Which SourceKnowledge publishers send bots and invalid traffic?
Not all SourceKnowledge traffic is equal. ValidVisit scores every visit 0–100 and pins it to the exact publisher that sent it — so you can tell real humans from bots and invalid clicks, worst publishers first.
SubID in SourceKnowledgeThe buyer adds each bad SubID to a Block list (or Global Block List) — or an Allow list to whitelist only good ones — either one-by-one from the Sub ID Summary dropdown or in bulk via the Tools > Allow & Block Lists interface, and can alternatively bid down/pause a SubID or automate it with Rules.
How invalid traffic slips in through SourceKnowledge publishers.
SourceKnowledge places native and display units across a publisher network built for performance buyers, and every click it sends carries a {clickid} identifying the click itself, a {subid} naming the sub-source behind it, and an {oadest} for the destination the unit pointed at. What it does not send is anything about the visitor. Where most networks pass back at least an operating system, a browser family or a connection address somewhere in their reporting, SourceKnowledge’s click parameters expose none of it — your view of who arrived is limited to which sub-source sent them. That gap is the whole difficulty of judging quality here: with no platform-side device or network data to compare against, one sub-source’s traffic looks exactly like another’s until something measures it independently. ValidVisit is that independent measurement. Every inbound visit is scored against 100+ data points covering the network the click arrived from, the device sitting behind it, and how the session behaves once the page loads, resolving to a single 0–100 quality score for that one visit. Because the score attaches to the {subid} that produced it, the dimension you can measure is exactly the dimension SourceKnowledge lets you block. On a network that tells you nothing about the visitor, a per-visit score is not a second opinion — it is the only opinion available.
The invalid-traffic patterns on SourceKnowledge are shaped less by the native and display formats it sells than by how little the platform reveals about a sub-source. The first consequence is that a {subid} is opaque by construction: it identifies the sub-source that delivered the click, but not the domain, the placement or the method by which those visitors were acquired. You cannot inspect a bad sub-source and see what is wrong with it, which means the only route to a verdict is to measure the visits it sends and let the pattern in the scores stand in for the site you cannot see. Sub-sources that are quietly reselling acquired traffic are indistinguishable from direct publishers in the reporting; they are readily distinguishable in the scores.
The second is that every device claim arrives unverified. On a network that reports an OS and browser family, a mismatch between the declared profile and the observed one is a cross-check you can run for free. Here there is no declared profile at all, so the entire burden falls on first-hand measurement: what the browser announces about itself versus how it actually renders, times and handles the page. That is not a weaker position than it sounds — the announcement is the part an automated session controls, and the behaviour is the part it struggles to fake — but it does mean an assessment of a SourceKnowledge sub-source depends completely on your own measurement rather than on reconciling two sources.
The third pattern is acquired-audience inflation, common wherever native inventory is resold. A sub-source buys visitors cheaply somewhere else, lands them on pages carrying your unit, and a share of the resulting clicks come from people with no interest in the offer — or from automated sessions that exist to produce click volume. These arrive in real browsers on real URLs, which is why address-based filtering alone rarely separates them. ValidVisit reads the network origin of each click as part of the score, and that is usually where the pattern surfaces first: sub-sources running it tend to route through residential proxy pools to disguise a datacenter origin, and the routing marks the wider signal set before any single filter reacts.
What to watch on SourceKnowledge
{subid}Segment your ValidVisit report by the {subid} token. Because it is the only dimension SourceKnowledge exposes and the exact unit its block lists operate on, it is where nearly every actionable finding will surface. A sub-source carrying a disproportionate share of your click volume alongside scores well below your campaign baseline is the primary candidate for a block list, and there is no finer grain to fall back on.
For each weak {subid}, look at whether the poor scores trace mainly to where the clicks came from or to what the sessions did after landing. A sub-source whose weakness sits almost entirely in its traffic origin has a structural sourcing problem and belongs on a block list. One whose weakness appears mostly in on-page behaviour is more likely a smaller automated operation, which a bid reduction may absorb without giving up the volume.
{subid} over timeA sub-source that scores well for weeks and then degrades sharply has usually changed how it acquires traffic, not what it is. Because you cannot see the underlying sites, the score trend is your only early warning that a previously good sub-source has started reselling. Reviewing the trend before blocking also keeps you from cutting a sub-source over a single bad day.
SourceKnowledge supports allow lists as well as block lists. If your scores show quality concentrated in a small number of sub-sources while a long tail contributes weak volume, whitelisting the good ones is fewer decisions and holds its shape as new sub-sources enter the network — whereas a block list needs extending every time one appears. The score distribution across your {subid} values is what tells you which side of that trade you are on.
Pinpoint the bot publishers & placements in SourceKnowledge.
SourceKnowledge itself isn’t the problem — bots and invalid traffic concentrate in a handful of its sub-sources: the publisher, site or zone, and the placement or widget within it. So we roll the score up by those SourceKnowledge tokens, not by creative (which says nothing about whether a click was human).
