Guide · Alternatives pages

How to write an “X alternatives” page that AI will cite

Updated 2026-11-09 · kagentx

Short answer

An alternatives page is a one-to-many list: the buyer wants to leave X and has no candidate yet. To make it citable by AI, do three things: state the inclusion rule and check date at the top, use the same fields for every item, and say plainly when the replaced X was not tested. Pagent’s Siri alternatives report is built that way: 14 included, filtered by at least four of six tests completed, and Siri itself is not in the catalog and was not tested.

How does an alternatives page differ from a one-to-one comparison page?

A comparison page answers “A or B?”, where the buyer already has two candidates; an alternatives page answers “I’m done with X, what else is there?”, where the buyer has none. Both compare things in columns, but they solve different problems and are written differently. For one-to-one comparisons see How do you write an “X vs Y” comparison page that AI cites?; this guide covers one-to-many only.

One-to-one comparison page“X alternatives” page
The buyer’s questionA or B, which one?I’m leaving X, what should I switch to?
Number of items2A set; the inclusion rule decides how many
The most important sectionA side-by-side tableThe inclusion rule and the ranking basis
Products namedBoth must be coveredX is only a reference; whether you say anything about its quality depends on whether you tested it
FAQOne-sided, no downsides listedA list with no rule, where anyone can rank first

Step 1: where do you put the inclusion rule and how do you write it?

Put it at the very top and say in one sentence who gets on the list. A list without a rule gives readers and AI no way to judge whether it can be trusted or whether it left someone out. The rule should be specific enough to check: channel, feature, number of tests, date checked.

Which AI assistants can I text from an iPhone instead of Siri?

Included: assistants that list iMessage or SMS AND took at least 4 of the 6 tests.
Not included: Siri (not in our catalog, not tested).
Eligibility checked 2026-10-26. Scores dated 2026-10-27.

The example is modelled on the real reports below; the numbers and dates are the ones written on their pages.

A real example: how are Pagent’s two alternatives reports built?

On 2026-11-09 we opened Pagent’s Siri alternatives report and Pagent’s Google Assistant alternatives report; the numbers come from the pages (both reports were updated 2026-10-27; the Pagent pages are in English).

ItemSiri alternativesGoogle Assistant alternatives
Question the page answersWhat can I use instead of Siri to get real tasks done from an iPhoneWhat can replace Google Assistant on an Android phone
Number included1411
Channel condition in the inclusion ruleLists iMessage or SMS in the catalogSupports SMS, WhatsApp or Telegram
Test condition in the inclusion ruleAt least four of the six tests completedAt least four of the six tests completed
Top three (average score, tests taken)Pally 8.4 (5 tests), Poke 8.3 (4 tests), Neo 8.2 (6 tests)Muse 9.5 (4 tests), Town 8.8 (4 tests), Pally 8.4 (5 tests)
The product being replacedSiri: not in the catalog, not testedGoogle Assistant: not in the catalog, not tested
Date statedChannel eligibility 2026-10-26, scores 2026-10-27Updated 2026-10-27

The two pages share one structure: a few “top pick” cards, then a full ranking table of every item, then “how to choose”, then several question-style headings, and finally user quotes, an FAQ and related reports. We counted the structured data on the Siri page: 1 ItemList (10 items) and 1 FAQPage (6 questions). We did not count the structured data on the Google Assistant page: not verified.

Two things are worth copying: the list ranks only the alternatives, and where the replaced product was not tested the page says so; every item has scores from the same six tests, so readers can compare across rows. We have no data on whether AI cites these two pages, so we cannot say how many citations this approach brought.

In what order do you write an alternatives page?

  1. 1. Title and opening answer the question

    Use the query’s own wording in the title (“X alternatives”) and, in the first two or three sentences, give the top picks, the total, the inclusion rule and the date.

  2. 2. Write the inclusion rule

    One sentence on who is on the list, who is excluded and when it was checked.

