Key takeaways
- AdPix invalid traffic detection gives every paid click a verdict the moment it lands and scores every traffic slice (source, medium, channel, campaign, device) in a nightly run, using data the AdPix tag already collects.
- There is no binary fraud label: each click or slice is Clean, Review, Potential or Confirmed, and the invalid rate counts only Confirmed plus Potential.
- The system is built for precision: a source that converts at 2% or more, a slice under three sessions and plain direct traffic are not auto-flagged, and a shared IP or a VPN alone never flags anything.
- Monitoring (verdicts, sources, networks, settings) starts on Starter at 499,000 toman a month; blocking, Google Ads pushes and dispute evidence files are on Pro and AdPix 360.
- Start in Monitor mode for one to two weeks, import ad cost so wasted spend can be estimated, and allowlist your own office before you trust any number.
AdPix invalid traffic detection tells you how much of the traffic you pay for is not a real, interested visitor, and how much of your ad budget it cost. Every paid click gets a verdict within a fraction of a second of landing, every traffic slice is scored again in a nightly run, and the result is one of four tiers: Clean, Review, Potential or Confirmed. It reuses the data the AdPix tag already collects, so there is nothing new to install. Monitoring starts on the Starter plan, and blocking and refund evidence are on Pro.
Invalid traffic is a visit or click that cannot be taken as evidence that a real person was interested in your business, whether it came from a bot, a click farm, an agency’s placement checker or someone who clicked by accident.
What question does AdPix invalid traffic detection answer?
It answers “which part of my paid traffic should not count, and what did it cost me?” The Fraud Protection page reports the share of paid clicks (or sessions, when no paid clicks were recorded) that was invalid in the selected date range, the estimated spend that went with it, and which source, campaign or network it came from (AdPix docs: overview and severity tiers).
The word “invalid” is an accounting label, not an accusation. Your own uptime monitor, your QA bot and a person who left after two seconds are all invalid traffic with nobody acting in bad faith. A click farm burning your budget is invalid traffic too. AdPix puts both in the same column because intent cannot be recovered from data; what can be recovered is that the money went to something that will not become a customer (AdPix docs: what invalid traffic is).
Who is AdPix invalid traffic detection for?
It is for teams that pay per click or per impression and want a second opinion on the traffic their ad networks bill them for. Three situations make it worth turning on early:
- You buy display or native inventory. Placements on publisher sites carry more risk than search, and AdPix recognises Iranian networks such as Yektanet and Tapsell by source or referrer, even though they pass no click ID (AdPix docs: paid-click log).
- Competitors may click your ads. Local services and lead generation are the classic targets; AdPix ships stricter presets for both.
- Your conversion rate dropped without a reason. A new placement or a burst of fake visits can dilute every rate on your dashboard before anyone notices.
If you only run organic traffic, the manual checks in our bot traffic detection guide are usually enough, and they work in any analytics tool.
How does invalid traffic detection work in general?
The industry splits invalid traffic into two layers. The Media Rating Council, which sets measurement standards for ad verification, calls them general invalid traffic (GIVT) and sophisticated invalid traffic (SIVT) (MRC, 2020). GIVT is what a list catches: known crawlers, data-center addresses, proxies, Tor. SIVT is traffic that only its behaviour gives away, such as a real person in a real browser paid to click.
Ad networks run their own filters. Google says its traffic quality team uses automated filters, machine learning and manual review, removes invalid traffic detected before the end of the billing cycle from your metrics and invoice, and credits you when it finds invalid traffic after billing (Google Ads Help: about invalid traffic). That protection is real, but it is the seller grading its own goods. An analytics-side check sees what happened after the click: whether anyone scrolled, whether the same device came back from eight networks, whether a conversion fired ten seconds after landing.
How does AdPix detect invalid traffic?
AdPix detects invalid traffic in three passes: a real-time verdict on each paid click, browser signals a few seconds later, and a nightly analysis of traffic slices and networks. All three feed the same four tiers.
