How IVT detection and filtration works
The industry standard is the Media Rating Council's (MRC) Invalid Traffic Detection and Filtration Standards. Invalid traffic filtration, usually shortened to IVT filtration, happens at two points. Up-front detection blocks known bad traffic early, for example before a bid. Back-end detection reviews traffic after the event and removes what it finds. The MRC says both have a place, and it requires known general invalid traffic to be removed from monetized counts.
Retail media networks apply this to sponsored ads. Amazon Ads says it blocks "a vast amount" of IVT before the bid, using signals such as sudden changes in traffic velocity, suspicious IP lists, and user agents shared by industry bodies. Most remaining IVT is invalidated in real time after the impression, and checks continue for up to 30 days. The result is a gross and a net number: Amazon reports clicks as gross clicks minus invalid clicks. Industry lists such as TAG's monthly Data Center IP List help companies filter routine cases.
General vs sophisticated invalid traffic (GIVT and SIVT)
The MRC sorts IVT into two categories by how hard it is to catch. General invalid traffic (GIVT) is the routine kind, and sophisticated invalid traffic (SIVT) takes deeper investigation. The standard treats them differently for accreditation:
| GIVT | SIVT | |
|---|---|---|
| How it is caught | Lists and standardized parameter checks | Advanced analytics, multi-point corroboration, human review |
| Examples | Known data-center traffic, known bots and crawlers, non-browser user agents | Hijacked devices, bots posing as real users, click farms |
| MRC status | Detection required for accreditation | Strongly encouraged; buyers are encouraged to add third-party coverage when a measurer checks only GIVT |
The boundary shifts over time: the MRC expects some SIVT techniques to become GIVT once industry bodies turn them into standard lists. Human activity can be invalid too, which is why click farms count as SIVT.
Invalid clicks on sponsored product ads: the Amazon example
Amazon Ads defines IVT as "any fraudulent, involuntary, non-human, duplicate, or otherwise illegitimate ad traffic". Its examples fit a retailer's site: crawlers that gather product listings and prices, bots built to target ads "with nefarious motives such as exhausting competitors' budget," and accidental double-clicks.
On a CPC (pay-per-click) sponsored product campaign, bots built to drain budgets are a direct cost, because Amazon warns that IVT its filters miss can be billed to the campaign. Amazon removes invalid events from invoices and advertiser reports once it detects them. Since September 2022, advertisers can pull a gross and invalid traffic report for Sponsored Products, Sponsored Brands, and Sponsored Display that shows invalid clicks and the invalid click rate.
Walmart Connect and Instacart hold MRC accreditation for their click metrics. Walmart Connect's Sponsored Search accreditation, announced in October 2025, includes GIVT detection and filtration. Instacart's net clicks are gross clicks filtered for invalid traffic, and they cover sponsored product ads.
Why closed-loop sales data helps expose invalid clicks
On a retail media network, the platform that counts the click also records whether a sale followed, so closed-loop attribution data can flag bot clicks that filters miss. Amazon Ads says that for Sponsored Ads, a drop in conversion rate, together with a significant fall in the invalid traffic rate and unusual rises in click-through rate, impressions, or clicks, is "a higher confidence indicator for undetected IVT". Criteo's retail media help center agrees: when IVT is suspected, start with conversion rate, which is likely to drop suddenly.
Coverage differs by placement. As of September 2026, Criteo's MRC accreditation, including MRC IVT filtering, covers onsite campaigns only. Offsite ads bought programmatically also lean on ads.txt, a public list of a publisher's authorized sellers, and on ad verification.
Why it matters for brands and sellers
IVT touches three decisions:
- Budget. IVT that filters miss can be billed, and Amazon names bots built to exhaust competitors' budgets.
- Optimization. IVT skews performance metrics, and invalid events can be removed for up to 30 days, so recent click and cost figures can still change. Revisit bid and keyword cuts once the data settles.
- Comparison. Each network filters with its own methods and scope: Walmart Connect's accreditation names GIVT filtration, and Criteo filters general and sophisticated IVT. Measurers differ too: for Amazon DSP campaigns, Amazon expects its invalid traffic rates to differ from third-party verifiers' by up to 5%, because filtration varies between implementations. Keep this in mind when you compare retail media network profiles.
A click spike with a conversion drop on one SKU or keyword can vanish inside a blended campaign total, so review these metrics per SKU. Our Amazon Ads guide covers how sponsored ads run on Amazon.
Related terms
- General invalid traffic (GIVT): IVT caught by lists and routine checks.
- Sophisticated invalid traffic (SIVT): IVT that needs advanced analytics or human review.
- Bot traffic: automated visits and clicks generated by software.
- Click farms: organized human clicking, a form of sophisticated IVT.
- Ad verification: third-party measurement that can add SIVT coverage.
- Brand safety: controls on where ads appear, the adjacent trust problem.
Frequently asked questions
What does invalid traffic (IVT) mean in advertising?
In advertising, invalid traffic (IVT) means impressions, clicks, and other ad events that should be excluded from measurement and billing, such as bot clicks, crawler visits, accidental double-clicks, and fraud. The Media Rating Council splits it into general and sophisticated IVT.
What is IVT filtration?
IVT filtration, or invalid traffic filtration, is the process of finding invalid impressions and clicks and removing them from counts and bills. The MRC standard combines up-front detection with back-end review. Amazon Ads blocks IVT before the bid, invalidates most of the rest in real time, and rechecks for up to 30 days.
What is the difference between GIVT and SIVT?
General invalid traffic (GIVT) is caught with lists and standard checks, such as known data-center traffic. Sophisticated invalid traffic (SIVT) needs advanced analytics or human review; examples include hijacked devices and click farms. Accredited measurers must filter GIVT, and SIVT detection is strongly encouraged.
Do advertisers pay for invalid clicks on retail media networks?
Policies vary by network. Amazon Ads removes invalid events from billing invoices and reports, including traffic it identifies up to 30 days later. Criteo's retail media platform excludes identified invalid traffic from reporting and billing. Check each network's documentation, since accreditation scope and filtering methods differ.
