Build a 'best X for Y' comparison that answer engines cite

For one category (X) and one buyer (Y) the person names: find the vendors buyers and answer engines weigh; choose only metrics a vendor can prove at a public address and read every value with its address and date; score the vendors with a vendor proof row of rules (tools/workflows/read-data/candidate-rules.json, classified by tool:classify-a-record), where a metric with no proof never counts; write a page whose quotable answer, structured comparison table (each cell linked to its proof), method and FAQ are in the served HTML with ItemList and FAQPage data, so answer engines can quote and cite it; keep sponsored placement in its own labelled block that never changes a score or a rank; publish only what the person approved, check it the way answer engines read it, hand it to workflow:keep-pages-cited-by-answer-engines and recheck the proof on a schedule.

What it asks for

  • site (Needed, words): Where the person publishes the comparison: their site's address or the id of their public collection
  • buyer (Needed, words): Y: who the comparison is for, in their own words (a team size, a trade, a need)
  • rules (Needed, words): The id of the person's writing rules every word is checked against
  • vendors (Optional, list): Vendors the person already knows buyers weigh, by name or address; the scope step adds those buyers and answer engines name
  • category (Needed, words): X: the kind of product compared, in buyers' words
  • sponsors (Optional, list): Paid placements the person sold: which vendor, where it shows and the disclosure words it carries
  • questions (Optional, list): The questions a buyer asks an answer engine before choosing; the scope step writes three when none are given
  • source_row (Optional, details): The person's own vendor proof row for this category (tools/workflows/read-data/candidate-rules.json shape: facts, keep rules and weighted signal rules for tool:classify-a-record); vendors-by-verifiable-proof of tools/workflows/read-data/candidate-rules.json when left out
  • recheck_days (Optional, number): Every proof is read again after this many days (30 when left out)
  • answer_engines (Needed, list): The search address of each answer engine the person wants to be cited by, with {question} where the question goes

The steps your AI walks

  1. Name the vendors buyers weigh and the questions they ask
  2. Choose the proof metrics: only numbers anyone can check
  3. Read each vendor's proof and copy every value with its address and date
  4. Score the vendors with the proof row
  5. Write the comparison page: a quotable answer, a proof-linked table, sponsored placement kept apart
  6. Publish the page once the person approves it, and check it the way answer engines read it
  7. Hand the page to the citation workflow and set the proof recheck

Use it from your AI

Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this workflow by what it does and runs it.

Say to your AI: “Use Trillion to for one category (X) and one buyer (Y) the person names: find the vendors buyers and answer engines weigh; choose only metrics a vendor can prove at a public address and read every value with its address and date; score the vendors with a vendor proof row of rules (tools/workflows/read-data/candidate-rules.json, classified by tool:classify-a-record), where a metric with no proof never counts; write a page whose quotable answer, structured comparison table (each cell linked to its proof), method and FAQ are in the served HTML with ItemList and FAQPage data, so answer engines can quote and cite it; keep sponsored placement in its own labelled block that never changes a score or a rank; publish only what the person approved, check it the way answer engines read it, hand it to workflow:keep-pages-cited-by-answer-engines and recheck the proof on a schedule”