The real-time click verdict
A paid click is a page view that carries an ad click ID (gclid, gbraid, wbraid, msclkid, ttclid and similar), arrives from a known paid network, or has a paid utm_medium such as cpc or display. An fbclid on its own does not count, because Meta adds it to organic links too (AdPix docs). Each paid click is checked against counters per IPv4 address (per /64 range for IPv6). With the General preset:
| Reason | Raised when |
|---|---|
| Repeat clicks (10 min) | more than 3 paid clicks from one address in 10 minutes |
| Repeat clicks (7 days) | more than 7 from one address in 7 days |
| Repeat clicks (30 days) | more than 10 from one address in 30 days |
| IP rotation | one device arrives from more than 8 networks in 7 days (never for mobile or carrier addresses) |
| Click burst | more than 8 paid arrivals from one /24 network in an hour (never for carrier addresses) |
| Replayed click ID | the same ad click ID arrives again from another device or network |
| Datacenter IP, VPN / proxy / Tor, Known fraud network | the address or its network belongs to one of those classes |
The number and kind of reasons decide the tier. One reason means Review. Two reasons, or a data-center address on its own, mean Potential. A replayed click ID, a data-center address plus any other reason, or three reasons make it Confirmed. A data-center address alone stops at Potential because real users of iCloud Private Relay or WARP also exit from data-center ranges.
Here is how that plays out on a hypothetical Tuesday. An address sends four paid clicks to your sale page in eight minutes. The fourth click exceeds the allowance of three, so it gets Repeat clicks (10 min) and lands in Review. If that address also belongs to a hosting provider, the same click carries two reasons, one of them Datacenter IP, and becomes Confirmed. The Click evidence drawer shows the counters the verdict was taken on, so you can see the “4” against the threshold of 3 yourself.
Browser signals and shields
When the visitor has granted consent for the Marketing category, a small module loads after the first beacon and checks for automation (a browser that declares webdriver, a headless user agent, and weaker signs such as a zero-by-zero window) and counts pointer moves, scrolls and key presses in the first five seconds. Either of the two strong probes moves the click to Confirmed with the reason Automated browser. Without consent the signals are never sent, and their absence is never held against a visit (AdPix docs).
Shields decide before any counter. An address or device that converted on your site in the last 30 days stays Clean however often it clicks, and so does anything you allowlisted.
The nightly analysis of slices and networks
Roughly once a day AdPix scores traffic slices: one combination of source, medium, channel, campaign and device, such as google|cpc|Paid Search|spring-sale|mobile. That grain lets you say “this campaign is bad on mobile” instead of “paid is bad” (AdPix docs: why was this flagged). Structural signals include one fingerprint seen across five or more organisations on the AdPix network, more than half of sessions from data centres or proxies, eight or more sessions from one /24 in one hour with an interaction rate under 10%, and dwell times so uniform (coefficient of variation under 0.1 across 20 or more sessions) that no group of people would produce them. An ensemble of three anomaly detectors adds a model opinion, which never flags anything without a behavioural signal behind it.
A separate layer scores each network (ASN) within each channel, so one mobile carrier can be normal in organic traffic and suspect on a single paid channel.
What AdPix never auto-flags
These guards are hard rules, not settings (AdPix docs: overview):
- a slice that converts, with conversion measured by purchases, key events, session revenue or a source-level conversion rate of 2% or more;
- a slice under three sessions;
- a slice where at least half the sessions are engaged (10 seconds of dwell, two page views or a key event), when the only evidence is the model;
- direct traffic, unless a hard structural signal fires;
- a shared IP or a VPN on its own.
Key events are one of the signals that mark a slice as converting, so set them up before you judge the verdicts; the conversion tracking guide covers which actions to mark and how to count each order once.
A slice that converts but also carries a structural signature goes to Review, because a bot can fake key events.
How do you read the Fraud Protection page?
Read the headline strip first, then the trend, then the evidence. The page sits under the Fraud protection group in the sidebar and has seven tabs: Overview, Clicks, Sources, Networks, Blocks, Evidence and Settings.
