A serious analysis therefore states which signals were observable, which were only reported server-side and which conclusions cannot be supported.
A better KPI question
Instead of maximizing completion in isolation, advertisers should ask which deliveries had a plausible opportunity to be perceived, came from traceable inventory and showed a consistent usage context across multiple signals.
This article describes principles and questions. Specific fraud, bot or quality rates cannot be transferred directly to individual campaigns without understanding methodology, scope and data basis.