How the machine moves
12 animated diagrams. Each plays by itself when you scroll to it; hover to pause, or step through with the arrows.
The arbitrage loopOne visitor, start to finishWhere the advertiser’s dollar goesThree ways a search feed appearsThe feed supply chainClick ID out, revenue backFrom estimate to money in the bankAnatomy of a click-fraud schemeCatching invalid clicksFrom violation to terminationThe advertiser’s sideWhat AI answers change
The arbitrage loop
The whole business in one circle: buy a visitor cheaply in one place, earn more from that visitor’s click on a search ad, keep the difference.
The arbitrage loop
1/7
Two prices for the same person
A click on a social app might cost $0.20. A click from the same person on a search ad might be worth $1.00 to an advertiser. Search arbitrage lives in that gap, like a trader who buys fruit at the farm gate and sells it in the city (all figures here are illustrative).
- Two prices for the same person: A click on a social app might cost $0.20. A click from the same person on a search ad might be worth $1.00 to an advertiser. Search arbitrage lives in that gap, like a trader who buys fruit at the farm gate and sells it in the city (all figures here are illustrative).
- Buy a visitor: The arbitrageur runs an ad on a traffic source and pays for each click. Say the cost per click is $0.20. That money is spent whether or not the visitor ever earns anything back.
- The visitor lands on a page: The person who clicked arrives on the arbitrageur’s landing page: usually a short article with a block of related search topics underneath.
- A search shows sponsored results: If a topic interests them, the visitor taps it and sees a results page. The ads on it come from a search feed: the same advertisers who bid on a big search engine, shown on this smaller site.
- The advertiser pays for a click: The visitor clicks one sponsored listing and goes to the advertiser’s site. The advertiser is charged, say, $1.00. This is the only moment in the loop when new money enters.
- A share comes back: The search engine and the feed provider keep their cuts and the arbitrageur receives, say, $0.56 for that ad click: the RPC. But only about half of visitors click an ad, so the average visitor earns $0.28 (RPV).
- What is left is the spread: $0.28 earned minus $0.20 paid leaves $0.08 per visitor: the spread, a 40% ROI. Across 10,000 visitors a day that is $800. If the ad price rises to $0.30 or fewer people click, the same loop loses money just as fast.
One visitor, start to finish
Follow a single person from a social media ad, through an article and a related-search block, to an advertiser’s website.
One visitor, start to finish
1/7
An ad in the feed
Asha is scrolling a social app and sees an ad: “SUV lease deals compared for 2026”. The arbitrageur paid the app to show it. This is the paid traffic half of Social-to-search.
- An ad in the feed: Asha is scrolling a social app and sees an ad: “SUV lease deals compared for 2026”. The arbitrageur paid the app to show it. This is the paid traffic half of Social-to-search.
- She taps it: The tap costs the arbitrageur about $0.20 (illustrative). Asha lands on a content page: a real, readable article about leasing an SUV, on a site the arbitrageur runs.
- The article offers related searches: Inside the article sits a labelled block of related search terms such as “SUV lease deals near me”. This is Google’s RSOC unit. The page also tells Google the wording of the ad she came from (referrerAdCreative) so the terms can be checked against it.
- First click: choosing a topic: Asha taps one term. Nobody is paid for this click; it simply tells the system what she wants to search for. The share of visitors who do this is the lander CTR.
- A results page with ads: A search results page opens on the same site. Its sponsored listings are supplied by the search engine and marked as ads. Advertisers who bid on that phrase compete to appear here.
- Second click: the one that pays: Asha clicks a leasing company’s ad. The advertiser is charged, say, $1.00, and a share flows back to the arbitrageur. Two clicks, one paid: that is the two-click flow.
- The advertiser judges the result: If Asha asks for a quote, the advertiser got what it paid for: an advertiser conversion. If visitors like her rarely do, the engine learns that this traffic is worth less, and prices it down or cuts it off.
Where the advertiser’s dollar goes
Follow $1.00 of search ad spend from the advertiser through the search engine and feed provider to the arbitrageur, and on to the traffic source.
Where the advertiser’s dollar goes
1/7
An advertiser pays for a click
Someone clicks a sponsored listing on an arbitrage page. The advertiser’s account is charged $1.00. Every figure in this flow is an example: real shares differ by contract and most are confidential.
- An advertiser pays for a click: Someone clicks a sponsored listing on an arbitrage page. The advertiser’s account is charged $1.00. Every figure in this flow is an example: real shares differ by contract and most are confidential.
