Review an AI-Written Product Comparison Without Inventing Features

A comparison table can look authoritative because every cell is filled. Empty cells are often more honest when the evidence is incomplete. Build the table from inspected sources and keep your judgement separate from the product facts.

Choose criteria before collecting claims

Write down the reader's task and three relevant criteria, such as supported input types, export requirements and account restrictions. Avoid a long feature list that has no connection to the reader's decision. Use the same definition for every candidate.

For a fictional pair of note-taking tools, one source might confirm PDF export while another says only “Export your notes.” The second statement does not establish PDF support. Mark the specific format as unverified until the vendor documents it or an authorised test confirms it.

Attach evidence to each cell

Record the source URL, the wording that supports the claim and the date you checked it. Specify edition or platform restrictions. If a trial exposes only some features, describe that scope instead of presenting the observation as a complete product audit.

Ask an assistant to turn your evidence into a readable table with “Unknown” for missing fields. Instruct it not to fill gaps from memory. Review every row against your notes after it returns the table.

Write the recommendation last

Explain which criterion matters most for the stated task and where uncertainty could change the choice. Do not invent a numerical benchmark or imply you personally tested a product when you only read its documentation.

Before publication, recheck volatile details such as prices, limits and availability directly. If you cannot verify them, omit the claim or label the gap. A useful comparison helps a reader understand the trade-off without pretending the evidence is complete.