Comparison and Alternatives Pages

These are the highest-value pages most companies never write. Someone searching "X vs Y" or "alternatives to X" has a budget, a shortlist and a decision to make this week. And they're disproportionately the pages AI assistants cite, because a model answering "what should I use for Y" is looking for exactly this kind of structured, comparative source.

They're also the pages companies write worst, because the temptation to win the comparison overrides the job of helping someone decide.

The rule that makes them work

Recommend the competitor where the competitor is genuinely better.

This isn't a moral position, it's the mechanism. A comparison page that concludes "we win on everything" is recognisable as marketing in about four seconds. Readers discount the whole page, and models synthesising across sources weight it accordingly against the independent reviews saying otherwise.

A page that says "if you need X, use them, we're not built for it" earns the credibility that makes the rest of the page believable. It also disqualifies bad-fit buyers before they become bad-fit customers.

The practical form: for every comparison, name at least one dimension where the other tool wins, and be specific about it.

The three page types

They're not interchangeable. Each targets a different searcher.

If you're the underdog, write all three targeting the market leader. If you're the leader, write "best X for Y" pages and let others write the versus pages.

X vs Y structure

H1        X vs Y: which should you choose in 2026?

          [Direct answer, 50 words. Who each is for. State it plainly
           up top — a reader who leaves here should have the answer.]

          [Comparison table: 6-10 rows, the dimensions that decide it]

H2        Choose X if…        [3-5 specific situations]
H2        Choose Y if…        [3-5 specific situations, genuinely]

H2        Pricing compared    [Real numbers, dated, with the gotchas:
                               per-seat vs flat, what's gated, what
                               the real cost is at 20 people]
H2        [Dimension 1]       [The thing that actually differs]
H2        [Dimension 2]
H2        Where X falls short [Yes, including your own product]
H2        Migration           [What switching actually involves, honestly]
H2        FAQ                 [The specific questions, answered directly]

          [Disclosure: we make X. Stated plainly, in the sentence
           where it matters, not buried in a footer.]

What makes it credible:

• Disclose ownership early and plainly. "We build X, so read this with that in mind — here's our methodology" costs nothing and buys everything. Hiding it and getting caught costs the page.
• Date everything. "Pricing as of September 2026." Comparison pages rot faster than anything else on a site.
• Be specific. "Y limits you to 5 GB on the free plan; X gives 15 GB but caps API calls at 1,000 a day" beats "Y has storage limitations".
• Use real screenshots of both products. Signals first-hand use, which is the thing a model can't fabricate and a reader can verify.
• Say what you actually tested. How long you used it, on what, in what version.
• Don't compare against a version from two years ago. It's the most common form of dishonesty on these pages and competitors will point it out.

Alternatives pages

The trap is writing "9 alternatives to X" where your product is number one and the other eight are strawmen. Everyone recognises it.

The version that works:

• List genuine alternatives, including ones better than yours for some use cases.
• Lead with why people leave X. That's the query intent. If you can't articulate it honestly, you don't understand the market.
• Give each alternative a real evaluation — what it's good at, who it's for, what it costs, where it falls down.
• Put yourself in a sensible position, not automatically first. A product listed third with an honest "best if you need Z" converts better than the same product listed first with a puff paragraph.
• Include a decision table or flow so someone can self-select.

Best X for Y

The highest-value format for AI citation, because it maps directly onto how people prompt an assistant: "what's the best X for Y".

• Narrow Y until the answer is genuinely different. "Best CRM for real estate teams under 10 people" has a different answer than "best CRM", and far less competition.
• State the criteria before the list. What you evaluated on, and why those criteria and not others.
• Rank, don't list. A list of nine tools with no ranking is a page that failed to decide, and it won't be cited because it doesn't answer the question.
• One clear pick, then alternatives for specific edge cases.
• Show the work — what was tested, for how long, what was excluded and why.

Why these pages get cited by AI assistants

A model answering a recommendation question is synthesising across sources. What it can use:

• Structured, comparable facts. Tables get parsed cleanly. Prose comparisons often don't.
• Explicit "choose X if" statements. These map directly onto a user's stated constraint and are quotable verbatim.
• Named entities with specifics — product names, prices, limits, versions.
• Balanced judgement. Sources that only praise get weighted against sources that evaluate.
• Recency signals. Dated content, current pricing, current version numbers.

And the constraint that undoes all of it: the content must be in the HTML. Comparison tables rendered client-side, or behind a tab that fetches on click, are invisible to retrieval agents that don't run JavaScript. See featured-snippet-passages for the passage-level mechanics.

Legal and ethical lines

• Comparative advertising is legal in most jurisdictions when the claims are truthful and substantiated. Keep evidence for every claim.
• Don't use a competitor's logo in a way that implies endorsement. Fair use for identification is generally fine; check the jurisdiction if it matters commercially.
• Don't misrepresent pricing or features. Beyond the ethics, it invites a complaint and it's trivially disproved.
• Update when they change. A page claiming a competitor lacks a feature they shipped six months ago is a liability, and it's the thing their sales team will screenshot.
• Trademark terms in ads have different rules from organic content. Organic comparison pages are fine.

Maintenance

These decay faster than any other page type. Put a quarterly review in the calendar:

• Pricing on both sides
• Features shipped since last review
• Screenshots still matching the current UI
• The "as of" date
• Whether the ranking still holds

A stale comparison page is worse than none, because it's confidently wrong in public.

Where the data comes from

Two inputs decide whether these pages get written for the right pairs: which competitors people actually compare you against, and which of those comparisons have demand. Guessing produces pages for pairings nobody searches.

Keupera covers the competitor and prompt side — who ranks and gets cited for these queries, and the actual prompts people put to LLMs at the comparison stage of the decision — plus AI Brand Radar to check whether the page changed which product gets recommended. Over MCP:

claude mcp add --transport http keupera https://mcp.keupera.com/mcp -s user

Related skills

best-of-listicle-pages, featured-snippet-passages, ai-visibility-audit, topical-authority-map, keyword-research-briefs, eeat-author-pages