- The search engine takes its cut: The engine supplied the advertisers, the auction and the billing, so it keeps a slice first. Say it keeps $0.30 and passes on $0.70. What it pays out to partners is its traffic acquisition cost.
- The feed provider takes its cut: Most arbitrageurs do not contract with the engine directly. A feed provider does, and shares the feed onward for a revenue share. Say it keeps 20% of the $0.70: $0.14.
- What the arbitrageur receives: The arbitrageur gets $0.56 for that ad click, its RPC. Only about half of its visitors click an ad, so average revenue per visitor (RPV) is $0.28. That is its gross revenue.
- Most of it was already spent: Each visitor was bought from a traffic source for $0.20, paid up front. So the largest single share of the arbitrageur’s income goes straight back out to an ad platform.
- The spread: $0.28 in, $0.20 out: $0.08 per visitor is the spread, before staff, tools and content costs. Out of the advertiser’s dollar, the business that built the page keeps the thinnest slice and carries the most risk.
- Invalid clicks unwind the chain: If the engine later decides a click was an invalid click, it credits the advertiser and nobody downstream is paid for it. The arbitrageur sees that as a Clawback (revenue deduction), even though the $0.20 spent on the visitor is gone.
Three ways a search feed appears
A search box that returns ads (AFS), related searches inside an article (RSOC) and a parked domain (AFD): same advertisers, three different doors.
Three ways a search feed appears
1/8
One ad pool, three doors
A search feed lends a search engine’s advertisers to another website. Google has offered it in three main shapes. They differ in one thing: how the visitor’s interest (the search intent) is established before ads appear.
- One ad pool, three doors: A search feed lends a search engine’s advertisers to another website. Google has offered it in three main shapes. They differ in one thing: how the visitor’s interest (the search intent) is established before ads appear.
- Door 1: the visitor types a query: With AdSense for Search, a site has its own search box. The visitor types what they want and the results page carries search ads, drawn by Custom Search Ads code. Like a shop assistant answering the question you asked.
- One step from query to ad: Because the visitor wrote the query themselves, intent is clear and one ad click pays. Rules are strict about where the query came from: a query the visitor never typed or chose is not a real search.
- Door 2: the visitor reads first: With Related Search on Content, the visitor lands on an article. A labelled block suggests searches related to what the page is about. The page must have real content; the block is an add-on, not the whole page.
- Two clicks: topic, then ad: Tapping a suggested term opens a results page with ads for it; clicking an ad is what pays. This two-click flow is the format most paid-traffic arbitrage uses today, because the visitor’s own tap confirms the interest.
- Door 3: a domain with nothing on it: A parked domain is a web address with no real site. With AdSense for Domains, someone who mistyped an address or followed an old link saw a page of keyword links and ads guessed from the domain name alone.
- The door Google closed: Intent here was weakest: the visitor never asked for anything. Google Ads opted advertisers out of parked domains by default during 2025 (flipping the usual advertiser opt-out) and removed them from its search partner network on 10 February 2026.
- The clearer the intent, the safer the feed: Typed query, chosen topic, guessed topic: advertisers pay for intent, and the formats that prove it survive. That is why policy keeps pushing arbitrage toward real content and a genuine choice by the visitor.
The feed supply chain
How an advertiser’s search ad travels from Google, Bing or Yahoo through a syndication partner to a small publisher’s page, and who answers for it.
The feed supply chain
1/7
Advertisers bid at the search engines
An advertiser sets up keyword campaigns in Google Ads or Microsoft Advertising. Almost all the money in search arbitrage starts in these few advertising systems. Yahoo’s search ads have for years been supplied largely through Microsoft.
- Advertisers bid at the search engines: An Advertiser sets up keyword campaigns in Google Ads or Microsoft Advertising. Almost all the money in search arbitrage starts in these few advertising systems. Yahoo’s search ads have for years been supplied largely through Microsoft.
- Engines lend their ads out: Search engines also show those ads on other companies’ sites. This is search syndication, and the sites together form a search partner network. Think of a wholesaler stocking many small shops.
- The feed provider holds the contract: A direct feed is a contract with the engine itself. Few companies have one. A feed provider (or syndication partner) does, and is answerable to the engine for everything shown under its account.
- Sub-publishers plug in: The provider gives approved publishers a hosted feed in return for a share of revenue. Each publisher, site or campaign is tagged with a channel ID so the provider and the engine can see whose traffic is whose.
- The ad reaches a visitor: The sub-publisher brings visitors, usually with paid traffic, and the engine’s ads appear on its results pages. The visitor rarely knows four companies stand between them and the advertiser.
