Content operations for affiliate sites

Volume is not the strategy, and where that belief comes from

Almost every piece of affiliate advice reduces publishing to throughput. Post more, rank more. The belief has a real origin: for about a decade, a site that published fifty pages beat one that published five, because there was little competition for the long tail. That era ended. Google's spam policies now name scaled content abuse directly — many pages generated for the primary purpose of manipulating rankings rather than helping people — and the policy explicitly covers pages made by humans, by automation, or by both together.

What replaced throughput is not "quality" in the abstract. It is a pipeline: a repeatable way of deciding what to write, checking it is not a copy of something you already published, holding it to a floor, and dating it honestly. That pipeline is what content operations means. It is boring, it is mostly checklists, and it is the difference between a corpus that compounds and one that gets filtered.

This guide is a publishing pipeline, described so one person with a laptop can run it by hand. For the strategic argument about why first-hand content is the only durable position left, read our companion piece on what changed for affiliate content in the AI era. This one is the machinery.

The brief is the quality gate: what a usable content brief contains

The highest-leverage document in the pipeline is the brief, because it is the last cheap moment to kill a bad idea. Editing a finished 2,000-word article that had no angle costs hours. Rejecting a brief costs a minute.

A usable brief contains all nine of these fields. Anything less and the writer — including you, three weeks from now — fills the gap with filler.

One angle, one worked example, one defensible number

That is the minimum bar, and it is deliberately low enough to be checked by someone who has not read the draft closely. Open the piece and ask three questions. Can you state its angle in one sentence that would not equally describe a competitor's article? Is there a calculation with actual arithmetic in it? Is there at least one number you could defend if a reader emailed to ask where it came from — meaning it carries either a source link or an explicit illustrative label?

Three noes means the piece does not publish. In practice the third question fails most often, because writers reach for statistics they half-remember. The rule is blunt: a figure either links to where it came from, or the sentence introducing it contains the word illustrative. There is no third category.

Cadence: what a solo operator can sustain versus what a team can

Cadence is a budget problem before it is a strategy problem. Work out your real hourly output and multiply, rather than picking a number that sounds impressive.

Here is an illustrative build, with round numbers you should replace with your own. Say a 1,500-word guide that includes a worked example takes six hours end to end: two hours of research and brief, three hours of drafting, one hour of editing and linking. Say you have ten hours a week for content after everything else is done. That is roughly 1.6 pieces a week — call it six a month, or seventy-odd a year. Not fifty a week. Six a month.

Publish on a stagger rather than in batches: one queued item every two to three days per site, with the sites offset from each other so nothing goes out everywhere at once. The reason is not a ranking trick. A steady drip keeps the queue honest — if a queue starts emptying you notice within days, whereas a batch schedule lets a site sit untouched for a month before anyone looks. Each queue carries roughly one to two months of runway at any time, and refilling it is a calendar task rather than an emergency.

Uniqueness: the check that has to run before publish, not after

Duplicate content is usually not a penalty story. It is a selection story: given two near-identical pages, a search engine picks one and ignores the other, and you do not get to choose which. When both pages are yours, you have paid twice and earned once.

The check has to be automatic and it has to run before the piece goes live, because a human reading two articles a month apart will not notice that they share a skeleton. The check worth building runs in the deploy path and compares every candidate against the entire published corpus, on four axes:

• Slug collision — the same URL stem already exists on one of your other sites.
• Title similarity — normalised titles that differ only by a market name or a year.
• Shingle overlap — overlapping runs of consecutive words, which catches paraphrase that a hash misses.
• Exact-body hash — the cheap check that catches copy-paste and template reuse.

It fails closed. If the comparison cannot run at all — corpus missing, file unreadable — the publish is blocked rather than waved through. A gate that passes when it is broken is not a gate, and that is the most common way an automated check quietly stops protecting you.

Why running several sites makes duplication the default failure

With one site, duplication is something you have to go out of your way to do. With four, it is the path of least resistance: the same operator list, the same commission explanation, the same "how bonuses work" section, sitting in a folder waiting to be pasted.

We learned this expensively. One of our earlier sites shared a large editorial corpus with its sister sites, and when we finally measured the overlap, the category hub pages scored above 0.95 on a text-similarity scale where 1.00 is identical. That site stopped earning meaningful search visibility, and we eventually retired it into a redirect to its stronger sibling. Nothing about the writing was bad. It was simply the second copy, and there is no ranking position for the second copy. The rule now is one line: every article belongs to exactly one site, permanently, and nothing is translated or re-skinned across brands.

