← Blog
AI Tools · November 22, 2026 · 9 min read

One Brief, Ten Guides: Writing a Category Batch With an AI Writer So the Ten Pages Share a Reader Model and Not a Shape

Z
One Brief, Ten Guides: Writing a Category Batch With an AI Writer So the Ten Pages Share a Reader Model and Not a Shape

Writing one buyer guide with an AI writer is a brief and an edit. Writing ten in the same category is a different problem, and the two ways sites usually solve it both fail. Brief all ten from one template with the keyword swapped, and the ten converge: same shape, same seven picks, same opening sentence with a different noun.

Brief each one separately, and they drift: they disagree about which spec matters, name the same feature three ways, and overlap in picks by accident. The batch that works has one brief for the category and a short one for each guide, and the two layers do different jobs.

This post covers what goes in the category layer (shared by all ten) and what goes in the per-guide layer (different for each), why ten separate briefs drift and one template converges, a worked two-layer brief for a real category, the side-by-side check before any of the ten publishes, and the order and pace to publish them in.

The prompting post covers the six inputs a single brief needs; this is how those inputs split when there are ten.

The short answer

Write the category layer once: reader types, deciding specs, tiers in dollars, brands, vocabulary, voice, banned claims, method line, the full product list. Write a per-guide layer for each: keyword, the one reader it serves, its constraint, the picks it must diverge from, and its two internal links. Brief each guide with both.

Before publishing, put the ten pick lists side by side and merge any two that overlap by more than half. Publish the hub, then the main guide, then two or three modifier guides a week.

Two layers

Category layer (written once, shared by all ten)Per-guide layer (written per guide, 60 to 100 words)
Reader types in the category (commuter, trail rider, casual, kids' bikes)Which one of them this guide is for
Deciding specs (beam shape, runtime at mid mode, mount, charging) and the headline spec to dismiss (lumens)Which deciding spec leads for this reader
Price tiers in dollarsThis guide's band, if it has one ("under $50")
Brands to include; brands to avoid and whyNothing, unless this guide is a brand guide
Vocabulary: the words buyers useNothing
Voice briefNothing
Banned claims and the method lineNothing
The full product list for the category, with specs, from the APIThe picks other guides in the batch already own as their top three, which this one must diverge from or justify
The hub this category lives underThis guide's two internal links: the hub and one neighbour guide
Section rotation rules for the batch (the guides at scale post has them)Which opener and structure this guide draws

The left column is the category research, written as a brief. The right column is short because almost everything that varies between the ten guides is the reader and the constraint, and those are one line each.

Why separate briefs drift and one template converges

A worked example

Category layer (bike lights, about 200 words): "Readers: commuters on lit roads; commuters on unlit roads; trail riders; casual evening riders; parents buying for a child's bike. Deciding specs: beam shape (cut-off vs flood), runtime at the mid setting, mount type (quick-release vs bolt-on, bar diameter), charging port.

Lumens is the headline spec and mostly doesn't decide; say so in one line. Tiers: budget under $40, mid $40 to $100, premium over $100. Brands: Cygolite, Lezyne, Light & Motion, Knog, NiteRider; avoid unbranded listings and any brand with under 100 reviews across its range. Vocabulary: 'cut-off beam', 'be-seen light' vs 'see-by light', 'bar mount', 'helmet mount'.

Voice: plain, second person, no exclamation marks, verdicts blunt. Banned: 'blinding', 'brightest ever', any claim of testing, typed prices. Method: compared on specs, owner review patterns and current price; not tested. Hub: /bike-lights/. Products: [the API list, 40 items with specs]."

Per-guide layer, guide 3 (about 70 words): "Keyword: best bike lights for commuting on unlit roads. Reader: the commuter on unlit roads. Leading spec: beam shape, then runtime at mid. Band: none.

Diverge from: guide 1's top three (Cygolite Metro Pro 800, Lezyne Hecto, Knog Blinder) and guide 2's (Light & Motion Urban 500, Knog PWR, NiteRider Lumina); this reader needs more throw, so expect a different top pick or a justification. Links: the hub, and guide 2 (lit-road commuters) as the neighbour. Structure draw: opener B, structure 2."

Ten of the second kind take an hour; the first takes two. The brief the writer receives is the two together, and the guide that comes out agrees with the other nine about the category and disagrees with them about the reader.

The side-by-side check

  1. A sheet: ten rows, one per guide; ten columns, the picks in order.
  2. For each pair of guides, count shared picks. More than five means one guide is a section of the other: merge, or reassign its reader. Two to four shared is normal; the category has a few products that suit several readers.
  3. Top picks: no two guides share one unless the "pick this if" lines are clearly different readers. The modifiers post has the top-three test for whether a modifier earned its guide.
  4. Openers and structures: no two adjacent guides in publishing order draw the same.
  5. Vocabulary: search the ten for the category's terms; the same thing has one name everywhere.

Order and pace

Doing it with AffSERP

AffSERP's Keyword Grouping tool produces the reader assignment the per-guide layer needs: it clusters a category's keywords by the reader behind them, so each guide in the batch gets its reader and its neighbour before anything is written.

The X Best Review tool takes the category layer as the site's shared settings (voice, banned claims, method, tiers) and the per-guide layer as each guide's brief, selects picks from the live product list with the divergence instruction applied, and rotates openers and structures across the batch so the ten draw differently.

The side-by-side sheet is still worth the twenty minutes, and the publishing pace is yours; the Refresh option keeps the ten in agreement when the category's products change.

FAQ

What is a category brief for AI content?

The part of the brief that's true of every guide in a category: who buys these products and why (the reader types), the three to five specs that decide the choice, the price tiers in dollars, the brands to include and avoid, the words buyers use, the voice, the banned claims and the method line. It's written once from the category research and reused by every guide in the batch, so the ten pages agree about the category. The per-guide brief then adds only what differs: the keyword, the reader this guide serves, its constraint, and the picks it should diverge from.

Why do AI-written guides in the same category end up sounding the same?

Two opposite mistakes produce it. Ten guides briefed from one template with only the keyword swapped converge, because the writer fills the same slots the same way and selects the same bestsellers ten times. Ten guides briefed separately, each from scratch, drift: they disagree about which specs matter, call the same thing by different names, and overlap in picks by accident. The fix for both is the same two-layer brief: a shared category layer so they agree, and a per-guide layer so each has its own reader, constraint and picks.

How do I keep ten buyer guides from recommending the same products?

Assign readers and constraints before any guide is written, so each guide has a different reason to pick differently; put the category's full product list in the shared layer and tell each guide which picks the others already own as their top three, so it must diverge or justify; and run the side-by-side check before publishing (ten guides, ten picks each, in a sheet). More than half overlap between any two means one of them is a section of the other, not a page.

In what order should I publish a batch of guides?

Hub first if it doesn't exist, then the category's main 'best X' guide, then the modifier guides in order of search volume, two or three a week rather than all ten in a day. Internal links go in as each publishes (every guide to the hub and to one neighbour). A batch dropped in one day with identical templates and dates is the pattern Google's scaled-content systems look for; a batch over three or four weeks with real differences between the pages is a site growing.

Readers assigned before anything is written

AffSERP's Keyword Grouping clusters a category's keywords by reader, and X Best Review takes the shared category settings plus each guide's brief, with picks selected to diverge and structures rotated across the batch.

Start with free words
Z
Written by Zoe N. Content Lead, AffSERP

Zoe writes about affiliate content, SEO and the tooling behind niche sites, drawing on what works (and what quietly fails) across the guides published with AffSERP.

More about Zoe →