- Money flows back up the chain: A click charges the advertiser, say, $1.00. In our example the engine passes $0.70 to the provider, and the provider passes $0.56 to the sub-publisher. Each hop is a toll booth.
- Responsibility flows down it: The engine audits clicks and pages across the chain. One sub-publisher with bad traffic can put the provider’s whole contract at risk, so providers police feed approval, traffic sources and volumes closely.
Click ID out, revenue back
How a tracker joins a click bought on Meta to the revenue it later earned on a search feed, and tells Meta which clicks were worth buying.
Click ID out, revenue back
1/8
Two systems that never meet
Meta knows who clicked the ad and what it cost; the search feed knows what was earned; neither can see the other. Without a join, the buyer cannot tell which of 500 ads is profitable. A tracker is that join.
- Two systems that never meet: Meta knows who clicked the ad and what it cost; the search feed knows what was earned; neither can see the other. Without a join, the buyer cannot tell which of 500 ads is profitable. A Tracker (campaign tracking platform) is that join.
- A click ID goes out: When a visitor taps the ad, Meta adds a unique click ID to the link. The tracker records it with the campaign, ad and cost (say $0.20), like a cloakroom ticket stapled to the visit.
- The ID rides along: The tracker forwards the visitor to the lander and passes its own reference in a sub ID or channel ID. Whatever the visitor does on the feed is now stamped with that reference.
- An ad click earns revenue: The visitor clicks a sponsored result. The feed provider records about $0.56 against that sub ID. At this point it is only estimated revenue.
- Revenue comes back as a postback: The provider sends the figure to the tracker by Postback (server-to-server tracking) or reporting API. It can arrive hours late and is often reported per channel or sub ID, not per click, so trackers estimate the split. This lag is the revenue reporting delay.
- The tracker closes the loop: The tracker matches the reference to the original click: this ad cost $0.20 and earned $0.56. That is Attribution. It can now show profit by ad, keyword, device and hour.
- Meta is told what the click was worth: The tracker sends a conversion event with the click ID and the value to Meta through the Conversions API. No browser cookie is needed; one server talks to another.
- The algorithm hunts for more: With values attached, Meta’s system can favour people like those who earned money (value-based bidding). The catch: it learns from estimates. If revenue is later clawed back, the machine was trained on money that never arrived.
From estimate to money in the bank
Why the number on today’s dashboard is not what gets paid: estimated revenue, traffic-quality review, finalised revenue, clawback and Net payment terms.
From estimate to money in the bank
1/8
Money goes out first
Through March the arbitrageur buys 100,000 visitors at $0.20 each. The ad platform bills its card every few days: $20,000 of ad spend is gone before any income arrives. (An illustrative month.)
- Money goes out first: Through March the arbitrageur buys 100,000 visitors at $0.20 each. The ad platform bills its card every few days: $20,000 of ad spend is gone before any income arrives. (An illustrative month.)
- The dashboard shows an estimate: Each day the feed reports estimated revenue. By month end it reads $28,000, a paper profit of $8,000. It is a running tally, like a restaurant bill before the manager checks it.
- The engine reviews the clicks: After the month closes, the search engine’s systems finish checking for invalid clicks: bots, accidental taps, repeated clicks, traffic that broke policy. The publisher does not get an itemised list of what was removed.
- Advertisers are credited: Clicks judged invalid are credited back to the advertisers who paid for them. Say that comes to 5% of this publisher’s clicks. Nobody in the chain earns anything on a refunded click.
- The clawback: 5% of $28,000 is $1,400, deducted from the publisher: a Clawback (revenue deduction). The $20,000 spent buying those visitors is not refunded by anyone. A heavy clawback can turn a profitable month into a loss after the fact.
- Revenue is finalised: $28,000 minus $1,400 gives finalised revenue of $26,600. Real profit is $6,600, not $8,000. Google’s own AdSense timeline posts finalised earnings around the 3rd of the following month; feed providers set their own dates.
- Payment arrives on Net terms: The provider pays on Net terms, here Net 30: about 30 days after month end. Money spent on 1 March comes back around 30 April. Some contracts are Net 45 or Net 60.
- Growth eats cash: By the time March is paid, April’s $20,000 has also been spent. This cash-flow float means a growing arbitrageur needs working capital of one to two months’ spend, and a late payout or large clawback can sink a business that looks profitable.
Anatomy of a click-fraud scheme
How a dishonest publisher hides fake and junk visits inside a search feed, who ends up paying, and the signs that give it away.