Word and structure floors, and why they should fail closed

Word count is a bad target and a decent floor. As a target it produces padding; as a floor it catches the piece that promised a guide and delivered four paragraphs. Enforce these in code, so a queued item that misses them cannot be promoted into the live file:

• Blog posts: at least 1,000 words of body text.
• Guides: at least 1,200 words, and in practice 1,500 to 2,500 for anything carrying a calculation.
• At least three H2 sections, each carrying a definition, a worked example, a table or a checklist.
• At least three internal links with descriptive anchor text — never "here", never a bare "more".
• A clear majority of the substance net-new relative to any source the piece cites.

Notice what these floors are not. They are not a quality measure; a 2,000-word article can be worthless. They are a tripwire. They catch the systematic failures — thin, unlinked, unstructured — that otherwise accumulate silently across a hundred pages while each individual piece looks fine.

Dating and updating: never bump a date without an edit

Freshness bumping is a common self-inflicted wound in affiliate publishing: republish the archive with today's date and hope for a lift. Google's helpful content guidance puts it as a self-assessment question: are you "changing the date of pages to make them seem fresh when the content has not substantially changed"? If that is the only reason the date moved, the answer is the problem.

Our rule has two halves. First, the publishing script stamps the real run date at the moment of promotion, which makes backdating structurally impossible rather than merely discouraged — nobody types a flattering date into a field, because there is no such field. Second, the modified date moves only when the visible text actually changed: a corrected figure, a new section, a rewritten example. Re-saving the file is not an update.

When you do update, make it visible. State what changed and when, in the body, in one line. It costs nothing, and it is the only version of freshness a reader benefits from.

Editing: the pass that removes the sentences carrying no information

Do one editing pass with exactly one job: delete every sentence carrying no information. Not "tighten the prose" — delete. The sentences that go are always the same five species:

1. The restatement of the heading you just read ("In this section we will look at cadence.").
2. The universal truth ("Content is important for any affiliate site.").
3. The hedge that cancels the claim ("Results may vary depending on many factors.").
4. The transition that transitions nothing ("With that said, let us move on.").
5. The unsourced statistic ("Studies show most affiliates fail.").

On a first draft this pass typically removes something like a tenth of the words and improves every remaining paragraph, because what is left is denser. If a section survives at half its original length, that is the correct length. If an entire H2 disappears, the brief was wrong, and you have learned something about your briefs.

The second editing job is links. Every piece needs at least three internal links with anchors that describe the destination, placed where a reader would actually want them: a commission term points at the affiliate glossary, a claim about search performance points at the SEO guide for iGaming affiliate sites, and a claim about a commission model points at the revshare, CPA and hybrid comparison rather than restating it.

Where AI fits in the pipeline, and where it must not

Google's stated position is that generation method is not the issue: content produced with AI is acceptable, and content produced at scale primarily to rank is not, whoever or whatever made it. Some stages benefit from a model; at others, using one destroys the only asset you have.

The practical test to apply to any draft: remove every sentence a model could have written without access to your accounts, your contracts and your own measurements. If a full article remains, publish it. If two paragraphs remain, those two paragraphs are the article and the rest was padding.

Measuring whether a piece did anything, honestly

Content measurement in affiliate publishing is slower and blunter than anyone wants. A new page on a new domain is frequently not indexed for weeks, and impressions before indexing are zero by definition — so the first question is never "how did it rank" but "is it in the index at all". Check coverage first, then impressions, then clicks, and only much later conversions.

A workable review rhythm on a small site:

• Week 2 after publish: is the URL indexed, and does it have at least one internal link from another page? If not, fix the link, not the copy.
• Month 2: does it have impressions for anything? Impressions without clicks is a title and description problem. No impressions at all is a relevance or indexing problem.
• Month 6: clicks and on-page behaviour. This is the first honest read on whether the angle was right.
• Month 12: revenue attribution, if your tracking survived that long.

Two warnings from our own operation, stated plainly. First, early traffic on a new affiliate site is frequently not human: examined closely, a large share of the activity usually sits in countries the site does not serve and never converts. A geo mismatch between your audience and your clicks is the cheapest bot tell there is. Second, do not measure content against revenue in year one — measure it against indexing, then impressions. Revenue attribution needs a real depositing player, and that sits downstream of everything else on this list. Before committing a year to a publishing plan, put your real hourly cost and a conservative conversion assumption into the break-even calculator and see what that plan has to produce to pay for itself.

What we would do this week

1. Write the brief template once, then use it for the next five pieces. Nine fields: slug, title, angle, reader, outline, numbers, worked example, internal links, word floor. Reject any idea that cannot fill the angle field in one sentence.
2. Run a duplication audit across everything you have already published. Compare every page against every other page on titles, opening 200 words and section headings. Merge or redirect the near-duplicates, and write down the rule that stops it recurring: one topic, one page, one site.
3. Pick a cadence you can hold for six months and put it in a calendar. Four solid pieces a month beats twelve in one month and nothing after. Then read the anatomy of a dead funnel before you start a second one, because a second site doubles the number of places where tracked clicks can quietly stop converting and nobody notices for a quarter.