For affiliate publishers and review sites
Comparison pages for affiliate sites, built from your own test notes
Readers who search two product names want a pick for their situation and a reason they can check. Put each pair, use case, pick and test date in a row, design one comparison layout, and update the pages when a retest changes the answer.
- Free on WordPress.org
- Update picks from the CSV
- Duplicate URL check before writing
Why comparison pages get skipped
A spec table with two columns is not a comparison
Readers and search engines both pass over pages that rearrange the manufacturer’s spec sheet. What they look for is a verdict for a specific situation, evidence that you used the products, and a date that shows it still holds.
The usual comparison workflow
- Copy specs from two product pages into a table and call the longer battery the winner
- Write one page per pair, then copy it for every use case with a new heading
- A new model ships and nobody knows which posts still recommend the old one
- Prices typed into the copy are wrong within a week
With PageForge
- One row per pair and use case, with your pick, the reason and the test date
- Use-case pages only where the use case changes the answer
- Edit a pick, change TestedOn and refresh that page in one update run
- Prices left to your affiliate program’s live widget, never typed into the CSV
The data
One row per page, every detail that changes
Four fictional robot vacuums, six pairs and seven verdicts. ProductA, ProductB and UseCase build the title and URL, Pick and WhyPick give the verdict and its reason, the two runtime columns are the evidence, and TestedOn tells the reader how fresh the verdict is.
| ProductA | ProductB | UseCase | Pick | WhyPick | RuntimeA | RuntimeB | TestedOn | |
|---|---|---|---|---|---|---|---|---|
| 2 | Kestrel V5 | Tidy Owl Max | Pet Hair | Tidy Owl Max | Rubber rollers resist tangling with long hair | 120 min | 150 min | 2026-09-14 |
| 3 | Kestrel V5 | Brisa One | Small Apartments | Brisa One | Smaller body fits under low sofas | 120 min | 90 min | 2026-09-14 |
| 4 | Tidy Owl Max | Novo Sweep 2 | Large Homes | Novo Sweep 2 | Maps several floors and returns to recharge mid-clean | 150 min | 180 min | 2026-09-21 |
| 5 | Brisa One | Novo Sweep 2 | Tight Budgets | Brisa One | Costs less and handles hard floors well | 90 min | 180 min | 2026-09-21 |
| 6 | Kestrel V5 | Novo Sweep 2 | Carpet | Novo Sweep 2 | Stronger carpet mode and a larger dust bin | 120 min | 180 min | 2026-10-02 |
| 7 | Tidy Owl Max | Brisa One | Quiet Cleaning | Brisa One | Quieter mode for night runs | 150 min | 90 min | 2026-10-02 |
| 8 | Kestrel V5 | Tidy Owl Max | Hard Floors | Kestrel V5 | Mop pad and edge brush reach into corners | 120 min | 150 min | 2026-10-09 |
Placeholders
{ProductA} {ProductB} {UseCase} {Pick} {WhyPick} {RuntimeA} {RuntimeB} {TestedOn}
Patterns
Title{ProductA} vs {ProductB} for {UseCase}
URL/{producta}-vs-{productb}-{usecase}/
Campaign size
10 models make 45 pairs; 45 pairs × 3 use cases = 135 possible pages
Pages this creates
- Kestrel V5 vs Tidy Owl Max for Pet Hair
- Kestrel V5 vs Brisa One for Small Apartments
- Tidy Owl Max vs Novo Sweep 2 for Large Homes
- Brisa One vs Novo Sweep 2 for Tight Budgets
- Kestrel V5 vs Novo Sweep 2 for Carpet
How it works
From a testing log to published comparisons in four steps
01
Log every test as a row
One row per pair and use case you actually compared. Record the pick, a one-line reason, the figures that support it and the date. Leave untested pairs out of the file.
02
Build the comparison template
Lay out the short answer, the reason, the figures, who should buy the other one and your disclosure. Set the Title Pattern to {ProductA} vs {ProductB} for {UseCase}.
03
Check, then read against your notes
Check and preview flags rows that produce the same URL and shows the first title. Generate drafts and read each verdict against your testing notes before you publish.
04
Retest, edit the row, update
When a new model or a retest changes a pick, edit the row, change TestedOn and run Update it from the CSV and current template. Pages you rewrote by hand are skipped.
Built for comparison sites
What a comparison page needs, filled from your testing log
Your template holds the layout, the method and the disclosure. Your rows hold the verdicts. PageForge joins them and builds the URL, title and metadata for each pair.