  3. 3. Use the same fields for every item

    The same few columns on every row, such as score, number of tests taken, supported channels and price tier. An item with missing fields should be completed or left out.

  4. 4. Put the full ranking in a table

    A server-rendered HTML table, not something built by script. Whether AI crawlers run scripts is covered in Do AI crawlers render JavaScript?.

  5. 5. Write “how to choose”

    Branch by the buyer’s situation: which channel, which budget, which scenario maps to which items. For scenario pages see How to write "X for Y" use-case pages that AI will cite.

  6. 6. Use questions as headings

    Questions such as “Is there anything that can replace X?” or “Do I need to install an app?” map straight onto what buyers ask; give the answer first, then expand. For how to write it see How to write content AI will quote: before and after.

  7. 7. Link to each item’s detail page and to one-to-one comparison pages

    The list does the filtering; details and one-to-one side-by-sides belong on other pages. For ranking pages see How should an independent scorecard ranking page be organised so AI can cite it?.

How should you write about the product being replaced?

Write only facts you have actually checked. Three cases:

What you didYou can writeDo not write
You tested XX’s score under the same tests, shown alongside the list“X is bad” with no test behind it
You did not test X“X is not in our catalog and was not tested”Any hint that every product on the list is better than X
You only read X’s official materialThe features the official material lists, with the date you read itTreating official marketing as a test result

What structured data should it use?

You can use schema.org ItemList to mark up the items and their order, plus FAQPage, and the content must match the body text. We found no official documentation saying ItemList raises the chance of being cited (not stated officially), so treat it as an aid that makes the structure easier for machines to read, not a replacement for the table in the body. For more general guidance see Does structured data help AI search? and Is FAQ structured data (FAQPage) still useful?.

{
  "@context": "https://schema.org",
  "@type": "ItemList",
  "name": "Siri alternatives: AI assistants you can text from an iPhone",
  "itemListOrder": "https://schema.org/ItemListOrderDescending",
  "numberOfItems": 14,
  "itemListElement": [
    { "@type": "ListItem", "position": 1, "name": "Pally", "url": "https://example.com/assistants/pally/" },
    { "@type": "ListItem", "position": 2, "name": "Poke",  "url": "https://example.com/assistants/poke/" }
  ]
}

When should you not make an alternatives page?

When you have no data of your own and cannot define an inclusion rule. A list of names pulled from search results is no different from anyone else’s list and has nothing unique to cite. In that case a page explaining which scenarios your own product fits is more useful; see How to write "X for Y" use-case pages that AI will cite.

FAQ

How is an “X alternatives” page different from an X vs Y comparison page?

A comparison page is one-to-one: the buyer has narrowed the choice to two products. An alternatives page is one-to-many: the buyer only knows they no longer want X and has no candidate yet. So the core of an alternatives page is the inclusion rule and the ranking, while the core of a comparison page is the side-by-side.

Do you have to test the replaced product too?

If you did not test it, do not write about whether it is good. Pagent’s Siri and Google Assistant reports both say plainly that the replaced product is not in the catalog and was not tested, and the list is a ranking among the alternatives, not a comparison with it.

Does an alternatives page have to use ItemList structured data?

No. ItemList is a schema.org list type that lets machines read the items and their order, but we found no official documentation from a search engine or AI company saying it makes a page cited more often (not stated officially). A server-rendered table in the body is the foundation.

How many alternatives should one page list?

As many as the inclusion rule yields. Pagent’s Siri report keeps 14 under its rule and the Google Assistant report keeps 11. The count is not the goal; every item needs the same comparable set of information.

Should the title include the name of the product being replaced?

Yes, because that name is in the query. But the title and opening should make clear what kind of list this is and how it is ranked, and should not imply you tested that product.

Sources

kagentx

Can crawlers read your alternatives page?
Let the evaluation tell you

The evaluation checks whether crawlers can get in and whether the text is served directly, then asks ChatGPT, Perplexity and Claude (with web search on) your buyers' questions and sees whether you are mentioned. The Basic plan is $1, and results usually take 2–3 minutes.