Search Arbitrage, Explained by Someone Who Ran It

Search arbitrage is one of those businesses that is simple to describe and unforgiving to run. You buy attention on a social platform, you send that attention to a page that carries sponsored search results, and you get paid when someone clicks one of those results and lands on an advertiser's site. If what you get paid exceeds what you paid for the traffic, you have a business. If it does not, you have an expensive hobby. That is the entire model, and almost everything interesting about it lives in the gap between those two sentences.

I spent about two years doing this full time, starting on Snapchat, moving to TikTok, and later running Meta as well. What follows is the version I would give a smart friend who had never heard of it, including the parts that are usually left out of the recruitment pitch.

The Three Parties

There are always three parties in the chain, and understanding who pays whom removes most of the confusion.

The first is the traffic source: Meta, TikTok, Snapchat. You pay them to put your ad in front of people. This is your cost, and it is the only number in the entire model you control directly.

The second is the feed provider. These are the intermediaries that resell sponsored search inventory, mostly originating from Google, to publishers who can send it traffic. There are only a handful of them at any given time. They provide the search feed itself, the page technology that displays it, and the reporting dashboard. Critically, they also decide whether you are allowed in at all.

The third is the advertiser, whom you never meet. They are bidding to appear against a search term. When a visitor clicks through to their site, they pay the auction, the feed provider takes its cut, and what remains flows to you.

Why the middle party has all the leverage

The feed provider is the part of the chain outsiders underestimate. You cannot simply decide to enter this business the way you can decide to open a store. You have to be accepted by a provider, and providers are selective because they carry compliance risk for everything their publishers do. They will ask about your traffic sources, your ad creative, your spend capacity and your track record. A new operator with no history and no budget is a liability to them, not an opportunity.

This is the real barrier to entry, and it is a human one rather than a technical one. Everyone can build a landing page. Very few people can get a feed.

What Changed When RSOC Replaced AFD

For years the standard implementation was AFD - AdSense for Domains, the parked-domain product. The visitor clicked an ad and arrived on a page that was essentially a list of related search terms. They picked the term closest to what they wanted, that took them to a results page, and a click from there onto an advertiser paid out. It worked, but it was a thin experience: an ad, then a page of keywords, then results.

The replacement is RSOC. Google's own name for the product is related search for content, though the industry usually says related search on content and writes it RSOC either way. Under RSOC the visitor no longer lands on a bare list of terms. They land on an actual article. They read part of it, encounter a block of related search terms partway down, and continue reading below. The search unit is embedded in content rather than replacing it.

It is worth being precise about the sequence, because the industry shorthand of "RSOC arrived in early 2025" compresses two separate events. Google reclassified related search for content as an AdSense for Search product-integrated feature on 15 November 2024, which is what made it a first-class thing to build on. What actually emptied the old model out was economic and came a few months later: in February 2025 Google began opting advertisers out of showing ads on parked domains by default, which pulled the demand out from under AFD. Operators did not migrate because RSOC was new. They migrated because AFD stopped paying. Google finished the job on 10 February 2026, when parked domains ceased to be an ad surface within the Search Partner Network at all.

The practical consequence of the format change is that relevance became enforceable end to end. The ad has to match the article, the article has to match the search terms, and the search terms have to match what the visitor actually wanted. Under AFD you could get away with a loose connection between the ad and the page. Under RSOC a mismatch anywhere in that chain is visible to the platform, to the provider and to the visitor, and all three punish it.

That single change moved the skill requirement of the business from media buying toward editorial judgement. Writing the article became as consequential as writing the ad.

The rules tightened again after the migration

Anyone reading a description of RSOC written in early 2025 is reading about a permissive version of the product that no longer exists. Two changes since then matter more to a working operator than anything in the original transition.

The first is a tightening of the unit itself. Google added what it calls Restricted Access Features to related search in August 2025. These limits are not written up on Google's public policy change log, so treat the specifics as trade reporting rather than published policy: Domain Name Wire reported in November 2025 that publishers are held to five suggested search terms per ad block and one related search ad block per page, with constraints on styling and on the number of reporting channels available, and that Google lifts the restrictions case by case for publishers in good standing. If that is accurate, the ceiling on your page is now a function of your standing with Google rather than of your design decisions.

The second is documented by Google directly and is the more consequential of the two. Under a November 2025 policy update, publishers sending traffic to a related search for content page from any source under their control - including any third-party network, service or affiliate they partner with - are required to set a referrerAdCreative parameter to the precise and complete creative that produced the click. In practice the upstream ad has to be declared verbatim. The loose coupling between what an ad said and what the page behind it contained was the exploitable seam in the whole model, and it has been closed deliberately.

How the Money Actually Works

Your revenue is driven by revenue per click, usually shortened to RPC, and it is not a fixed number you can look up. It reflects the underlying advertiser auction, so it moves with the topic, the country, the device, the time of year and the general level of competition. Categories where advertisers have high customer value tend to pay more per click. Categories where they do not tend to pay less. The same campaign can earn very different amounts in two countries for reasons entirely outside your control.

