How to Build Content Briefs for Comparison and Review Pages

The problem a brief actually solves

Most affiliate sites do not have a quality problem in the way editors imagine. They have a consistency problem. Ten reviews written across three months by two writers produce ten different section orders, four ways of presenting pricing, and no reliable place to put the comparison table. Six months later you want to add a "who should skip this" module to every review, and there is nothing to hook into. The pages are not bad. They are just not the same shape as each other.

That shape is what a brief protects. On a general content site a brief is a writing aid. On an affiliate site it is closer to a schema definition: the contract that makes internal linking, templating, and bulk updates possible later. Skip it and every structural change becomes a manual pass over hundreds of URLs, which in practice means it never happens. A site we run publishes 319 articles alongside a separate directory of tool pages, and restructuring anything is only affordable where the pages already agree with each other.

Step 1: read the SERP for format, not for keywords

The most common briefing error is treating the keyword as the specification. It is not. The specification is the format Google currently rewards for that query, and the only place to read it is the results page. Open the query in a clean session, in the country you actually target, and answer four questions before writing a line.

• What page type holds positions one through five? Vendor pages, listicles, single-product reviews, forum threads, and video results are five different answers demanding five different documents. If the top five are vendor pages, the query is navigational and your review is fighting for scraps.
• Is the top result a comparison or a single review? "X vs Y" queries frequently return single-product reviews, and "X review" queries return roundups. The keyword's grammar lies; the results do not.
• Are there forum results in the top ten? That is Google saying marketing copy is not satisfying the query and first-hand experience is. A polished, sourceless review will not displace a thread where four people describe what broke for them.
• What sits above the organic results? A product carousel, an AI summary, or a video pack consumes the clicks you were forecasting. Sometimes the right output of this step is a decision not to write the page at all.

Write the answers into the brief as a short SERP read block. Months later, when the page underperforms, this is the only artifact that tells you whether the strategy was wrong or the execution was.

Step 2: count the competing documents instead of judging them

Human judgment about "depth" is unreliable and unrepeatable. Counting is neither. Take the three to five results that are the same page type you intend to publish, discard the vendor's own domain, and record hard numbers.

What to count	Why it matters
Number of H2 sections	Sets the floor for coverage. If every competitor runs nine sections and your brief specifies five, you are ordering a thinner page on purpose.
How pricing is exposed	Exact figures, tier names, or "starts at" language? The field most likely to make your page wrong within a quarter. Decide the format once, site-wide.
Subheadings shared by every competitor	A heading in one document is optional colour; a heading in all of them is the query's definition.

The output is not a word count. It is a section list with a rationale attached to every entry. If you cannot state in one clause why a section is in the brief, it does not go in the brief.

Where a SERP analysis tool pays for itself, and where it does not

Everything in Step 2 is mechanical, and mechanical work is what tooling is for. Surfer is what we reach for at this exact step: pull the ranking set for a query, extract their heading structures, and report term and entity coverage across the group. Forty minutes of tab-switching and manual tallying becomes closer to five, and it is the same five minutes every time, which is the entire point of having a process.

Be precise about what that buys. It is good at three things: extraction and tallying, gathered without the fatigue that makes humans sloppy on the eleventh page of the day; surfacing vocabulary gaps, which on a comparison page regularly exposes a missing dimension rather than a missing phrase; and enforcing consistency, since a generated outline gives two freelancers the same starting structure.

It is not good at three others, and does not claim to be. It cannot decide intent — no tool tells you that a SERP returning three forum threads means your polished review is the wrong artifact. It cannot judge whether to compete, because coverage scores say nothing about whether a SERP is winnable with the authority you have. And it cannot supply evidence: if competitors win on hands-on testing, term coverage is no substitute, and the tool will score a well-structured but entirely derivative page highly, because structure is what it measures.

The failure mode to watch for is optimising toward the score. A content score is a proxy for "resembles the documents that currently rank" — a floor, not a ceiling. We use it as a pre-flight check and never as an acceptance criterion. Writers who chase the number produce pages that read like an average of page one, and an average of page one is not a reason to rank above it. Evaluate any SERP tool on that basis: does it shorten the mechanical step, or does it quietly become the editorial standard?

Step 3: the fields a review brief must contain

This field set is non-optional. Anything beyond it is house style; anything missing from it is a page that will need rework.

• Target query, page type, and the SERP read. Single-product review, head-to-head comparison, or category roundup — stated explicitly, never merged mid-draft.
• The reader's decision state. Not a persona — the actual question in their head. Choosing between two shortlisted products, replacing something they pay for, or checking whether a tool does one specific thing. Each needs a different opening.
• Section list in order, with a rationale per section. Derived from Step 2. The load-bearing field.
• Table specification. Exact column headers and row order, fixed at brief time. Columns invented during drafting are why your tables do not match across the site.
• Named alternatives and required internal links. Which competing products get mentioned, which of your pages they link to, and which pages must link back. Decide this in the brief and your link graph builds itself; decide it in the draft and it does not.
• The disqualifier section. Who should not buy this. The section that most reliably separates a real review from restated marketing copy, and the first to disappear without a brief.
• Claims requiring verification, and the refresh trigger. Every assertion the writer must source, plus the event that will make this page wrong.

The handoff and the definition of done

A brief is only worth writing if it is also the acceptance checklist. When a draft returns, the reviewer does not react to the prose. They walk the brief field by field: every section present and in order, the table matching the specified columns, pricing in house format with sources attached, the disqualifier section actually disqualifying someone, internal links in place both ways. Anything the reviewer wants that is not in the brief is a change request against the brief, not against the writer — a distinction that keeps freelance relationships functional at volume.

One last discipline, learned by failing at it: store the brief with the published page, retrievable by slug. When a page underperforms, it tells you what you intended. When a vendor changes pricing, it tells you which sections are now wrong. A brief discarded at publication solved a one-day problem; a brief that is kept solves the next three years of them.