Do I need to paste the full paper or just the abstract?

The abstract is usually sufficient for generating lay summaries and social media content. However, for more accurate plain-language numbers (especially if the abstract doesn't include effect sizes), include the key results from the paper's Results section as well. More input = more accurate output.

The AI used the phrase "statistically significant" — is that okay?

No — that violates Rule 1. Add a follow-up prompt: "Replace all instances of 'statistically significant' with plain-language equivalents as required by the De-Jargon Guard." The correct phrasing depends on context: "the difference was unlikely due to chance" or "the pattern was consistent and reliable."

Can I use this output directly for my institution's social media?

Use it as a strong first draft, but always have a communications professional or PI review before posting. Verify that all numbers exactly match your published paper. Check your institution's policies on social media posting about research.

My study found a null result (no significant difference). Can I still use this?

Yes — null results are important and often under-communicated. Paste your abstract and specify in KEY_NUMBERS that the finding is a null result. The AI will frame it appropriately: "This large study found no evidence that X affects Y — an important finding that challenges assumptions in the field."

Should the WeChat section always be in Chinese?

Yes — WeChat content is platform-specific and serves the Chinese-speaking public. Even if OUTPUT_LANGUAGE is set to English, the WeChat section will be generated in Chinese. All other sections follow the OUTPUT_LANGUAGE setting.

The AI output contains words like "delve" or "landscape." What should I do?

These are AI-isms that signal machine-generated text to experienced readers, undermining credibility with medical and scientific audiences. Add a follow-up prompt: "Rewrite removing all AI-Isms Blacklist words (Rule 6): delve, tapestry, realm, beacon, crucial, vital, multifaceted, game-changer. Replace lazy transitions (Furthermore, Moreover, In conclusion) with specific connective language. Vary sentence lengths — mix short punchy sentences (3–5 words) with longer ones to break uniform AI rhythm." These patterns emerge from RLHF training and require explicit suppression.

The AI added background information about the disease that was not in my abstract. Should I keep it?

No — this is the Elaborative Explanation Phenomenon (Rule 5 violation). AI adds unrequested context to "be helpful," but that content may be inaccurate, outdated, or implicitly attribute claims your paper never made. Delete any information not explicitly stated in your abstract and add a follow-up prompt: "Remove all background information not explicitly stated in the provided abstract. Base every sentence only on the facts in the source text."

Do I need to disclose AI assistance when submitting the finished work?

Yes — ICMJE (2023) guidelines and most major journals (NEJM, Lancet, Nature, JAMA, BMJ) require authors to disclose AI-assisted writing in the Acknowledgements or Methods section. AI tools cannot be listed as authors. A typical disclosure: "AI-assisted writing tools [specify name] were used in the preparation of [specify sections] of this manuscript. The authors reviewed, revised, and take full responsibility for all content." Check your target journal's specific policy, as requirements vary.

Lay Summary & Social Media Generator -- medPrompt

Usage Guide

Fill in your target platform(s), audience, and any key numbers to highlight.
Click AI Run — the assistant will introduce itself and ask you to paste your paper's abstract or key findings in the chat.
Share your abstract in the chat.
Receive lay summaries and social media posts tailored to your audience — ask for revisions as needed.

Medical Research Assistant

Fill in variables and run directly with AI

Wiki

Communicating research findings beyond the scientific community is increasingly expected — and in many funding frameworks, required. But translating clinical findings into plain language for the general public, patients, or policymakers carries its own risks: overclaiming, stripping essential nuance, or introducing background not supported by the original data. This tool provides a structured framework for converting medical research papers into accurate, platform-optimized lay summaries and social media content.

Six guards are enforced throughout all output. The De-Jargon Guard prohibits statistical and clinical terminology inappropriate for lay audiences — terms like "hazard ratio," "p-value," and "confidence interval" are replaced with plain-language equivalents. The Anti-Overclaim Guard blocks language that exceeds the evidence: "cure," "breakthrough," and "proves" are replaced with accurate hedges. The Number Clarity rule enforces absolute numbers over relative risks wherever possible, since relative risk language systematically misleads general audiences. The Limitation Honesty rule requires at least one limitation from the original paper to be mentioned explicitly.

The Absolute Fidelity Guard prevents the AI from adding background information not stated in the source abstract — a common AI failure mode where the model helpfully expands context with potentially inaccurate or outdated information. The AI-Isms Blacklist bans a curated list of high-frequency AI vocabulary ("delve," "tapestry," "realm," "crucial," "multifaceted") that signals machine-generated text to experienced readers and undermines scientific credibility.

Content is generated for four platforms with distinct format requirements. Twitter/X threads are structured as 5-8 numbered tweets, each under 280 characters, with a hook tweet optimized for engagement. LinkedIn posts follow a professional narrative structure suitable for a health-professional or researcher audience. WeChat content is generated in Chinese regardless of OUTPUT_LANGUAGE, structured for China's social media conventions. ResearchGate summaries follow the plain-language abstract format used by the platform.

This tool is appropriate for communication professionals, principal investigators preparing grant impact statements, and researchers required to submit lay summaries to funders such as NIH, Wellcome Trust, or NIHR. It is not a substitute for institutional communications review — always verify that numbers exactly match the published paper and have a subject-matter expert review the output before posting.

FAQ