The headline strip shows paid clicks, invalid clicks, the invalid rate, estimated wasted spend and active blocks. There is one definition of the invalid rate on the whole page: Confirmed plus Potential divided by all units in the range. Review sits in the denominator only. The risk badge beside the title reads Low under 8% of sessions flagged, Medium from 8% to 20%, High from 20% to 40%, and Critical at 40% or more (30% when at least one entity is Confirmed).
| Tier | What AdPix claims | What to do |
|---|---|---|
| Confirmed | several detectors agree and a structural signal corroborates them | exclude the network or placement and keep the evidence for a dispute |
| Potential | real evidence, short of certainty | investigate; cap the campaign’s budget before a bigger decision |
| Review | something showed up and AdPix did not rule on it | look at it; allowlist it or report a false positive if it is yours |
| Clean | no evidence | nothing |
On the trend chart, watch for a one-day jump rather than the level. A high, flat rate usually reflects your channel mix; a sudden jump means a new campaign, a new placement or a click farm that has just found you.
How is wasted spend estimated?
For each flagged channel-and-network pair, AdPix multiplies the channel’s spend in the range by that network’s share of all the channel’s sessions, sums the results, and caps the total at the cost you imported (AdPix docs: ad spend wasted). The denominator is all of the channel’s traffic, which keeps the number on the low side. Treat it as a cost to subtract before you compare channels, the same way the performance marketing guide works budget decisions back from clean cost per order.
A hypothetical example: your Paid Search channel spent $6,000 last month and brought 9,000 sessions. Network A, a hosting provider, was flagged with 450 of those sessions, which is 5%, so $6,000 × 0.05 = $300. Network B was flagged with 180 sessions, 2%, so $120. Estimated waste is $420, or 7% of paid spend. AdPix spreads spend evenly across sessions because it has no per-click price, so if the flagged clicks were your most expensive keywords, the true figure is higher. Turn any channel figure into CPC terms with the free calculator below.
Which AdPix plan includes invalid traffic detection?
Invalid traffic detection is included from the Starter plan up, at 499,000 toman a month or 4.9 million toman a year (AdPix pricing). The AdPix documentation splits it into two tiers (AdPix docs):
| Tier | Plan | What you get |
|---|---|---|
| Monitoring | Starter and up | Overview, Clicks, Sources, Networks, Settings, the allowlist, blocklist export and a verdict for every paid click |
| Enforcement | Pro and AdPix 360 | Blocks and Evidence tabs, Block mode, automatic pushes to ad platforms, the dispute evidence export |
Detection is identical in both tiers. On the Free plan the Fraud Protection page shows an upgrade card, and the documentation adds that the Home page of a Free site carries a card with the paid clicks confirmed invalid in the last 30 days.
AdPix is a web analytics platform that replaces Google Analytics 4: realtime, acquisition, engagement, conversion and e-commerce reports, plus attribution, lift testing and marketing mix modelling, with no data sampling and a free plan. The AdPix features overview shows how invalid traffic detection fits next to AdPix lift tests and MMM: clean traffic first, then measure what the spend caused.
How do you get started with AdPix invalid traffic detection?
- Put the website on Starter or higher and confirm the AdPix tag is live (the AdPix install guide shows how to check). Check the website picker on the page, since another website may be on the right plan.
- Tag paid links properly. Keep auto-tagging on in Google Ads and give Meta and Telegram campaigns a paid
utm_medium, or their clicks never enter the paid-click log. - Open Settings and pick an industry preset. Lead generation and Local services are stricter than General; B2B is looser because offices put many buyers behind one IP. Add your target countries (for example IR) so off-target clicks are checked (AdPix docs: protection settings).
- Allowlist your office, testers and agency by IP, CIDR range or device, instead of loosening thresholds for everyone.
- Import ad cost for the same range, or Est. wasted spend stays at zero.
- Stay in Monitor mode for one to two weeks, the page’s own advice. Open a few Confirmed and Review items and read their evidence.
- On Pro, switch to Block once you agree with the verdicts, and push the exclusion list to Google Ads. Google allows up to 500 excluded IP addresses per campaign (Google Ads Help), which is why AdPix shows the active block count against that limit.
Leave the sensitivity dial alone at first. A site that never saved a setting is scored Conservative, where only the top 10% of traffic by anomaly score can be flagged and two detectors must agree; Aggressive opens that to 35% with one detector (AdPix docs: sensitivity). Structural signals fire on every setting.
