Write a non-exaggerated talk track for data charts

Author: AILesson6 min setupTested with:ChatGPTReviewed: 2026-08-28

Quick answer

Explain what a chart shows, what it does not show, and what decision it can reasonably support. Provide: Chart data and encoding, Method and limitations, Audience and decision context. Expected result: A concise chart narration with context, comparison, limitations, and an evidence-aligned takeaway.

Add your context

Your text stays in this browser. AILesson Prompts does not send it to a model or server.

Chart data and encodingMethod and limitationsAudience and decision context

Your prompt

Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt

Write a spoken explanation for the supplied data chart.

Chart title, type, encoding, values, denominators, filters, missingness, and source:
[chart]

Period, sample, comparison, uncertainty, revisions, and causal limits:
[method]

Audience, time, question, decision, and prohibited claims:
[context]

Audit the chart before narrating it: verify units, denominators, axis baseline/range, category order, aggregation, time intervals, sample size, missing values, and whether the visual implies a comparison the data cannot support. Report any material chart problem first and provide corrected wording; do not normalize misleading design through careful narration alone.

Write a natural talk track in this order: orient the listener to measure/time/population; state the most decision-relevant pattern with exact magnitude and comparison basis; identify a meaningful exception or uncertainty; explain the supported implication; state what the chart cannot establish; connect to the next decision or slide. Separate percentage points from percent change, correlation from causation, forecast from actual, and count from rate. Include a 15-second version, a full timed version, accessible text alternative, likely misinterpretation with correction, and a literal claim-to-cell/source audit.

Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.

From input to outcome

A worked example

See how concrete context turns this recipe into a usable result

Actual input

Chart data and encoding: Clustered bars titled 'Reservation collection before expiry'. Prior period: 107/198, 54%; reminder pilot: 131/214, 61%. Y-axis starts at 50% and ends at 65%. Labels show 54 and 61 but not n. Source cells Pilot!B4:C5; adult patrons who opted in only. Six records with unknown collection status were excluded: 2 prior, 4 pilot.
Method and limitations: Two consecutive six-week periods in 2026, not randomized. Seasonal title mix differed. Percentages use known-status records as denominators. No confidence intervals calculated. Pilot log revised 26 Aug after duplicate removal. This comparison cannot establish that reminders caused the difference.
Audience and decision context: Library director, 45 seconds. Decide whether to fund design of a randomized follow-up test. Avoid 'increase', 'improved', 'worked', or 'significant'. Next slide covers opt-outs and staff workload.

Example output

Chart audit: the truncated 50-65% axis visually magnifies a seven-percentage-point gap, and missing n labels hide different denominators. Redraw from zero or use a dot plot with the full meaningful scale; label 107/198 and 131/214, plus six excluded unknown-status records. Do not present until the denominator note is visible.

15-second version: “Among opted-in adults with known status, 61% collected in the six-week pilot period versus 54% in the prior six weeks—a seven-percentage-point difference. The periods were not randomized, so this is a reason to test, not evidence that reminders caused the difference.”

Full version, about 40 seconds: “This compares opted-in adult patrons with known collection status across two consecutive six-week periods. In the prior period, 107 of 198 reservations—54 percent—were collected before expiry. During the reminder pilot, it was 131 of 214—61 percent. That is a seven-percentage-point difference, or about 13 percent relative to 54, but the percentage-point expression is clearer here. Six unknown-status records were excluded. Because the periods were not randomized and the title mix differed, the chart cannot tell us that reminders produced the difference. It supports designing a stronger test. Before that decision, the next slide shows the opt-out and staff-workload guardrails.”

Accessible alternative: “Table: prior 107/198, 54%; pilot 131/214, 61%; known status only; 2 and 4 unknowns excluded.” Likely misreading: “collection improved by 7%.” Correction: it differed by 7 percentage points; causation is unknown. Audit: 107/198 and 131/214 map to Pilot!B4:C5; all caveats come from the supplied method.

Why this works

1
A fixed narration order makes denominators and limitations part of the explanation rather than optional footnotes
2
Claim-to-source auditing catches persuasive phrases that silently exceed the plotted data

Check the result

Are measure, population, period, denominator, and comparison basis spoken clearly
Are percentage points, percent change, causality, forecasts, and uncertainty handled correctly
Does the takeaway support the requested decision without going beyond the chart

Use it with confidence

Frequently asked questions

Practical answers about when to use this recipe, what to provide, and where human review still matters

What should I prepare before using “Write a non-exaggerated talk track for data charts”?

For “Write a non-exaggerated talk track for data charts,” prepare Chart data and encoding, Method and limitations, and Audience and decision context. Replace placeholders only with information you can verify. If a detail is unknown, preserve that uncertainty explicitly instead of asking the model to infer it.

When is the “Write a non-exaggerated talk track for data charts” result not ready to use?

The result is not ready if it does not yet deliver the stated outcome—A concise chart narration with context, comparison, limitations, and an evidence-aligned takeaway—from the supplied evidence, or if it relies on unresolved assumptions, missing approvals, or invented details. Use the checks as release gates: revise the source inputs or assign a named, authorized reviewer instead of polishing an unsupported output.

Which AI tools have recorded tests for “Write a non-exaggerated talk track for data charts”?

The published test record for “Write a non-exaggerated talk track for data charts” lists ChatGPT as of 2026-08-28. This confirms recorded runs, not guaranteed compatibility or identical results in later product versions. For another tool or version, keep every constraint visible and repeat the result checks before use.

More ways to explore