The generated page
What a reader sees on Kestrel V5 vs Tidy Owl Max for Pet Hair
The first row becomes a page that gives the verdict before anything else: Tidy Owl Max, because its rubber rollers resist tangling with long hair. The runtimes, 120 minutes against 150, and the 2026-09-14 test date show the verdict is first-hand and recent. A reader who searched two product names is close to buying, so the link to the pick sits right under the answer.
- The pick and its reason in the first lines
- Runtime figures side by side as evidence
- A last-tested date the reader can see
Before publishing
Catch the pair you entered twice
Check and preview builds every URL before anything is written. If the same pair and use case appear twice in the same order, the second row is skipped as a duplicate URL, and a title placeholder with no matching column stops the run.
- See which rows will be skipped, and why
- Read the first title exactly as it will appear
- Nothing is written until you start the run
When picks change
Retest, edit one row, refresh one page
New firmware or a new model can flip a verdict. Edit Pick, WhyPick and TestedOn, upload the CSV and run the template in update mode: the changed page is refreshed and the rest are reported as unchanged.
- Status and URL stay as they are
- Pages you rewrote by hand are skipped
- A new product name makes a new URL, so redirect the old page yourself
Compared
PageForge versus copying comparison posts by hand
A reader cannot tell how a page was built. They can tell whether it has a verdict, a reason and a recent test date.
| PageForge | Copying comparison posts | |
|---|---|---|
| Verdict and reason on every page | From the Pick and WhyPick columns | Rewritten each time, if at all |
| Test date shown to readers | From the TestedOn column | Usually missing |
| Duplicate URL check before writing | Yes | No |
| Changing one pick | Edit a row, run update | Find the post and edit it |
| SEO title and meta description per page | Built from patterns | Typed by hand, often skipped |
| Disclosure on every page | Written once, in the template | Pasted into each post |
What comparison pages for affiliate sites need to do
Comparison pages for affiliate sites serve one reader: someone who has narrowed a purchase to two products and wants to be told which one suits their situation, and why. They search the pair, “kestrel v5 vs tidy owl max”, often with a qualifier such as “for pet hair” or “for a small apartment”. The page that earns their click gives a verdict in the first lines, shows the evidence, and says who should buy the other one.
PageForge builds those pages in WordPress from your testing log. Each row is one pair judged for one use: your pick, the reason, the figures and the date you tested. One template holds the layout, your method and your disclosure. Each page gets its own title, URL, meta description and SEO plugin fields.
The sample uses four fictional robot vacuums. Your rows should come from products you have used or tested.
The searches behind comparison pages
| Query shape | Example | What the reader wants |
|---|---|---|
| X vs Y | “kestrel v5 vs tidy owl max” | The better product overall |
| X vs Y for a use | “tidy owl max vs novo sweep 2 for large homes” | The better product for their home |
| X or Y | “brisa one or novo sweep 2” | A fast decision |
| Best for a use | “best robot vacuum for pet hair” | A shortlist |
The first three shapes fit a template, because the page structure is the same for every pair. The fourth is a roundup. It needs an editor’s judgment across many products and is better written by hand, with links out to your pair pages. Programmatic SEO keyword research covers how to find which pairs people actually search.
Plan the pairs before you build
Pairs grow faster than most publishers expect. Ten models make 45 pairs, and judging each pair for three uses makes 135 possible pages. That number is a ceiling, not a target.
The sample shows the right ratio. Four models make six pairs, and the file names seven uses: pet hair, small apartments, large homes, tight budgets, carpet, quiet cleaning and hard floors. Six pairs times seven uses is 42 possible pages. The file has 7 rows, one in six, because those are the combinations that were tested.
Give a use case its own page only when it changes the answer
Kestrel V5 vs Tidy Owl Max has two rows, and that is justified: the pick flips. Tidy Owl Max wins on pet hair, Kestrel V5 on hard floors. If the same product won every use case for the same reason, one pair page with a short section per use would serve the reader better, and separate URLs would only compete with each other for the same search.
Fix one order for every pair
Row 6 is Tidy Owl Max vs Brisa One. If someone later adds Brisa One vs Tidy Owl Max for quiet cleaning, it gets a different URL, so Check and preview has no reason to flag it, and you now have two pages answering one question. Pick an order rule, such as alphabetical, before the first run, and sort the sheet by ProductA and ProductB to catch reversals. The sample does not follow alphabetical order, which is fine as long as nobody adds the reverse of an existing pair. Duplicate URLs and skipped rows explains what PageForge does catch.
