Skip to main content

High-impact AI in visualization and images

The guidance below highlights impactful applications of AI when seeking to visualize data or create imagery, with practical considerations and prompt templates you can copy and paste to adapt to your own needs.

Aim: Improve clarity without altering meaning.

Data Analytics Statistics Technology Information Concept

High-impact use

  • Improve figure layout and readability without loss of detail (panel order, consistent axes, clear labels, legible text, clean legends, less clutter).

  • Check for misleading choices (axis truncation, wrong scale type, distorted aspect ratios, color scales that exaggerate differences).

  • Generating conceptual diagrams or schematics

  • Suggest better chart types for the same data (e.g., distribution plots instead of bars; raw data + intervals instead of mean-only).

  • Flag practical readability issues and address.

Safety check

  • AI may assist presentation — it must never alter evidence.

  • Check that AI-assisted edits do not introduce artefacts, distortions or invented detail.

  • Manually verify each element and keep a record of what was generated vs confirmed.

  • Treat conceptual diagrams as claims: ensure every arrow and label is evidence-aligned, appropriately caveated (supported vs hypothesized), and no stronger than the results/limitations in the manuscript (avoid implied causality where evidence is correlational).

  • Retain original, unedited versions.

  • Do not alter raw data images beyond accepted journal standards.

  • Do not use AI to fabricate experimental images.

  • Clearly label AI-generated conceptual figures.

  • Follow journal and funder disclosure requirements.

Copy and paste prompts

📑 Copy and paste prompt:

Review the following figure (description and/or image).

1. Briefly state the main message the figure appears to communicate.

2. Identify any risks of misinterpretation related to axes (truncation, inconsistent scales, inappropriate scale type), aspect ratio, colour use (contrast, gradient meaning, accessibility), legend clarity, overplotting, or omitted uncertainty.

3. Flag any elements that could unintentionally exaggerate, downplay, or imply causation beyond the data.

4. Check whether uncertainty, sample size (n), and definitions of error bars are clearly visible.

5. List the top five specific, practical fixes that would improve clarity and integrity without changing the meaning of the data.

Do not suggest altering the underlying data, statistical analysis, or results. Focus only on presentation quality, transparency, and interpretability.

📑 Copy and paste prompt:

Create a [conceptual diagram / schematic / visual layout improvement] for [topic/study area].

This figure is illustrative only and must not fabricate or modify underlying data.

Do not invent experimental results, measurements, or images.

If improving an existing figure: preserve all data values, axis scales, labels, and proportions exactly as provided.

Clearly label the output as conceptual / illustrative / AI-generated, where appropriate.

⚠️ Unsafe copy and paste prompt example:

Generate a realistic image/figure that shows [desired result] and makes the findings look clearer or stronger.

Why this is unsafe:

- Encourages fabrication of results.

- Encourages exaggeration.

- Blurs the line between illustration and evidence.

Explore all research stages