Dictating a Prompt for a Comparison Table

A spoken method for building a comparison table that supports a decision.

Voiskey Voiskey Team
Voice PromptingChatGPT WorkflowDictating Prompt Comparison

Fix the Options and the Decision

Dictating a Prompt for a comparison table works best when the options and the decision are fixed first. Say how many options are in scope, what they are, and who will choose between them. Then say what the table must help decide: which option to buy, which vendor to shortlist, which approach to test first. A table without a decision becomes a wall of cells. A table with a decision becomes a page a reader can act on. Speak the options and the decision before you mention a single criterion, because the decision decides which criteria matter. A single named decision also keeps the option list from growing during the conversation.

Keep the option list closed and explicit. Name each option exactly as it should appear, including version numbers or plan labels if they matter. Say whether the list is final or whether one more option may be added. If the options are not comparable, say which dimension makes them comparable and which differences should be ignored. A table that mixes a small tool with an enterprise platform produces misleading rows. Comparability is a decision you make before the table, not something the table discovers on its own. Write the list at the top of the prompt so the assistant treats it as closed.

Name the reader and the level of detail they need. An executive reader wants three criteria and a recommendation; an evaluation team wants ten criteria and notes on evidence. Say which reader the table serves and how much supporting text belongs beside it. If the table will be attached to a decision memo, ask for a short recommendation line under the table. If it will be used in a working session, ask for the evidence column. Reader and detail level decide the table's size before any content is drafted, so state them early. If the table serves two readers, make two tables rather than one crowded page.

Set the rule for a tie. Comparisons often end with two options that score the same on the stated criteria. Decide in advance what breaks the tie: cost, time to start, support quality, or a must-have requirement that only one option meets. Say the tie-breaker in the prompt and ask for it to be marked. A stated tie-breaker prevents the table from ending in a shrug. It also forces you to name the criterion you actually care about, which is often the most useful part of the exercise. A tie rule stated early also stops the discussion from reopening every criterion.

Choose Criteria Before Looking

Choose the criteria before you look at the options in detail. If you choose after, the criteria quietly bend toward the option you already prefer. Speak the criteria as a list and say which are must-haves, which are nice-to-haves, and which are disqualifiers. Keep the list to a number a reader can scan, usually five to seven. A longer list dilutes the comparison; a shorter list hides the trade-offs. The criteria are the argument of the table, and they deserve to be stated before the cells are filled. Say which criteria are non-negotiable before you describe any of the options in detail.

Define each criterion in observable terms. Fast is not a criterion; pages that load in under two seconds is. Easy is not a criterion; a new user can complete the task without training is. Say the test or the evidence that would show each criterion is met, and ask for that evidence beside the rating. Observable criteria make the table checkable, and they expose options that only look good because the standard was vague. A short definition beside each criterion keeps the comparison honest. Ask for the evidence to be quoted, not summarized, so a reviewer can check it quickly and easily.

Assign weights only if the decision needs them. A weighted score can clarify a trade-off, but it can also hide a disqualifier behind a good average. If you want weights, say the scale and the weights explicitly, and ask for the raw ratings to remain visible. If you do not want a single score, say so and ask for a table with a recommendation and reasons instead. The choice between a score and a reasoned table is a decision about how the reader should think, not a formatting preference. If weights are used, show the weighted and raw scores in separate columns.

Add the criteria that are easy to forget. Total cost over the expected period, time to get started, the effort to switch later, and the support available when something breaks are often missing from first drafts. Say which of these matter to your decision and ask for a row or a column for each. A table that compares features but ignores switching effort can recommend a choice the team will regret. The forgotten criteria are often the ones that decide the outcome. Add the criterion that would make you reject the leading option, even if it scores well everywhere else.

Dictate the Table Shape

Describe the table in words before you ask for content. Say the number of columns, what each column represents, and the order. A common shape is options as rows and criteria as columns, with a rating and a note in each cell. Another is criteria as rows and options as columns, which is easier to read when there are many criteria. Say which shape you want and how the first column should be labeled. The shape decides how the reader scans, so it is worth one clear sentence. Draw the table on paper first if the columns are hard to describe.

Specify the cell content and the rating scale. Say whether each cell holds a short phrase, a number, a yes-or-no mark, or a note with evidence. If you use a scale, define the endpoints in words and keep the number of points small. Ask for a source or an evidence note beside any rating that is not obvious. A table with unexplained numbers looks precise and is often unverifiable. A table with short notes and sources is less tidy and far more useful. Ask for a dash or a question mark where evidence is missing instead of a confident guess.

Ask for a summary row or a short verdict under the table. The summary should state which option leads, on which criteria, and what would change the recommendation. Say the length: one sentence, three bullets, or a short paragraph. A verdict without reasons is an opinion; reasons without a verdict are homework for the reader. Ask for both and keep them separate from the cell content. The combination gives a reader a fast path and a way to check the reasoning. The verdict should name the runner-up, the margin between the top two, and the condition that would flip the result.