What goes in each row
| Column | What it holds | Value in row 1 |
|---|---|---|
| ProductA, ProductB | The pair, in your fixed order | Kestrel V5, Tidy Owl Max |
| UseCase | The situation you judged | Pet Hair |
| Pick | Your verdict | Tidy Owl Max |
| WhyPick | The reason, in one sentence | Rubber rollers resist tangling with long hair |
| RuntimeA, RuntimeB | Figures that support it | 120 min, 150 min |
| TestedOn | When you last checked | 2026-09-14 |
Columns worth adding
- OtherFor: who should buy the product you did not pick. Every page needs this, and the sample has no column for it.
- TestNotes: one sentence on what happened in your test, in your own words.
- Tester: who ran it, so the page can say so.
- RuntimeMode: the setting the runtime was measured on, so the figures can be compared.
- LinkA, LinkB: the affiliate link for each product.
Keep prices out of the file. They change daily, and a typed price is wrong within a week.
Google’s spam policies name thin affiliation: affiliate pages that copy merchant descriptions and reviews without adding original value. Its guide to writing high-quality reviews asks for the opposite: evidence from your own use, measurements, what sets one product apart, and who each one suits. Every column above maps to one of those.
A worked example: four robot vacuums, seven verdicts
The testing log, with fictional products:
ProductA,ProductB,UseCase,Pick,WhyPick,RuntimeA,RuntimeB,TestedOn
Kestrel V5,Tidy Owl Max,Pet Hair,Tidy Owl Max,Rubber rollers resist tangling with long hair,120 min,150 min,2026-09-14
Kestrel V5,Brisa One,Small Apartments,Brisa One,Smaller body fits under low sofas,120 min,90 min,2026-09-14
Tidy Owl Max,Novo Sweep 2,Large Homes,Novo Sweep 2,Maps several floors and returns to recharge mid-clean,150 min,180 min,2026-09-21
Brisa One,Novo Sweep 2,Tight Budgets,Brisa One,Costs less and handles hard floors well,90 min,180 min,2026-09-21
Kestrel V5,Novo Sweep 2,Carpet,Novo Sweep 2,Stronger carpet mode and a larger dust bin,120 min,180 min,2026-10-02
Tidy Owl Max,Brisa One,Quiet Cleaning,Brisa One,Quieter mode for night runs,150 min,90 min,2026-10-02
Kestrel V5,Tidy Owl Max,Hard Floors,Kestrel V5,Mop pad and edge brush reach into corners,120 min,150 min,2026-10-09
The patterns:
| Field | Pattern |
|---|---|
| Title Pattern | {ProductA} vs {ProductB} for {UseCase} |
| Slug Pattern | {producta}-vs-{productb}-{usecase} |
| Meta Description Pattern | {ProductA} or {ProductB} for {UseCase}? We pick {Pick}: {WhyPick}. Tested {TestedOn}. |
What the seven rows produce:
| Row | Page title | URL | Pick |
|---|---|---|---|
| 1 | Kestrel V5 vs Tidy Owl Max for Pet Hair | /kestrel-v5-vs-tidy-owl-max-pet-hair/ | Tidy Owl Max |
| 2 | Kestrel V5 vs Brisa One for Small Apartments | /kestrel-v5-vs-brisa-one-small-apartments/ | Brisa One |
| 3 | Tidy Owl Max vs Novo Sweep 2 for Large Homes | /tidy-owl-max-vs-novo-sweep-2-large-homes/ | Novo Sweep 2 |
| 4 | Brisa One vs Novo Sweep 2 for Tight Budgets | /brisa-one-vs-novo-sweep-2-tight-budgets/ | Brisa One |
| 5 | Kestrel V5 vs Novo Sweep 2 for Carpet | /kestrel-v5-vs-novo-sweep-2-carpet/ | Novo Sweep 2 |
| 6 | Tidy Owl Max vs Brisa One for Quiet Cleaning | /tidy-owl-max-vs-brisa-one-quiet-cleaning/ | Brisa One |
| 7 | Kestrel V5 vs Tidy Owl Max for Hard Floors | /kestrel-v5-vs-tidy-owl-max-hard-floors/ | Kestrel V5 |
Seven rows give seven pages and no duplicate URLs, covering all six pairs. Brisa One wins three times, Novo Sweep 2 twice, and Tidy Owl Max and Kestrel V5 once each. The tests ran on four dates between 2026-09-14 and 2026-10-09. Titles run from 37 to 44 characters, and meta descriptions from 112 to 141, so none runs past the length search results tend to show. Row 1’s reads “Kestrel V5 or Tidy Owl Max for Pet Hair? We pick Tidy Owl Max: Rubber rollers resist tangling with long hair. Tested 2026-09-14.”