Anatomy of a click-fraud scheme
1/7
Paid per click, so clicks get faked
A feed pays whenever a sponsored result is clicked. An honest publisher earns that by bringing interested people. A dishonest one looks for the cheapest thing that resembles a click, which is click fraud.
- Paid per click, so clicks get faked: A feed pays whenever a sponsored result is clicked. An honest publisher earns that by bringing interested people. A dishonest one looks for the cheapest thing that resembles a click, which is click fraud.
- Visits that cost almost nothing: The fraudster sources bot traffic, or people with no interest: pop-unders nobody asked for, or incentivised traffic paid a reward to click. A real interested visitor costs $0.20 in our example; these cost a fraction of a cent.
- Blended with real visitors: Sent alone, fake visits stand out. So they are mixed with a layer of genuine traffic until the averages look normal (traffic blending), or passed through other sites first to hide where they came from (traffic laundering).
- Searches and clicks nobody meant: On the page, the fake visits “search” and “click” sponsored results. Queries no human chose are query fraud. To the feed, each one arrives looking like an ordinary ad click.
- The advertiser pays for nothing: Each click bills an advertiser about $1.00. No customer arrives, so its money buys nothing. Honest publishers in the same feed suffer too: when partner traffic converts poorly, engines pay less for all of it (smart pricing).
- What gives it away: Reviewers look for clicks that never convert, visitors from data centres or a single device model, click rates that are too high and too steady, night-time activity as busy as daytime, and one-second sessions. No single sign proves fraud; together they form a pattern of invalid traffic.
- Refunds, clawbacks and a closed account: When the engine acts, invalid clicks are credited to advertisers, the publisher’s revenue is clawed back and the feed is usually terminated (feed suspension). The feed provider above it can lose its contract as well.
Catching invalid clicks
The two lines of defence in a search feed: filters that act before an ad is shown, and audits that judge a click after it happened.
Catching invalid clicks
1/7
A visit arrives
Every click on a search ad starts as a visit from some device. Most are people; some are scripts, crawlers or hijacked phones. IVT detection is the work of telling them apart.
- A visit arrives: Every click on a search ad starts as a visit from some device. Most are people; some are scripts, crawlers or hijacked phones. IVT detection is the work of telling them apart.
- Line 1: easy catches, before the ad: Pre-click filtering checks what can be known instantly: declared bots, data-centre addresses, impossible device details. This routine layer is GIVT. Blocked visits never see an ad, so no advertiser is charged.
- Deeper signals: Harder cases need more: device fingerprinting to spot one machine posing as many, signs of a headless browser, or a residential proxy hiding the true origin. Visits that pass go on to the page.
- An ad is clicked and logged: The visitor clicks a sponsored result. The click is recorded with its context: time, device, the query, how long the page had been open, where the pointer or finger was.
- Line 2: judged after the fact: Post-click analysis looks at what only shows up in bulk: clicks that never convert, identical timing, the same few devices returning. Disguised, deliberate fraud of this kind is SIVT.
- Invalid clicks are refunded: A click judged invalid is credited back to the advertiser and removed from publisher earnings. A valid click is billed as normal, say $1.00. Two outcomes from the same button press.
- Each verdict sharpens the filter: Findings roll up into a traffic quality score per source or placement. ClearTrust’s TQI Score™, for example, combines 150+ filters into one number. A publisher who scores its own traffic can cut a bad source before the engine does it for them.
From violation to termination
What typically happens after a search feed publisher breaks a rule: review, warning, traffic cap, clawback and, at the end of the road, loss of the feed.
From violation to termination
1/8
A feed comes with a rulebook
A search feed is lent, not owned. The engine’s AFS and RSOC policies, plus the provider’s contract, say what the ads, pages and traffic must look like. Compliance is the price of keeping the tap open.
- A feed comes with a rulebook: A search feed is lent, not owned. The engine’s AFS and RSOC policies, plus the provider’s contract, say what the ads, pages and traffic must look like. Compliance is the price of keeping the tap open.
- Something breaks a rule: Typical causes: a misleading ad that promises what the page does not deliver, search terms unrelated to the article, an unapproved traffic source, or wording that pushes people to click ads. Each is a policy violation.
- It gets noticed: Automated checks, manual reviewers and advertiser complaints all feed a review. The provider is watching too, because the engine holds it responsible for its publishers.
- First rung: a warning: For a first or minor problem the usual result is a notice naming the issue and a deadline. Fixing the ad or page, and showing it, normally ends the matter. This is the cheap exit.
- Second rung: limits: If issues repeat or quality looks weak, the feed may be throttled: a feed cap on daily volume, fewer ads per page, or a ban on one traffic source. Revenue falls immediately while ad spend may still be running.