Dictate the prompt as a sequence of blocks. Say the decision, the options, the criteria, the table shape, the cell rules, and the verdict format. Open the window where you will review the answer, press the Voiskey shortcut, and speak the blocks in order. Voiskey places the text at the insertion point and removes filler words, stutters, and self-corrections. It also learns names, jargon, acronyms, and spelling preferences, so option names and criterion labels stay consistent. Read the prompt once for column count and scale definitions before the assistant responds. Check the block order once; a missing block produces a lopsided table.

Check Cells Against Sources

Treat every cell as a claim with a source. Pick the rows that carry the most weight and check each one against the original material: the specification, the quote, the test result, or the conversation note. If a cell has no source, mark it as unverified rather than leaving it blank. A blank cell reads as missing information; an unverified cell reads as a warning. Check the numbers first, then the labels, then the notes. A table is only as good as its weakest checked cell. Mark the two cells a reader will look at first, and verify those before the rest.

Look for cells that compare different things. A price for one option may cover a year while another covers a month; a speed figure may come from a different test. These mismatches are common and easy to miss. Ask for the unit and the basis beside every number, and check that the units agree across a row. If they do not, convert or footnote them. A table that compares unlike units is worse than no table, because it looks authoritative while misleading the reader. Put the unit in the column heading so every row is read against the same standard.

Check that the table answers the decision from section one. Read the verdict and ask whether the criteria that matter most are visible in the cells. If the recommendation depends on a criterion that is not in the table, add it or remove the recommendation. If the table supports two options equally, say so and use the tie-breaker. A comparison that does not resolve the decision has failed its purpose, even if every cell is correct and every note is careful. Write the reason for the verdict in the table notes so the reader can see why it follows.

Record the check and the version. Note who verified the cells, when, and against which source. If a source changes later, the note tells you which cells to revisit. Keep the table and the notes together in one document, and date the version. Anyone who wants to run this workflow can download Voiskey and dictate the decision, options, criteria, and shape into the document where the table will live. Keep a short verification note beside each table you share, with the date, the checker, and the source used for the most contested numbers in each row. Share the note with the table.

Reuse the Table Prompt for New Shortlists

Save the prompt as a template for the next comparison. Keep the decision line, the criteria definitions, the table shape, the cell rules, and the verdict format. Leave slots for the option names and the evidence. A good template turns a two-hour comparison into a twenty-minute update when a new option appears. Test the template on a second shortlist before you rely on it. If the criteria no longer fit, change them deliberately and note the change. Name the template by the kind of decision so the right one is easy to find. Keep the slot list at the top so the blanks are obvious.

Keep a library of criteria for common decisions. A vendor comparison, a tool comparison, and a hiring comparison each have a stable core of criteria that rarely change. Write those cores down and reuse them, adding only the criteria specific to the current decision. A criteria library speeds up the prompt and makes comparisons across time more consistent. Review the library each quarter and remove criteria that never influenced a decision. Add a line for the criterion that was hardest to verify last time, and a note on where to find the evidence. Review the library with one recent decision in hand.

Share the template with the people who make similar decisions. Put it in the team space with one filled example and a short note on the tie-breaker rule. The team plan offers shared vocabulary and workflows, which keeps option names and criterion labels consistent across tables. Ask each person to run the template once and report the cell that was hardest to verify. Those difficult cells reveal where the criteria need a clearer definition. Ask for one example of a cell that was easy to verify and one that was not. Compare the two examples before changing the criteria definitions.

Review the results of past tables against the choices that were made. If an option that scored lower was chosen, find out why; the reason may be a criterion the table missed. Add it to the template and the criteria library. This feedback loop is how a comparison prompt improves over time. To start, download Voiskey, dictate one real comparison into the document you already use, and check every cell before you share it. Note the reason in one line so the next table starts from a better criterion list. If the same reason appears twice, add it to every template.

FAQ s

Voiskey is an AI voice typing app that works in the document window you already use. Press a shortcut and speak the decision, options, criteria, table shape, cell rules, and verdict format; the prompt appears at the insertion point. Filler words and self-corrections are removed, so a structured brief arrives as clean instructions.

Voiskey learns names, jargon, acronyms, and spelling preferences. Say each option name and criterion label once, and it keeps your spelling across the brief and later prompts. Check the drafted table for drift, then correct the term in the template so future comparisons use the same labels.

Yes. Voiskey removes filler words, stutters, false starts, and self-corrections, and it adapts punctuation and formatting, so a spoken list of criteria arrives as an ordered list rather than a run-on paragraph. Read the prompt once before the assistant responds, and confirm the count of criteria and columns.

Voiskey supports more than one hundred languages and translation. You can dictate the comparison brief in one language and ask for the table in another, keeping option names and ratings consistent. Review translated criterion labels with a fluent colleague before sharing the table, because a label can shift the meaning of a whole column.

Say less. Get more
Say less. Get more
Say less. Get more
Say less. Get more
VoiskeyVoiskey

Your voice, amplified.

Download Voiskey for Mac and Windows and dictate your next comparison brief.

Use it across macOS, Windows, iOS, and Android

Use it across