Look at the runtimes. In four of the seven rows (2, 4, 6 and 7), the pick has the shorter battery life. Brisa One has the shortest runtime in the set, 90 minutes, and still wins three use cases, for fitting under low sofas, for price and for noise. A verdict taken from the spec sheet, where the longest runtime wins, would have picked the other product in all four. That gap is the reason your testing is worth reading.
With no Service column, the SEO title is the page title and the focus keyphrase follows it. The schema is your business item and a WebPage item. PageForge adds no Review schema, and its basic Product item only appears on product content types, not on comparison Pages. See Schema markup and Title, URL and meta description patterns.
Template details the data forces
- WhyPick values start with a capital letter, so put them after a colon or a label:
Why we pick {Pick}: {WhyPick}. - Runtimes carry their unit, so
Battery life: {RuntimeA} for {ProductA}, {RuntimeB} for {ProductB}.reads “Battery life: 120 min for Kestrel V5, 150 min for Tidy Owl Max.” - TestedOn prints exactly as typed. Keep 2026-09-14 if you like sorting by date, or type September 14, 2026 if you want words on the page.
Rows that need more data
- All seven rows lack “who should buy the other one”. Add an OtherFor column. Row 7 already hands row 1 its answer: Kestrel V5 suits a reader whose floors are mostly hard.
- Rows 2, 4 and 6 pick the 90-minute Brisa One. Each of those pages has to say why runtime does not decide that use case, or a reader comparing 90 with 120, 180 or 150 will doubt the verdict.
- Row 4 says “Costs less” with no figure. Keeping prices out of the CSV is right, but that page needs a live price for both products near the verdict, so the reader can check the claim.
What to put on every comparison page
- The verdict in the first two lines: which product, for which situation
- The reason, in one sentence a reader could test
- The figures behind it, with how and when you measured them
- Who should buy the other product instead
- What you tested and the date, visible on the page
- Photos or notes from your own use
- Current prices from a live source, never typed text
- Links to both products, marked as paid links
- The affiliate disclosure near the top
- Links to the other comparisons that include each product
Disclosures, links and prices
Put the disclosure in the template, near the top, so every generated page carries the same wording. A footer line alone is easy to miss. Check the disclosure rules where you publish and the terms of each affiliate program you use.
Put each product’s affiliate link in a column, LinkA and LinkB, and write the links into the template with rel="sponsored". Google’s guide to qualifying outbound links explains the value for paid links. Generate one draft and click both links before you run the rest.
For prices, place your affiliate program’s price widget or shortcode in the template. The template is copied to every page, so the widget appears on each one and stays current without a CSV change.
How to run it in PageForge
With PageForge Free:
- Open PageForge > Settings and set Business Type (Schema) to Organization.
- Open PageForge > Data Sources, choose the testing log under CSV Upload and click Upload CSV.
- Open PageForge > Templates and click Add New Template. Build the comparison layout with the disclosure at the top, or choose a designed page as Base Layout to Clone. In PageForge Template Settings, enter the patterns above, set Target Post Type to Page and Generated Post Status to Draft. To nest the set under a hub such as /robot-vacuums/, publish that page first and choose it as Parent Page.
- On the Generate screen, choose the template and CSV and click Check and preview. Expect
Ready: 7 new, 0 to update, 0 skipped.with the first page at /kestrel-v5-vs-tidy-owl-max-pet-hair/. - Click Generate 7 pages now and read every verdict against your testing notes before you publish.
- Add the PageForge Sitemap block to a comparisons hub with grouping set to none. In Free the list holds every published page PageForge generated, so several product categories share one list. Pro can include or exclude pages by ID. See Hub pages with the sitemap block.
Keeping comparison pages for affiliate sites current
A comparison is true on the day you test it. Firmware changes, a model is replaced, a price drops. Sort the sheet by TestedOn every month and retest the oldest rows first.
When a pick changes, edit Pick, WhyPick and TestedOn, upload the file and run the same template with Update it from the CSV and current template. Only the pages whose row changed are rewritten. The rest are reported as unchanged, and pages you rewrote by hand are skipped unless you choose to overwrite them. See Update pages you already generated.
A new model is a new product name, so its rows make new URLs. When Novo Sweep 3 replaces Novo Sweep 2, add rows for the new pairs and decide what happens to the old pages: update them to mention the successor, or redirect them yourself.
