Sign in Start free
AI SEARCH (GEO)

Build a comparison table assistants can lift

Use when you want a comparison page that assistants summarize accurately rather than mangle.

comparison-table-builder.md
Download .md 
You are building a comparison table for a page comparing {{ITEM_A}} and {{ITEM_B}} (and optionally {{ITEM_C}}).

Audience and their decision: {{AUDIENCE_AND_DECISION}}
Facts I can verify: {{KNOWN_FACTS}}

Produce:
1. A table with one row per decision criterion and one column per item. Choose 8 to 12 criteria that actually change the decision, not feature counts. Every cell must be a short factual statement, under 15 words, no marketing adjectives.
2. A "verdict by use case" section: 4 to 6 short paragraphs, each starting "Choose X if you ...".
3. A one-paragraph summary of 50 to 70 words that a reader could quote as the answer to "which should I pick".

Constraints:
- Any cell you cannot verify from {{KNOWN_FACTS}} must read "not verified" rather than a guess. Do not invent pricing, limits or dates.
- Do not declare an overall winner. Verdicts are per use case.
- Criteria must be phrased as neutral nouns, not as leading questions.
- No em dashes, no superlatives, no "the clear winner".

Fill in before running

Replace each placeholder with your own detail. The more specific you are, the less the model invents.

  • {{ITEM_A}}
  • {{ITEM_B}}
  • {{ITEM_C}}
  • {{AUDIENCE_AND_DECISION}}
  • {{KNOWN_FACTS}}

Getting a better result

  1. Assistants lift row labels as much as cell values, so make criteria names literal and searchable.
  2. Mark verified pricing with the date you checked it; stale numbers get cited for years.
  3. Publishing the table as real HTML rather than an image is what makes it liftable.

Questions about this prompt

When is this the right prompt rather than the citation gap one?

When you are building the comparison page. The citation gap prompt diagnoses an existing pair of pages after an assistant has picked the other one. Reach for this when the page does not exist yet and the risk is a summary that misstates one product because your own table was vague.

What do I need before running it?

A written list of facts you can actually verify, and one sentence on who is deciding what. Every cell you cannot support from {{KNOWN_FACTS}} comes back as not verified, so a thin fact list produces a thin table. The decision context is what keeps criteria to things that change the answer.

What comes back, and what is the useful bit?

Eight to twelve decision criteria as rows, a set of Choose X if you paragraphs, and a 50 to 70 word summary answering which should I pick. The criteria labels matter as much as the cells, since assistants lift the row name along with its value when summarising a comparison.

What is the mistake here?

Adding an overall winner because the per use case verdicts feel indecisive. A comparison that crowns a favourite reads as a sales page and gets discounted as one. The next most costly move is shipping the table as an image or a script-rendered widget, which cannot be quoted however good the cells are.