Lesson 1 of 8 · 8 min read · intermediate
What counts as invalid traffic in a search feed?
Invalid traffic is any click that does not reflect a real person's real interest. Learn Google's definition, the MRC's GIVT and SIVT split, and why feeds care.
In search arbitrage, money changes hands on one event: a person clicks a sponsored listing in a search feed. The advertiser pays the search engine, the search engine shares the money with the feed provider, and the provider shares it with the publisher who brought the visitor. Every link in that chain assumes the click came from a real person who actually wanted what the ad offered. When that is not true, the industry calls it invalid traffic, or IVT.
Picture a market stall that pays a boy a coin for every shopper he walks to the counter. If he brings his cousins, who look at nothing and leave, the stallholder has paid for nothing. If he brings people who were dragged there against their will, the same. And if he simply got confused and counted the postman, it is an honest mistake, but still not a shopper. All three are "invalid". Only some are cheating.
Google's definition, in plain words
Google runs the largest search feeds, so its wording sets the tone. On the advertiser side, Google Ads describes an invalid click as a click that does not represent genuine interest in the advertiser's business, and lists three kinds: accidental clicks caused by poor ad placement, automated clicks from bots or deceptive software, and manual clicks meant to run up an advertiser's costs or inflate a publisher's earnings. On the publisher side, AdSense defines invalid traffic as any clicks or impressions that may artificially inflate an advertiser's costs or a publisher's earnings.
Notice what the definition does not require. It does not require a bot. It does not require bad intent. A bored human paid to click, a real user tricked by a misleading ad, and a toddler mashing a phone screen can all produce invalid clicks. The test is whether the click reflects genuine interest, not whether a human finger was involved.
| Kind of invalid activity | Search-feed example | Deliberate? |
|---|---|---|
| Accidental click | Ads placed so close to a button that people tap them by mistake | Usually not, but the layout is the publisher's responsibility |
| Bot traffic | Scripts that load a lander, click a keyword and then click an ad | Yes |
| Self-clicking | A publisher or their staff clicking ads on their own pages | Yes |
| Incentivised click | Visitors rewarded with points, cash or unlocks for clicking | Yes |
| Deceived humans | A person who clicked because the ad promised something the page never offers | The user is innocent, the traffic buyer is not |
| Query fraud | Search queries the user never typed or chose, sent to the feed to pull high-paying ads | Yes |
GIVT and SIVT: the industry's two buckets
The formal vocabulary comes from the Media Rating Council, the US body that audits ad measurement. Its Invalid Traffic Detection and Filtration Standards, updated in June 2020, split IVT into two categories. The split is about how hard the traffic is to catch, not how harmful it is.
GIVT (general invalid traffic)
- Caught by routine filtration: lists and standard checks
- Known data-center traffic, declared crawlers and spiders
- Non-browser or unknown user agents, pre-fetched pages
- Simple activity rules, such as an impossible number of clicks in a minute
SIVT (sophisticated invalid traffic)
- Needs advanced analytics, multi-point corroboration or human investigation
- Bots that imitate humans, hijacked devices and sessions
- Adware and malware, hidden ads, ad stacking, domain spoofing
- Incentivised or manipulated human activity, falsified location or measurement
For a search feed, GIVT is the easy part. A crawler that announces itself or an IP address inside a known hosting range is filtered almost instantly. The problems that end feeds are nearly always SIVT: traffic that looks human on the surface and only shows its nature when someone studies patterns across thousands of clicks.
Why search feeds are stricter than display ads
A search ad click is expensive. An advertiser may pay several dollars for one click on a legal or insurance keyword because a searcher is assumed to be close to buying. That assumption is the whole value of search advertising. A syndicated feed borrows that value: the click on a partner site is billed through the same Google Ads or Microsoft Advertising account as a click on the search engine's own results page. If partner clicks convert worse, advertisers notice, complain, or switch off partner traffic with an advertiser opt-out.
That is why the search engines treat "low quality" as close to "invalid". A click can be made by a real person and still be worthless to the advertiser, for example when the person was nudged into clicking by a design trick. Google's policies put the burden on the publisher. Its AdSense help pages say plainly that it is the publisher's responsibility to make sure traffic to their ads is valid, even when a third party sent it.
Who pays when traffic is invalid
- The advertiser pays first, although Google says it filters invalid clicks from billing and issues credits when it finds them later.
- The publisher pays next. Filtered clicks earn nothing, and later deductions show up as the gap between estimated revenue and finalised revenue, known as a clawback.
- The feed provider carries the reputational risk. Too much bad traffic under its account can cost it the feed for every partner at once.
- The honest arbitrageur pays twice: the click was bought with real money from a traffic source, and the revenue that was meant to cover it disappears.
The rest of this track walks through the deliberate kinds one by one, then turns to how quality is scored, how you can detect problems early, and what the documented cases teach.
Key takeaways
- Invalid traffic is any click or search that does not reflect a real person's genuine interest, whether or not a bot was involved.
- Google lists accidental clicks, automated clicks and manual clicks meant to inflate costs or earnings as invalid.
- The MRC splits IVT into GIVT (caught by lists and rules) and SIVT (needs advanced analysis); feed-ending problems are usually SIVT.
- Publishers are held responsible for the traffic on their ads, even when a third party supplied it.
- Invalid traffic costs the arbitrageur twice: the click was paid for, and the revenue is removed.
Questions people ask
What is invalid traffic in search arbitrage?
Invalid traffic is any click, search or impression on a search feed that does not come from a real person with genuine interest in the ad. It covers accidental clicks, bots, paid or rewarded clickers, publishers clicking their own ads and visitors tricked into clicking. Search engines filter it, do not pay for it, and can deduct revenue later or close the feed.
What is the difference between GIVT and SIVT?
GIVT, general invalid traffic, is caught with routine checks such as lists of known crawlers and data-center addresses. SIVT, sophisticated invalid traffic, imitates real users and needs advanced analytics or human investigation to find, for example hijacked devices, adware or manipulated human clicks. The Media Rating Council defines both categories. The line is about detection difficulty, not about harm.
Does Google refund advertisers for invalid clicks?
Google says invalid clicks it detects are filtered out of campaign reports and billing automatically. When it identifies invalid clicks after charging for them, it issues a credit where appropriate and possible. In April 2026 Google also added an Invalid Activity Credit Report for Search and Performance Max campaigns, so advertisers can see those credits in the report editor.
Is a real human click always a valid click?
No. A click is valid only if it reflects genuine interest. A person paid to click, rewarded for clicking, clicking by accident, or lured by an ad that promised something the page does not deliver is human but not valid. Search engines judge the intent and the outcome for the advertiser, not just whether a person touched the screen.