Pro On Professional and Enterprise, a linked Google Sheet updates pages when their rows change, and removing a discontinued product’s rows can move its pages to draft once you have previewed and applied a sync. See Google Sheets to WordPress. Every Pro plan adds no per-run row limit, a queue for drip publishing, and AI content per page for a first draft of the comparison text that you then check against your notes. See AI content for WordPress pages.
Common mistakes
- Spec-sheet verdicts. Picking the longer battery or the bigger number, when your test said otherwise.
- Publishing untested pairs. If you have not compared them, leave the row out.
- Reversed pairs. Two orders of the same pair make two URLs for one question.
- Typed prices. Use a live source.
- Use-case pages that do not change the answer. Fold them into the pair page.
- A hidden disclosure. Put it near the top of the template.
- Review markup the page does not support. If your SEO plugin adds review schema, it must match what the page shows.
Free or Pro
The free plugin runs up to 100 rows per run and covers the whole workflow: CSV upload, the template, Check and preview, update mode, SEO plugin fields and the Sitemap block. The 135-page ceiling from ten models passes that, and Pro has no row limit. Move to Pro when your log runs past 100 rows, when you want a sheet that keeps pages in step on its own, or when you want AI first drafts from your notes. More page sets built this way are in programmatic SEO examples and the ecommerce landing pages use case. Compare editions on Free vs Pro.
Use cases
Blueprints with a sample CSV included
{Category} for {Audience}Landing pages by use, audience or occasion that show live products from your store.12 categories × 6 audiences = 72 pages, keeping only the pairs you stock wellSee the blueprintSaaS integration, use-case and alternative pagesPollywise + {Integration} integrationIntegration, use-case and alternative pages built from your real product data.40 integrations + 12 industries + 8 alternatives = 60 pages from three templatesSee the blueprintDirectories and listing sites{Category} in {City}, {State}Category and city pages for directories, published only where you have real listings.8 categories × 60 cities = 480 possible pages, published only where listings existSee the blueprintWill Google treat generated comparison pages as thin affiliate content?
It will if the pages only repeat what the manufacturers publish. Google’s spam policies name thin affiliation: pages that copy product descriptions and reviews without adding original value. Each row should carry your own verdict, the reason, figures you measured and the date you tested, and pairs you have not compared should stay out of the file.
How many product pairs should I publish?
Publish only the pairs you have actually compared. Ten models make 45 possible pairs, and three use cases each make 135 pages, but a page per combination is not a goal. Start with the pairs readers search for most, usually the current best sellers against each other, and add use-case pages only where the use case changes the pick.
Should every use case get its own page?
Only when the use case changes the answer or the reasons. In the sample, Kestrel V5 vs Tidy Owl Max gets two pages because the pick flips: Tidy Owl Max wins on pet hair, Kestrel V5 on hard floors. If the same product wins every use case for the same reason, one pair page with a section per use serves the reader better.
Can I put prices in the CSV?
No, keep prices out of the data. Prices change daily, and a page that quotes a figure the retailer changed last week loses the reader’s trust on the spot. Place your affiliate program’s price widget or a link to the current price in the template, and keep the CSV for facts you control, such as your pick, your reasons and your test date.
Where should the affiliate disclosure go?
Put it in the template, near the top, so every generated page carries the same wording. A disclosure in the footer alone is easy to miss. Check the rules that apply where you publish and the terms of each affiliate program you use, and mark affiliate links with rel=”sponsored”, the value Google asks for on paid links.
Does PageForge add review or product schema to comparison pages?
No. PageForge adds your business item and a WebPage item to each page, and it does not generate Review, FAQ or Breadcrumb schema. A basic Product item only appears on content types whose name includes product, such as WooCommerce products, not on comparison Pages. If you add review markup through your SEO plugin, make sure it matches what the page shows.
What happens when a manufacturer releases a new model?
Retest, then edit the rows that change. Update Pick, WhyPick and TestedOn, upload the CSV and run Update it from the CSV and current template, and only the changed pages are refreshed. A new model name is a new URL, so its rows create new pages, and the old pages stay until you update or redirect them yourself.
Can AI write the comparisons for me?
On Pro it can write a first draft, not the verdict. AI content per page sends each row’s values to your own AI provider, so the draft is only as good as your notes. Keep the pick, the reason and the figures in your columns, read every sentence against your testing, and remove anything you cannot back up. Every Pro plan includes it.
Publish the comparisons you have actually tested
Install the free plugin, upload five rows from your testing log and generate them as drafts. Read each verdict against your notes before anything goes live.