- Third rung: revenue taken back: Earnings tied to the violating traffic can be withheld or deducted, sometimes for weeks already reported. If $28,000 was estimated and $8,000 of it is judged non-compliant, that Clawback (revenue deduction) alone wipes out the month’s profit in our example.
- Last rung: termination: Serious or repeated breaches end in the feed being switched off, often without a second chance, and unpaid balances may be kept (feed suspension). Severe cases can skip every earlier rung. This is platform risk at its plainest.
- The damage travels: A terminated publisher rarely gets another feed quickly: providers share the same few engines and ask about history. And a provider with too many bad publishers risks its own contract, which is why approval is slow and monitoring constant.
The advertiser’s side
How an advertiser’s search ad ends up on an arbitrage page through the search partners setting, how to see where it ran, and how to exclude sites.
The advertiser’s side
1/8
An advertiser buys a keyword
A car leasing firm creates a search campaign in Google Ads and bids on “SUV lease deals”. It expects its ad to appear when people search for that on Google.
- An advertiser buys a keyword: A car leasing firm creates a search campaign in Google Ads and bids on “SUV lease deals”. It expects its ad to appear when people search for that on Google.
- One setting widens the reach: Search campaigns include an option to show ads on Google’s search partners as well as Google.com. It is switched on unless the advertiser turns it off. It is all or nothing: partners cannot be picked one by one, only excluded.
- The ad appears on an arbitrage page: A visitor on a partner site taps a related search for “SUV lease deals”. The firm’s ad appears in the results there, and a click costs it about $1.00 (illustrative), the same budget as a click on Google.com.
- Does it convert?: What matters to the advertiser is the conversion rate: do these clicks become enquiries? Some partner traffic performs well. Where it converts less, Google says it lowers the price automatically (smart pricing), and invalid clicks are credited back.
- Seeing where the ads ran: For years partner sites were a black box. Since 2025 Google has reported the individual partner sites where Search, Shopping, App and Performance Max ads appeared, so an advertiser can see each Placement by name.
- Excluding what does not work: A site that spends money without results can be added to an account-level placement exclusion list, and ads stop showing there. The blunt alternative is the opt-out: switch search partners off for the campaign.
- One whole category removed: Google went further with parked domains: advertisers were opted out by default in 2025, and on 10 February 2026 parked domains stopped being part of the search partner network altogether.
- Why arbitrageurs should care: Every exclusion removes a bidder from that page, and fewer bidders mean lower prices. A page that sends advertisers real customers keeps its demand. A page that does not is slowly switched off, one exclusion list at a time.
What AI answers change
Search arbitrage depends on a chain of query, results and click. AI summaries shorten that chain, and the ad money is starting to move with it.
What AI answers change
1/8
The old path: query, list, click
For twenty years a search produced a page of links and ads (a SERP), and the searcher clicked through to a website. Search advertising, and search arbitrage with it, is built on that click.
- The old path: query, list, click: For twenty years a search produced a page of links and ads (a SERP), and the searcher clicked through to a website. Search advertising, and search arbitrage with it, is built on that click.
- An answer appears first: With AI Overviews and other generative search, many queries now return a written answer above the links. The searcher can read it and stop.
- Fewer clicks leave the page: A Pew Research Center study of US browsing in March 2025 found people clicked a result link on 8% of Google visits that showed an AI summary, against 15% without one. Google disputes the method. More searches end as a zero-click search.
- Ads move into the answer: The auction has not gone away. Google has placed ads in and around AI Overviews and in its AI Mode, so the same advertisers can appear inside the answer instead of beside the links.
- One budget, more places to spend it: An advertiser’s budget is finite. Each dollar spent inside an engine’s own AI answer is a dollar that did not need a partner site. Syndication exists because engines wanted extra reach; if they need less of it, feeds shrink.
- What it means for arbitrage: Arbitrage visitors mostly come from paid social and native ads, not from Google results, so the first click is less exposed than it is for sites living on organic traffic. The exposure is on the selling side: the feed is one supplier’s decision. That is concentration risk.
- Ads inside AI assistants: Chat assistants have begun carrying ads too; OpenAI started testing ads in ChatGPT in the US in early 2026 (ads in AI assistants). Whether any of them will syndicate ads to outside publishers, as search engines did, is an open question.
- Plan for a shorter chain: Nobody knows how far this goes. The safe reading for an operator: treat today’s feed economics as rented, keep costs flexible, and build pages people would value even with no ad on them. Platform risk is the business model’s main risk.