What is Invalid Traffic (IVT)?
Also known as: IVT, Bot traffic, Non-human traffic (NHT)
What is invalid traffic?
Invalid traffic (IVT) is any ad impression or click not produced by a real human with genuine interest. The IAB Tech Lab IVT detection guidelines define IVT as the industry-standard umbrella covering bots, spiders, hijacked devices, and manipulated traffic. Most large platforms report 1 to 4 percent IVT after pre-bid filtering.
IVT does not require malicious intent. A search crawler hitting an ad is invalid. A QA script testing a page is invalid. So is a botnet faking 10 million impressions to drain a budget.
The label matters because billing depends on it. If traffic is flagged as IVT by an MRC-accredited vendor, advertisers can claw back the spend. If it slips through, the ad-fraud loss stays on the books.
GIVT vs SIVT
The IAB Tech Lab splits IVT into two tiers. The split decides which detection method applies and which vendor catches it.
| Tier | Full name | Sources | Detection difficulty |
|---|---|---|---|
| GIVT | General Invalid Traffic | Known bots, search spiders, data-center IPs, browsers on public block lists | Low. Pre-bid filters catch most. |
| SIVT | Sophisticated Invalid Traffic | Hijacked devices, malware-driven clicks, click farms, ad injection, domain spoofing, cookie stuffing | High. Needs behavioral analysis and continuous updates. |
GIVT is the floor. Every MRC-accredited platform filters it before billing. SIVT is where the money leaks. The MRC's IVT detection standards require accredited vendors to detect both, with SIVT carrying stricter audit rules.
A clean campaign report will still show some GIVT. SIVT closer to zero is the goal.
How is IVT detected?
IVT detection rests on signal patterns no genuine user produces. Vendors blend dozens of inputs per impression. The IAB Tech Lab IVT detection guidelines document over 30 signal categories used in 2026.
Network signals. Data-center IPs, residential proxy chains, anomalous ASN patterns, and known botnet ranges. Pre-bid filters lean here.
Device fingerprints. User-agent inconsistencies, missing canvas APIs, headless browser markers, and impossible screen-resolution combinations. Catches most automation.
Behavioral biometrics. Mouse entropy, scroll cadence, dwell time, and click coordinates. Real humans drift. Bots are too clean or too random.
Network graph analysis. Cross-publisher impression patterns. If one device ID hits 4,000 impressions across 200 unrelated sites in an hour, it is SIVT. Vendors like HUMAN run this at the bot-traffic network level.
Detection is probabilistic. No single signal proves IVT. Stacked signals do.
Which vendors detect IVT?
Four MRC-accredited vendors dominate IVT detection. Each plays slightly differently.
HUMAN Security. Born from White Ops. Strongest in bot mitigation and SIVT network analysis. Took down Methbot and 3ve. Best fit for advertisers worried about sophisticated botnets.
Integral Ad Science (IAS). Pre-bid and post-bid IVT, viewability, and brand safety in one stack. Heavy DSP integration. Reports IVT alongside brand-safety metrics on the same dashboard.
DoubleVerify (DV). Direct competitor to IAS. Strong on CTV fraud, SSAI manipulation, and authenticated-impression measurement. Public company, well-funded R&D.
Pixalate. Specializes in CTV, mobile in-app, and OTT IVT. Publishes free quarterly fraud benchmarks. Often used as a second-source verifier alongside IAS or DV.
Most enterprise advertisers run two vendors. One pre-bid, one post-bid. The overlap catches what either misses alone.
How do refunds and credits work?
IVT refunds run on the make-good model. The MRC requires accredited sellers to credit verified invalid impressions. The path to the credit depends on where the IVT was caught.
Pre-bid filtered. The advertiser was never billed. Nothing to refund. Platform reporting shows the filtered volume in a separate line.
Post-bid detected. The advertiser pays, then a third-party tag flags the impression. The advertiser submits a discrepancy report. The seller reconciles and issues a credit, usually within 30 to 90 days.
