Affiliate SEO in highly competitive niches stopped looking like conventional SEO a long time ago. You don’t work on one project for years; you launch batches of sites, count how many survive, and cover losses with profits from the ones that gain traction. In terms of cost structure and mindset, it’s closer to traffic arbitrage than agency work. Below is a look at how this segment works from the inside: its economics, tools, risks, and the lessons that can be applied to white-hat projects.
The unit of work is a batch, not a site
The key difference from conventional thinking is that a site here isn’t a project—it’s a lottery ticket. According to teams that work at scale, out of a hundred sites launched, seven or eight make it into the search results. The rest remain buried in the third hundred of results or never get indexed at all, and there’s no set of parameters that can predict in advance which ones will make it.
This leads to a rule that sounds counterintuitive to someone with a white-hat background: investing in an individual site before it shows results is pointless. First, you launch a batch; then you see what has gained traction on its own, even if it’s only in the top thirty, and put more budget into those sites alone. The opposite approach—putting all your resources into one “perfect” project—regularly ends with the entire investment lost: competitors in these niches know how to knock out standalone sites before they have a chance to pay for themselves.
Practitioners estimate the minimum number needed for the statistics to mean anything at several dozen sites. Fewer than fifty launches isn’t a large enough sample to tell a failed approach from bad luck.
What it costs to get started
The cost structure has changed dramatically over the past two years—and the change has been in one line item: development.
- Site. Generating a template-based landing page or small site with an LLM agent costs about a dollar in tokens. You don’t need a dedicated developer at the outset; you need someone to design the pipeline: the build system, search-results scanner, and domain portfolio management dashboard.
- Content. Automated generation with analysis of search results and entities costs a few dollars per page. With the typical five pages per site, that comes to tens of dollars per site.
- Links. The main expense. A starting run typically budgets around four hundred dollars per site—and that’s just to find out whether it gains any traction.
- Budget for further optimization. A separate budget for the sites that take off; without it, the first batch will simply burn out.
Overall, the entry threshold for serious work is estimated at tens of thousands of dollars, with the lion’s share going to link-building and repeat rounds. You can save money by doing things in the right order: build links only to sites that have already shown organic traction, rather than to the entire batch.
Domains: newly registered versus expired
The two approaches differ in price by an order of magnitude.
New registrations — fresh domains containing the target keyword. They cost next to nothing and have no history, but they offer no head start either: you’ll have to build links from scratch. This is a sensible option for initial batches—it lets you get a feel for the semantics and the search results structure without burning through your budget.
Expired domains — lapsed domains with accumulated link equity. Prices start in the thousands and can reach tens of thousands of euros for a genuinely valuable domain. They’re selected based on several factors: enough live links, an acceptable DR, content in the language and geo of the target market, and a period of inactivity no longer than two years—after that, the accumulated weight stops working. The Russian segment presents an additional challenge: there are few full-fledged auctions, and notable lots sell quickly.
Two risks eat into investments more often than the rest:
- Loss of trust when changing topics. If a search engine sees a sudden topic change as abuse, the accumulated weight won’t transfer, and an expensive domain becomes an ordinary newly registered one. That’s why some teams change the topic gradually, while others switch it all at once and test the hypothesis quickly.
- Someone else’s trademark and reputation in the domain’s history. Restoring a site that features a registered brand ends in a complaint. Worse still, the domain may have belonged to a real person or organization that simply forgot to renew it—in that case, a public scandal comes on top of the legal complaint. Check the site’s history in the web archive before buying, not afterward.
Specialist services such as Expired Domains and Spamzilla are used to find domains; the latter shows domains that will become available in the next few days, along with snapshots of their previous content.
Behavioral signals: why emulators don’t pass
Behavioral manipulation runs into anti-fraud systems, not ranking algorithms. Those systems trigger when someone clicks a search result—at the redirect layer, before they reach the site.
At this layer, Google reads a device fingerprint that includes parameters an emulator can’t fake: hardware identifiers, the presence of a camera, battery, and SIM card slot. A farm built from motherboards without these components is immediately identified. Technically, the problem can be solved with a fleet of real phones using local IP addresses, but influencing search results at industrial scale requires tens of thousands of devices. The cost of such a fleet puts it out of reach for most teams.
