Bot Traffic vs Invalid Traffic vs Ad Fraud: A Practical Measurement Taxonomy
Bot traffic describes automation; invalid traffic describes activity excluded under a measurement definition; ad fraud implies manipulation. These categories overlap but are not synonyms. Name the event, purpose, measurement rule and evidence before reporting a percentage or turning a bot observation into an allegation of fraud.
Quick answer
Bot traffic describes automation; invalid traffic describes activity excluded under a measurement definition; ad fraud implies manipulation. These categories overlap but are not synonyms. Name the event, purpose, measurement rule and evidence before reporting a percentage or turning a bot observation into an allegation of fraud.
For marketing, analytics, media buying and security teams reviewing paid traffic quality. This guide is technical and operational information, not a legal finding about a specific source.
Bot traffic
Automation can fetch resources, monitor a service or submit actions without the same process as an ordinary browser user. A bot observation says little by itself about authorisation or intent. State how automation was identified and which requests or sessions were observed; do not extend a limited sample to all billable media activity.
Legitimate automation
Search crawlers, uptime checks and authorised service agents can perform useful work. Their requests need not be a business enquiry or a valid billable advertising interaction. A site may permit an automated request while excluding it from a particular audience or campaign metric. Access policy and advertising measurement answer different questions.
Invalid traffic
MRC's IVT guidance concerns activity that does not meet the quality or completeness requirements for the relevant measurement. Google also includes accidental and automated advertising interactions in its invalid-traffic handling. The exact denominator and filtration scope matter: a billed click, page request and rejected lead are not interchangeable units.
GIVT
General Invalid Traffic refers to categories addressed through routine filtration, such as recognised non-human sources and common list- or parameter-based techniques. It is not a synonym for all activity originating from a data centre. Mixed infrastructure and legitimate routing need context, and a provider must document the rule used for its metric.
Sophisticated invalid traffic
SIVT concerns activity requiring more advanced analysis or corroboration than routine filtration. MRC examples include disguised automation and manipulated measurement or inventory. Complexity describes the measurement challenge, not proof that every unusual session is fraud. Some findings may be determined after initial reporting; disclose that timing.
Suspected fraud
A repeatable pattern may support a hypothesis of deliberate manipulation while leaving actor and motive unresolved. Label that uncertainty and name competing explanations. A risk signal can justify a proportionate operational review without becoming a confirmed commercial or legal allegation against a named party.
Demonstrated manipulation
Use stronger language only when the evidence demonstrates the mechanism being asserted, such as fabricated events or a documented attribution replay. Preserve source records and their limitations. Proof of a manipulated event still may not identify its operator or settle advertising liability.
Why terminology matters in reporting
If a report combines routine crawler exclusions, suspicious sessions and sales rejections into one fraud rate, the resulting number has no consistent meaning. State the numerator, denominator, period, observed population and review status. Keep missing or unmatched evidence visible rather than counting it as legitimate by default.
Examples
These are illustrative classification boundaries, not client case studies or universal platform decisions. Apply the actual measurement rules and evidence to the case being reviewed.
| Observation | Supported interpretation | Not established by this alone |
|---|---|---|
| Recognised crawler fetch | Automation; possible exclusion from a defined metric | Malicious intent or a billed fraudulent click |
| Repeated shared-IP sessions | A network pattern to investigate | One actor, automation or competitor identity |
| Rejected sales enquiry | An outcome under the business's qualification rule | Invalid ad billing or a bot |
| Verified replayed conversion event | A measurement-integrity defect or manipulation, depending on evidence | Actor identity or guaranteed refund |
Recommended language for audit reports
Prefer statements such as observed automation, excluded under the stated IVT rule, suspected manipulation, or inconclusive due to missing identifiers. Attach the basis and scope to the finding. Using MRC terminology does not establish MRC accreditation or prove that AdFraud.pl implements every GIVT/SIVT method.
Check which definition a published benchmark actually measuresAgree the measurement question and reporting limits for an audit
Related reading
Sources and further verification
Platform rules, standards and detection practices change. Check the current primary source and your own measurement scope before applying a control.