What are the limits, and when should you use something else?
AdPix is open about what this feature does not do:
- It is not an edge filter. The nightly scoring has up to about a day of latency, and blocks act in the ad platform, never on your site. If bots are overloading your server, you need a firewall or bot manager.
- It does not catch everything. Every threshold that keeps false positives low lets some real fraud through, and AdPix chooses that side on purpose.
- A verdict is not a refund. Google reviews raw click logs and does not act on another vendor’s verdict, so send the click quality report (timestamps, IPs, GCLIDs) rather than a summary (AdPix docs: evidence and refunds). Raw IPs are kept for 90 days.
- Browser signals need marketing consent. Visitors who decline get the click counters and nightly analysis only.
- Different currencies are not converted. Only rial and toman are reconciled; keep each market’s cost on its own website.
If your traffic is mostly organic, or you want to see the problem before paying for a tool, start with the manual routine in the bot traffic detection guide and compare your analytics with your server logs, as the GA4 side-by-side guide explains. Once invalid clicks are out of the numbers, recheck your break-even point with the ROAS formula guide, because a cleaner denominator often changes which campaigns look profitable. To see the effect on the site itself, compare the AdPix conversion funnel for the weeks before and after you exclude a placement: if the excluded clicks were invalid, entries to step one fall while purchases hold, so every step rate rises.
CPM · CPC · CTR calculator
Cost and click metrics for any campaign. The free plan covers 100,000 page views a month for each website, no credit card. Setup takes under five minutes.
Frequently asked questions
What does AdPix invalid traffic detection do?
It separates the traffic you pay for that cannot count as a real, interested visitor: bots, data-center and proxy traffic, click flooding, replayed click IDs and similar patterns. Each paid click gets a real-time verdict and each traffic slice gets a nightly verdict, and AdPix estimates how much ad spend went to flagged networks.
Does AdPix block fraudulent clicks in real time?
No visitor is ever turned away from your site. On Pro and AdPix 360, Block mode turns Confirmed verdicts into an exclusion list that AdPix can push to Google Ads or a webhook; the exclusion takes effect in the ad platform, not on your pages.
Which AdPix plan includes invalid traffic detection?
Invalid traffic detection is included from the Starter plan (499,000 toman a month) up, with the monitoring tabs and a verdict for every paid click. Blocking, automatic pushes and the dispute evidence export need Pro or AdPix 360. On the Free plan the Fraud Protection page shows an upgrade card.
Why did AdPix mark a source with zero conversions as clean?
Because AdPix never treats a missing conversion as evidence of fraud. Bad targeting and awareness campaigns also convert near zero; conversion is used only in the other direction, so a source that converts is treated as real.
Can I use AdPix verdicts to get a refund from Google Ads?
Google decides refunds on its own review of raw click logs, so send the click quality report with timestamps, IPs and GCLIDs rather than a summary of verdicts. AdPix keeps raw IPs for 90 days, so build any claim within that window.
Sources and methodology
Screen names, thresholds, tiers and plan tiers for Fraud Protection come from the official AdPix documentation at docs.adpix.io, read on 8 November 2026 and linked per section. Prices and plan placement come from adpix.io/pricing (read on 13 November 2026). Google Ads behavior comes from the Google Ads Help pages in the sources. The store, its spend, its sessions and its clicks in the worked examples are hypothetical and chosen to keep the arithmetic easy to follow.
- AdPix Docs: What invalid traffic is
- AdPix Docs: Overview and severity tiers
- AdPix Docs: Paid clicks and the evidence behind each one
- AdPix Docs: Why was this flagged
- AdPix Docs: Protection settings, mode, industry preset and thresholds
- AdPix Docs: Sensitivity and false positives
- AdPix Docs: Ad spend wasted, what the number actually is
- AdPix Docs: Evidence, exclusion lists and refund claims
- Google Ads Help: About invalid traffic
- Google Ads Help: Exclude IP addresses
- Media Rating Council: Invalid Traffic Guidelines update, public comment announcement (2020)
This article was drafted with AI assistance and reviewed by the AdPix team before publishing. Editorial policy