Your cost is the price of a click on the ad platform, which is set by a different auction with different competitors. Neither auction knows the other exists. Your margin is simply the distance between them, and both ends move independently, daily.

Why the margin is thinner than it looks

Outsiders see the model and assume you find a spread once and then scale it. The reality is that the spread erodes constantly and from both directions. Audiences fatigue, so the same creative costs more each week to reach the same people. Competitors discover the same topic and bid the traffic price up. Advertiser demand in your category softens and the payout side drops. A campaign that is comfortably profitable in week one can be underwater in week three without you having changed anything.

The consequence is that this is not a passive business in any sense. It is closer to running a trading desk than running a website. The work is a continuous cycle of producing new creative, testing it, killing what has decayed and reallocating spend to what still holds. Stop working the cycle for a week and the account drifts to zero margin on its own.

What the Job Is Day to Day

Stripped of the framing, the daily work is narrower than people expect:

• Producing creative - short video and image ads, in volume, because most of them will not work and the ones that do will stop working. This is the single largest time cost in the business.
• Writing or commissioning the article - the RSOC page the traffic lands on, which has to be genuinely relevant to both the ad and the search terms beneath it.
• Watching the spread - comparing what you paid for clicks against what those clicks returned, at a granular enough level to act on before the day is lost.
• Cutting - turning off the majority of what you launch. Most operators find this psychologically harder than launching, and it is where most of the money is saved.

Notice that three of those four are content problems rather than technical ones. The tooling is provided; the judgement is not. If you are thinking about the creative production side, our guide to choosing a niche you will not regret covers the decision that sits upstream of all of it.

Why Most People Who Try This Lose Money

Three failure modes account for most of it.

The first is undercapitalisation. Finding a working combination of audience, creative and topic is a search process, and searches cost money before they pay. Operators who start with enough budget for a handful of tests will usually run out during the normal losing phase and conclude the model does not work.

The second is treating it as passive. The business rewards daily attention and punishes absence. People arriving from the affiliate world expecting to build something once and collect are describing a different business.

The third is relevance shortcuts. Under RSOC especially, running ads that oversell relative to the page is the fastest route to an account restriction. Platform enforcement is the largest uncontrolled risk in the model, and it is mostly self-inflicted. The operators who last are conservative about claims in a way that looks unambitious from the outside.

There is a fourth, quieter one worth naming: concentration. Your ad account can be restricted, your feed provider can drop you, or the payout in your category can fall. Any one of those can end the business overnight, and none of them are things you control. Anyone treating arbitrage income as stable income has misread the risk.

Is It Still Worth Understanding?

Even if you never run a single campaign, the model is worth understanding because it is an unusually clean demonstration of something that applies to any advertising-funded business: you are always buying one kind of attention and selling another, and your margin is the distance between two auctions you do not control.

The skills also transfer further than the niche does. Producing creative in volume, reading performance data honestly, killing things that are not working, and matching a message to a page are the core competencies of performance marketing generally. Search arbitrage just removes every other variable and leaves you alone with those four.

If you are evaluating the opportunity itself, the honest summary is this. The model works. The barrier is not technical, it is access to a feed provider, and that access is granted on the basis of a track record you do not have yet. The margins are real but they decay. And the whole thing rests on platform permissions that can be withdrawn without appeal. Go in with that understood and it is a legitimate business. Go in expecting passive income and it will be an expensive education.

Frequently Asked Questions

What is search arbitrage in simple terms?

You pay a social platform to send people to a page that carries sponsored search results. When a visitor clicks one of those results and lands on an advertiser's website, you earn a share of what that advertiser pays. Your profit is the difference between what you paid for the visit and what you earned from the click. Nobody buys a product at any point in the chain.

What is the difference between AFD and RSOC?

AFD is AdSense for Domains, the parked-domain product. A visitor clicked an ad and arrived on a page that was essentially a list of related search terms. RSOC - Google calls it related search for content - replaces that bare list with a real article carrying a block of search terms embedded partway down it. The distinction is no longer academic: Google began opting advertisers out of parked domains by default in February 2025, and parked domains stopped being an ad surface in the Search Partner Network entirely on 10 February 2026. AFD is not an alternative to RSOC now. It is gone.

Why is it hard to start in search arbitrage?

The barrier is access rather than technology. Sponsored search feeds come from a small number of intermediaries, and they vet publishers because they carry the compliance risk for whatever those publishers do. They ask for evidence of your traffic sources, your ad creative, your budget and your history. A newcomer with no track record and limited capital is usually declined, regardless of how good their pages are. Google added Restricted Access Features to related search in August 2025 as well; the reported limits are five suggested terms per ad block, one block per page and constrained styling and reporting, lifted case by case for publishers in good standing, so a new account starts capped.

Is search arbitrage passive income?

No. The spread between what traffic costs and what clicks pay erodes continuously as audiences fatigue and competitors enter. Campaigns that are profitable one week can be unprofitable two weeks later with no change on your side. Running it profitably means producing new creative constantly, testing it, and cutting most of what you launch. It resembles running a trading desk more than running a website.