The alternative that works is different: instead of simulating clicks, send real users from your own projects to the page. The search engine sees a spike in interest in the page from a particular geo and raises it for queries the page is relevant to—the effect takes hold within hours. The traffic is laundered through social referrers, and the original referrers are stripped so the source can’t be identified.
This kind of boost costs around a thousand dollars a month for a Tier-1 geo and less in developing markets. It doesn’t work everywhere: practitioners estimate it succeeds on roughly two out of three sites. The stronger the host and the better the other signals, the greater the response.
Parasite SEO
This is a strategy that uses someone else’s trust instead of your own domain: you rent a section on a major media site and publish your own content there. It’s not buying a link; it’s putting your landing page on someone else’s high-authority host.
The advantage is speed: a page on a trusted domain can reach the top as soon as it’s indexed, while your own site could take months. Spam filters have effectively already been cleared, since the host’s trust carries over to the page.
The limitations are clear, too. Few sites are willing to host content from restricted verticals, and renting a section costs thousands of dollars a month—or tens of thousands on top-tier media sites. For white-hat niches, this approach is more accessible and cheaper, yet often overlooked: arranging a placement is frequently faster than spending six months building up your own domain.
Where the legal line falls
One distinction every specialist should keep in mind, regardless of their hat color: violating a search engine’s rules and breaking the law are not the same thing.
Google’s and Yandex’s rules are the policies of commercial companies. The maximum penalty for violating them is spelled out in those same policies: removal of the site from search results. Unpleasant, but a commercial risk, not a criminal offense.
Two scenarios sometimes built into gray-hat schemes are criminal offenses: placing links by hacking into other people’s sites, and denial-of-service attacks. Under Russian law, these fall under statutes on unauthorized access to computer information and the creation of malicious software; similar offenses exist in US and EU legislation. They have nothing to do with optimization—they are separate crimes that merely use promotional infrastructure.
There’s another risk that’s rarely mentioned, and it doesn’t involve search engines: if traffic is directed to a licensed product in a way that violates its license terms, it’s the product itself—not the optimizer—that comes under fire, potentially even losing its regulator’s authorization. There are precedents in European jurisdictions, and the stakes for the operator are out of all proportion to the affiliate’s earnings.
What AI search has changed
For affiliate business models, the main problem with generative search results isn’t lost rankings but a break in the monetization chain. Assistants strip the referrer: the visit arrives unattributed, so the affiliate program can’t credit its source. That traffic is useful to a product promoting itself. It’s almost useless to an intermediary that sells the traffic.
A second observation concerns sources. Platforms featuring user discussions continue to have significant influence on assistant responses, but that influence is unstable: according to market research, Reddit’s share of ChatGPT citations fell severalfold over the course of 2025—from around 60% to single digits—yet it still ranks first among AI engines’ citation sources overall. In these conditions, building a strategy around a single platform is risky: source selection rules change faster than investments in a presence there can pay off.
Third, generative answers largely exclude restricted verticals at the model level. An assistant won’t recommend a product if it could later be held accountable to users and lawyers for doing so. For these topics, the classic ten-blue-links search results will stick around longer than they will for informational queries, where zero-click search has already become the norm.
What works for white-hat projects
Some lessons from the gray-hat segment carry over to conventional sites without any changes.
- Diversify instead of favoring one project. If three out of five areas take off, distribute resources among them instead of investing in the one the owner likes best.
- Topical relevance matters more than volume. HubSpot’s story is illustrative: an overgrown tangentially related blog saw traffic drop severalfold after algorithm updates—estimates vary from 70 to 80% of its previous tens of millions of monthly visits. Writing “about everything for traffic” stopped working even for very strong domains.
- LLMs can replace a junior, but not an architect. Generating code and templates no longer requires a dedicated developer. You need someone to design the entire system from the moment you have more than a dozen projects.
- Links still work. Regardless of the rhetoric around their obsolescence, in competitive niches you won’t reach the top for high-volume queries without a link-building budget.
The overall conclusion is sobering for anyone looking for predictability in SEO: in competitive niches, it has stopped being a craft with guaranteed results and become a discipline in risk management. The winner isn’t the one who finds a working tactic, but the one who can sustain enough attempts and recognize in time when a particular site isn’t going to take off.
