All packs

Kit · 32 pages · 10 prompts + house style

Be the answer AI gives about your product

Assistants answer questions about you before anyone reaches your site. Find out what they're saying, and give them the right answer to quote.

After using this: You'll know what AI assistants tell people about your product, you'll have fixed the wrong answers at the source, and you'll rerun the same test every month to see what moved.

Who this is for

You know people are asking ChatGPT and Claude about your product and getting answers you didn't write. You want your own pages to be the ones they quote.

Not for you if

You're hoping there's a form to submit or a ranking to buy. There isn't one. This is about making your pages worth quoting.

$89

One-time purchase · HTML and PDF

License

Instant delivery after payment.

Also in a bundleSave $97Make your help center AI-ready

How it works

Prompt 1 builds a test set of the questions people ask AI assistants about your product, and prompt 2 runs it and scores the answers. Then you trace each wrong answer to its source, rewrite the pages it should have quoted, and fix contradictions in the right order. You rerun the same test set every month and report what moved.

What's inside

  1. 01How AI decides who to quote
  2. 02What makes an assistant quote you, from the inside
  3. 03What you can and can't control
  4. 04The 5 properties of a citable page
  5. 05Your citation test set
  6. 06When the internet is wrong about you
  7. 07Fixing it in the right order
  8. 0810 prompts, from building your citation test set to the monthly citation report
  9. 09Your citation house style
  10. 10The short version

See the pages

3 pages from the PDF you download, as they are. Select one for a closer look.

Preview a prompt from the pack

The full prompt, exactly as it appears in the pack. Nothing removed.

You are an expert at writing documentation that AI systems can quote accurately. The article: [paste in full] The questions this article should be the answer to: [paste, from your test set] For each of those questions: 1. Quote the exact sentence you would lift to answer it. If there is no single sentence that answers it, say so explicitly, because that is the finding. 2. Where an answer is spread across several sentences or sections, show me how it is scattered and give me the one consolidated sentence it should be. 3. List every hedge in this article that removes a fact, such as "usually", "typically", "should", "may vary", "in most cases" and "fairly quickly". Replace each hedge with the specific answer or the condition. If the thing really varies, give the range and what it depends on. If you cannot tell the answer from what I gave you, keep the original wording and mark it [TODO: confirm] rather than guessing. 4. List every fact that a reader would want as a number or a limit and that is currently described in words. 5. Flag anything so qualified that a machine summarizing it would produce something wrong. Do not remove real uncertainty. Where something really does vary, the fix is to state what it depends on and give the range, not to invent a single number. Don't invent any fact I didn't give you. Where you'd need one, write [TODO: confirm]. At the end, list only the facts you needed and didn't have, not style choices, things to double-check, or anything I already gave you.

The thinking behind it

Want the thinking behind this kit? The course Deciding what goes where covers it in short parts, with an exercise in each.

Questions

Be the answer AI gives about your product, $89

Instant delivery after payment.