Definition
Invalid traffic covers any impression or click that does not represent a genuine opportunity for a person to see or act on an ad. Some of it is innocent: search-engine crawlers, monitoring services and prefetching browsers request pages without anyone looking. Some is careless, such as a placement that loads ads beneath a video player or inside a frame too small to see. Some is deliberate: hijacked devices, automated clicking and sites or apps that misrepresent who they are to the buyer. The term is broader than ad fraud, which is only the deliberate part, and the breadth matters because the remedies differ. Innocent and careless traffic is filtered; deliberate traffic is investigated, excluded and, where the contract allows, refunded.
General and sophisticated invalid traffic
The invalid traffic guidelines the Media Rating Council developed with industry bodies, which verification vendors and platforms reference, divide the problem by how hard it is to find.
- General invalid traffic (GIVT) can be identified by routine means: known data-centre addresses, declared crawlers and spiders, lists of known bad actors, and simple rules about activity no person could produce. Ad servers, exchanges and demand-side platforms remove it as standard practice, before or at the point of counting.
- Sophisticated invalid traffic (SIVT) imitates people. Malware that loads ads in the background, falsified site or app identities, stacked ads and manipulated measurement call for behavioural analysis, multiple signals or human review, which is the work accredited verification products are audited for.
The split describes detection, not seriousness. A large volume of general invalid traffic left unfiltered wastes as much money as a clever scheme.
In practice
Every system in the buying chain removes some invalid traffic with its own method and at its own stage, so the counts they report differ before anyone has done anything wrong. The ad serving basics guide lists invalid traffic filtering among the routine reasons a DSP, a publisher and the advertiser's ad server never agree on a total. Pre-bid controls in the DSP avoid inventory a verification partner scores as high-risk before money is spent; post-serve measurement tags the impressions that were bought and classifies them afterwards. Of the platforms mapped in the programmatic ecosystem guide, the verification vendor is the one whose job is to sit outside the transaction and count it independently.
Refunds follow from the contract, not from the report. A buyer who wants invalid impressions credited needs a named counting source, a named verification method and a clause excluding invalid traffic from billable delivery. The media auditing and verification guide places those terms within a wider audit of how a buy was delivered and billed.
Where it shows up
Invalid traffic is rarely spread evenly. It gathers where the payment model rewards volume that is cheap to fake:
- Click and traffic objectives, which reward any click. The traffic objective hub treats the share of bot and invalid traffic as a guardrail for exactly this reason.
- Unusually cheap open-auction inventory, especially long lists of apps and sites with little history.
- Placements with impossible engagement: plenty of clicks and almost no time on site, or visits that all arrive from one network in the same hours.
What buyers can require
Deal and insertion-order terms can name the independent counting source, require accredited measurement of invalid traffic, state that invalid impressions are not billable, and set how disputed volumes are credited. When deals are priced against each other, deducted volume belongs in the effective cost rather than being ignored, which is how the impression calculator tutorial for programmatic deals treats it.
Cheap clicks from one app network
Westerly Sailcloth runs a traffic campaign with a £6,000 monthly budget. One app network delivers 41,000 of the campaign's 52,000 clicks at £0.06 each, but those visits last under 2 seconds and produce no enquiries. The verification partner classifies 78% of that network's clicks as sophisticated invalid traffic: background app activity registering taps nobody made. Westerly excludes the network, the DSP credits £1,920 under the contract's invalid-traffic clause, and the budget moves to two sailing publishers whose clicks cost £0.40 but bring 35 enquiries in the following month. The campaign's reported click count falls by more than half; the number of people who actually reached the site barely changes.
Related terms
Every invalid ImpressionOne display of an ad or search listing, as counted by the platform. is still counted by somebody, which is why the term matters most when reports are reconciled. ViewabilityShare of measured ad impressions that had a real chance to be seen. answers a different question, whether a genuine ad had the chance to be seen, and the two are often measured by the same vendor. Brand SafetyControls that keep ads away from content that would harm the advertiser's reputation. concerns where an ad appeared rather than who saw it. An unusually low CPMCost per 1,000 impressions. is often the first hint of inventory with a high invalid share; the brand safety and viewability guide explains why made-for-advertising sites pass invalid-traffic filters while still wasting budget.
Common confusions
- IVT and ad fraud are not synonyms. Fraud is the deliberate subset; crawlers and accidental taps are invalid without being fraudulent.
- Filtered is not refunded. Traffic removed before billing never costs anything; traffic found afterwards is recovered only if the contract says so.
- Different counts are not proof of fraud. Each platform filters by its own method, so some gap between the DSP, the ad server and the verification report is expected.
- Real but uninterested visitors are not IVT. Made-for-advertising and arbitraged traffic involves real people; it is a quality and suitability problem that invalid-traffic filters do not address.
- A clean result on one measure says nothing about another. Valid impressions can still be unviewable or unsafe.