Disputed. Vendor reports disagree. The advertiser and seller compare methodology. Disputes go to MRC arbitration if the contract specifies it. Most settle bilaterally.
Contracts matter here. Without an IVT clause referencing MRC standards, sellers can refuse credits. Every IO and DSP master agreement should name the accredited measurement vendor and the dispute window. Advertisers also factor IVT into conversion-rate modelling so post-click metrics stay clean.
Real-world example
A US retail brand runs a 2 million dollar Q4 display campaign across open exchanges. DSP reporting shows 1.2 percent IVT. Their post-bid IAS tag reports 4.8 percent.
The gap is 3.6 percentage points. On 2 million dollars, that is 72,000 dollars of suspect spend. The brand's media team pulls the IAS report. Most of the discrepancy traces to three publishers. Each shows abnormal impression density in the 2 to 5 AM window with zero post-click activity.
The brand files make-good claims with the DSP. The DSP escalates to the supply-side platforms. Two of the three publishers issue full credits within 60 days. The third disputes. The brand drops the publisher from the inclusion list and adds it to its ad-revenue exclusion seed.
Net recovery: 51,000 dollars in credits. The bigger win was tightening the supply path for next year.
IVT in 2026
IVT is shifting faster than detection. Three trends define the year.
AI-driven SIVT. Headless browsers controlled by language models now mimic mouse paths, dwell times, and scroll behavior well enough to defeat legacy fingerprinting. The IAB Tech Lab added a generative-AI bot category in its 2026 guideline refresh.
CTV remains the soft target. Spoofed device IDs, server-side ad insertion abuse, and app-name manipulation push CTV IVT rates above display. Pixalate's quarterly benchmarks consistently report CTV IVT at 3 to 6 percent on open exchanges.
Privacy-driven signal loss. Apple Private Relay, third-party cookie deprecation, and Android Privacy Sandbox strip the IP and cookie signals detectors relied on. Vendors are rebuilding around server-side and on-device signals.
The upshot. IVT is not solved. It is moving into harder-to-see corners of the supply chain. Every advertiser running open-web or CTV media in 2026 needs an MRC-accredited measurement vendor, an IVT clause in every IO, and a quarterly audit. Skip any of the three and the unbilled fraud becomes a billed loss.
Related terms
Frequently asked questions
What is the difference between IVT and ad fraud?
Ad fraud requires intent to deceive. IVT is broader. It includes any non-human or non-genuine traffic, even when no fraudster profits. A search engine spider hitting an ad is IVT but not fraud. A botnet faking impressions is both. The MRC defines IVT as the umbrella term auditors use.
What is the difference between GIVT and SIVT?
GIVT is General Invalid Traffic. Known bots, spiders, and data-center IPs on public block lists. Easy to filter. SIVT is Sophisticated Invalid Traffic. Hijacked devices, malware, click farms, and adapted bots that pass filters. SIVT requires behavioral analysis and continuous detection updates from vendors like HUMAN or DoubleVerify.
Who certifies IVT detection vendors?
The Media Rating Council (MRC) audits and accredits IVT detection methods. Accreditation covers specific products and traffic types. HUMAN, Integral Ad Science, DoubleVerify, and Pixalate all hold MRC accreditation for IVT detection across display, video, and CTV. Buyers should check the MRC accreditation list before signing a vendor contract.
Do platforms refund IVT?
Yes, partially. Google, Meta, and the major DSPs filter known IVT pre-bid and never bill it. For IVT detected post-bid by a third party, advertisers file a make-good claim. The MRC requires accredited sellers to credit verified invalid impressions. Refund timelines run 30 to 90 days depending on the contract.
What is acceptable IVT rate for a campaign?
Industry benchmarks from IAS and DV put display IVT under 1 percent on premium inventory and 2 to 4 percent on open exchanges. CTV runs higher, often 3 to 6 percent due to spoofed device IDs. Anything above 5 percent on display warrants an audit. The MRC viewability standard requires IVT to be filtered before measurement.