Bought as one SourceKnowledge line, a buy reads as a single number. Scored per sub-source, a spread like this illustration runs from 82 down to 23 — the worst is nearly all bots. That’s the leak a blended average hides.
Illustrative: SourceKnowledge traffic scored 0–100 per sub-source, worst first — down to the placement you buy.
Bot / invalid-traffic score broken down by:
- Sub/Source ID
{subid}
Per-click id: SourceKnowledge passes a unique click id, so we also run velocity, deduplication and repeat-source checks on every visit.
Compare bot & invalid-traffic breakdown across every ad network →See your own SourceKnowledge sub-sources scored this way.
How ValidVisit attributes SourceKnowledge traffic
Each SourceKnowledge macro maps to a normalized parameter, so every scored click is pinned to the right campaign, creative and publisher.
https://yoursite.com/landing?utm_source=sourceknowledge&utm_medium=native&vv_click_id={clickid}&vv_publisher_id={subid}&vv_ad_id={oadest}| Token | SourceKnowledge macro | Maps to | Identifies |
|---|---|---|---|
| Click ID | {clickid} | click_id | click |
| Sub/Source ID | {subid} | publisher_id | publisher |
| Destination / Ad | {oadest} | ad_id | ad |
- SourceKnowledge macro
{clickid}- Maps to
- click_id
- Identifies
- click
- SourceKnowledge macro
{subid}- Maps to
- publisher_id
- Identifies
- publisher
- SourceKnowledge macro
{oadest}- Maps to
- ad_id
- Identifies
- ad
How the detection works.
Data points → one score
Every visit is weighed against more than a hundred independent data points and reduced to a single, sortable 0–100 quality score.
Many angles, combined
Each data point is combined rather than checked in isolation, so a genuine human almost never trips enough of them to be flagged — and bots that beat one rarely beat the rest.
Proprietary, not a black box
The detection model is ours and stays that way. What you get is a clear verdict on every visit — not a single brittle rule you can game, and not an unexplained number you can’t act on.
Pinned to the source
Every verdict maps to the campaign, publisher and placement that sent the click — so you know exactly which source to cut.
How ValidVisit helps you cut fraud and bad SubIDs on SourceKnowledge.
Scoring and attribution are the means — the point is cutting the SourceKnowledge traffic that wastes your spend. Here’s how ValidVisit gets you a list you can act on.
- Score
See what’s actually landing
You buy SourceKnowledge clicks; what arrives are visits. ValidVisit scores each one 0–100 so real humans stand out from bots and invalid traffic — one script, no funnel hop, no fingerprinting.
- Attribute
Pin the fraud to its source
Every scored visit is tied to the exact SourceKnowledge SubID and zone via the network’s own tokens — so the bad traffic has an address, not just a headline percentage.
- Cut
Take the SubIDs off your buy
You get the worst offenders as a ready-to-use list plus postbacks to your tracker — so you can exclude them in SourceKnowledge and put your next dollar behind the traffic that converts.
SourceKnowledge traffic quality, answered.
Does ValidVisit block SourceKnowledge sub-sources automatically, or is it manual?
It is manual. ValidVisit scores every visit and surfaces the {subid} values carrying weak quality scores in its reports and dashboard. You apply the block yourself under Tools > Allow & Block Lists, where you can create a list and associate it with specific campaigns, or add a sub-source to a Global Block List that applies across every campaign on the account. You can also act inline: each SubID in a campaign’s Sub ID or Channel Summary report has a dropdown with an "Add to list" option. There is no automated push from ValidVisit into SourceKnowledge. The workflow is: score in ValidVisit, identify the problem {subid}, block it there — which keeps the decision yours and avoids cutting a sub-source over an isolated spike.
SourceKnowledge doesn’t report device, OS or connection data. Does that limit what ValidVisit can score?
No, because ValidVisit does not read those fields from the network in the first place. The score is built from what the pixel measures directly on your page — the characteristics of the connection the visit arrived on, the consistency of the device presenting itself, and how the session actually behaves — so it does not depend on the ad platform reporting anything. What the missing data costs you is a cross-check, not the assessment: on a network that declares an OS and browser, a disagreement between the declared profile and the observed one is an extra signal. Here there is nothing to disagree with, so the measurement stands alone. In practice that changes very little, since the declared profile is the part an automated session can simply assert.
Which SourceKnowledge tokens give the most useful segmentation in ValidVisit?
The {subid} does the real work: it names the sub-source behind each click and it is the same unit SourceKnowledge’s allow and block lists operate on, so ValidVisit’s scores line up against the control you actually have without a translation step. The {clickid} gives per-click identity, ties each scored visit back to a specific SourceKnowledge click record, and is what carries a server-to-server conversion postback. Adding {oadest} lets you check whether a weak pattern spans the sub-source or is confined to one destination being promoted there — which is the difference between blocking the sub-source and changing what you send it to.
Detect fraud on other native networks
AdNow
Adskeeper
KadamFind the bots in your SourceKnowledge spend.
See which publishers and placements send real buyers vs bots — every visit scored 0–100, worst first.
Free trial at launch · just your email
One script · no cookies · no fingerprinting · raw IP never stored