What Is Invalid Traffic? GIVT, SIVT and the Ghost Devices Inflating Your Reach

In short: invalid traffic (IVT) is any ad impression or click with no genuine human behind it, split into easily-filtered GIVT and sophisticated, human-mimicking SIVT.

Invalid traffic (IVT) is the industry term for ad impressions and clicks that don’t represent a real person who could ever become a customer. It ranges from harmless crawlers to deliberate fraud, and the Media Rating Council splits it into two categories that matter enormously for how you defend against it.

The distinction is not academic. One category is a solved problem that every reputable platform handles for free; the other is where essentially all the lost budget sits. Vendors who blur the two can report an impressive-sounding filtration rate while catching nothing that was costing you money.

General Invalid Traffic (GIVT)

GIVT is the traffic that identifies itself, or is easy to identify: known search-engine crawlers, data-centre servers, bots on published lists, and obvious non-human patterns. It can be filtered with routine, list-based checks.

Crucially, most GIVT isn’t malicious at all. Googlebot crawling a publisher’s page generates an impression that no human saw, but nobody is defrauding anyone — it’s an artefact of how the web works. This is why GIVT rates can look alarming in isolation and mean very little.

Most reputable platforms strip GIVT automatically before you ever see a report. So if a vendor reports only GIVT, they’re catching the easy stuff — and charging you for work the platform already did.

Sophisticated Invalid Traffic (SIVT)

SIVT is the expensive problem. These are bots engineered to mimic human behaviour, hijacked real devices, domain and app spoofing, ad stacking, pixel stuffing and injected traffic.

What unites them is intent. SIVT is built specifically to pass the standard filters: it moves a cursor, it scrolls, it arrives from a residential IP address on a real consumer handset that someone else owns. Nothing in the bidstream distinguishes it from a customer, because it was engineered not to be distinguishable.

Catching it needs behavioural, device-level detection — evidence about how a device has actually behaved over time, rather than a list of things known to be bad.

Ghost devices and duplicate IDs

A growing slice of IVT isn’t “bots” at all in the classic sense. Every phone, smart TV, tablet, car and alarm system carries an advertising ID, and a single person can be linked to many of them.

Dormant, recycled and machine-to-machine devices inflate audiences with impressions no human will ever see. A handset sitting in a drawer still carries a valid ID. A recycled device can surface in an audience segment built around its previous owner. None of this is fraud, and all of it is reach on paper that doesn’t exist in the living room.

This category is expanding fastest, because the number of connected devices per household keeps rising while the number of humans in it does not.

Is invalid traffic the same as ad fraud?

Not quite, and the distinction cuts both ways. All ad fraud produces invalid traffic, but not all invalid traffic is fraud — a search crawler is invalid and entirely innocent.

More usefully for a budget holder: not all wasted spend is invalid traffic either. An impression served to a real, attentive human on a made-for-advertising page is perfectly valid and still worthless. IVT filtering is necessary, but it is not the whole of media quality.

What counts as a normal IVT rate?

There is no single benchmark, because rates vary enormously by channel, and because the figure depends entirely on whether a vendor is counting GIVT, SIVT or both.

For scale: Juniper Research puts invalid traffic at around 20.6% of programmatic ad traffic — roughly one impression in five. Connected TV runs worse, with invalid traffic reaching 18% globally and 24% in the UK in Q1 2025 (Pixalate). Treat any vendor quoting a flat industry average across all channels with suspicion; the variance between channels is larger than the average itself.

Why it matters

With automated traffic now past half of all web activity (Imperva, 2025), the gap between reported and real reach is wide enough to distort every downstream decision that follows from it.

Inflated reach makes frequency capping ineffective, because the cap is spread across devices that aren’t people. It corrupts lookalike modelling, because the seed audience contains machines. It flatters cost-per-thousand figures while raising true cost per human, and it makes channel comparisons meaningless when IVT rates differ sharply between them. Filtering to devices with recent, genuine human activity is what turns “impressions served” back into “people reached”.

Where IVT filtering stops and verification starts

Filtering removes what is demonstrably not human. Verification goes further: it confirms, positively, that a device was recently active with a real person in the location you targeted. The difference is between subtracting the known-bad and confirming the known-good, and it matters because SIVT is defined by not being known-bad yet.

What to ask your verification partner

In practice, the word to watch for is SIVT. Any competent vendor can filter GIVT; catching the sophisticated traffic engineered to look human is where budgets are genuinely protected or lost.

So when you assess a verification partner, don’t ask whether they “filter bots”. Ask specifically what SIVT rate they detect, how they detect it, whether they can show device-level evidence rather than an aggregate percentage, and — most importantly — whether they act before or after your money is spent. A report that arrives with the monthly reconciliation tells you what you lost. It doesn’t give any of it back.

If you want that measured on your own media, we’ll run it against a live campaign.