The short version
- Buyers search the comparison query whether or not you answer it. The result page fills up either way.
- The table is the page. An audit of 200 essays on this site found one table in the entire library and no internal links between them.
- Every claim about a competitor needs a source and a date beside it, because their releases make your page wrong without anybody touching it.
- Put it behind a publish gate. One account runs 44 content pages through an automated validator before anything goes live.
A comparison page answers the query a shortlisting buyer types with your name and a rival’s name in it. It gets refused internally because writing it means describing a competitor accurately. The version that works carries a real table, a source recorded against every claim, a review date, and one honest statement of who should buy the other product.
Why does nobody want to write it?
Because it means putting a competitor’s strength in writing, on your own domain. Every round of review pulls toward deleting that sentence, and what survives the reviews is a feature grid with every tick in one column, which no buyer has ever believed.
The page also collects approvals that a blog post does not. A claim about a named third party and any published price bring in people who were never in the content plan, so the page stalls without anybody deciding to stop it.
What a shortlisting buyer opens the page for, and what most versions give them:
| What the buyer came for | What most pages publish | What to publish |
|---|---|---|
| Where the other product is genuinely better | A grid with every tick in one column | The named case where they should buy the competitor |
| The price, or the model behind it | A link to a contact form | The number, or the model in one sentence |
| Evidence the comparison is current | An undated page | A review date, and a source against each claim |
| What switching costs them | Silence | The records that have to move, not the screens |
| A difference they can repeat internally | Pairs of adjectives | One mechanism, described plainly |
What has to be on the page?
A table. Answer engines lift structured rows straight into an answer, and this is the query where a buyer wants rows rather than prose. An audit of 200 published essays on this site found one table across the whole library, which is a fair description of writing nobody can extract from.
The GEO study from Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi (KDD 2024) measured a 40% lift in AI-answer visibility from adding statistics and 30 to 40% from citing sources, and found 44.2% of AI citations came from the first 30% of a page. On a comparison page that puts the table above the argument rather than under it.
How do you keep it defensible?
Record where each claim came from and the date you checked it. This is the one page that makes assertions about somebody else’s product, and those assertions start ageing the moment the other company ships a release.
A batch of twenty AI-written posts we commissioned came back structurally clean, and an adversarial audit after the repair pass still found invented supporting detail, because the source file held one-line entries and the writer closed the gap. On a page that names a competitor that is the expensive kind of wrong.
One account keeps 44 content pages behind an automated validator that runs before publish. It was added after the fact, which is the normal order and the costly one.
What do you say about the competitor’s strengths?
Name the buyer they are right for. A page arguing that a rival is wrong for everybody gets read by people who know at least one company it suits, and the table loses credibility on that one paragraph.
That sentence gives away deals that were never yours, and it is what makes the other rows believable.
The sentence that earns the rest of the page is the one where you tell a buyer to go and buy the other product.
How do you know the page is doing anything?
Not from position alone. A page can be shown and skipped: on one account the average organic click-through rate ran at 0.11% across a month while positions held steady. A comparison page with a title written for a crawler loses the click to whoever wrote a better one.
It can also be used without being visited. Citations into AI answers need their own report, because a page can supply the answer somebody acts on and appear in the traffic report as almost nothing.
The structural half of this is covered in why a library without tables is invisible and the gate that catches a wrong claim before publish.
Frequently asked questions
What is a competitor comparison page?
A page on your own site comparing your product against a named competitor, built to answer the search query that already pairs both names. Buyers reach it while shortlisting rather than while learning.
Should you name competitors on your website?
Yes, on a page built for the query that already pairs your names. That query gets answered by whoever wrote a page for it. The risk lives in unsourced claims, not in the naming.
What should a comparison page include?
A real HTML table rather than prose, the price or the pricing model, a review date, a source against each claim about the other product, and one honest statement of who should buy them instead.
How do you write about a competitor without getting it wrong?
Source every claim to their published material and record the date you checked. Describe what their product does rather than what it fails to do.
How often should a comparison page be updated?
On a fixed review cycle, and after any release on either side. Put the last-reviewed date on the page. A buyer who finds one wrong row stops trusting the rest.
Why do comparison pages get blocked internally?
They collect approvals that ordinary content does not. A claim about a named third party and any published price bring in people who were never in the content plan.
Sources
- Chua Network delivery data across 8 client accounts (internal fact bank)
- Chua Network engagement records, anonymized (internal experience bank)
- GEO study, Princeton, Georgia Tech, Allen Institute for AI and IIT Delhi, KDD 2024