AcademyHealth Action Resource: Improving Research Communications (With a Little Help from AI)
A practical guide to using AI as your first draft partner, not your final say.
Health services researchers are being asked to do something harder than ever right now: make the case for evidence at a moment when trust in science and public institutions is under pressure, funding is uncertain, and the information environment is crowded and contested.
Your findings matter. Getting them in front of the right people, in language that actually lands, has never been more urgent or more difficult.
You don’t have to be a communications expert to do this well. But you do have to know your audience, and that’s where most researchers get stuck. AI tools like ChatGPT can help. Not by doing the work for you, but by giving you a faster, more concrete starting point: a realistic picture of who you’re talking to and a first draft you can react to and improve.
If you’ve never used AI this way before, or if you’re skeptical, that’s fair. This resource explains what these tools can and can’t do, and how to use them responsibly.
Why Consider AI for This?
Most researchers are not starting from zero when it comes to communications. You know your findings. You understand the stakes. What AI offers is a thinking partner that can help you translate that expertise for a specific audience, faster than staring at a blank page.
AI tools are especially useful for two things: building out an audience profile you can actually use, and generating a first draft that gets your ideas out of your head and onto the page. From there, your judgment drives everything.
A Word on Common Concerns
Skepticism about AI is reasonable. Here are the concerns we hear most often, and how to think about them.
Bias. AI tools are trained on large amounts of text from the internet, which reflects the biases present in that content. AI-generated profiles of specific communities or populations can reproduce stereotypes. Use them as a starting point, not a final word. Adjust anything that feels reductive or off-base, and draw on your own knowledge of the communities your research involves.
Hallucinations. AI can produce confident-sounding claims that are simply wrong, including made-up statistics, incorrect citations, and plausible-but-false specifics. Never use an AI-generated fact, number, or example without checking it against your actual findings or a reliable source. The data comes from you; AI is helping you shape it.
Quality. The first output is rarely the best output. AI does not know your audience the way you do. Treat everything it generates as a draft to improve, not a product to adopt and immediately disseminate. Iteration and revision is important. Treat initial outputs as a baseline that you build from.
Privacy. AI tools you don’t control may store or reuse what you type. Never paste unpublished findings, embargoed data, participant information, or anything you wouldn’t want to become public. Stick to information that’s already published or that you’re comfortable sharing broadly.
Step 1: Build an Audience Profile
Effective communication starts with a specific, realistic picture of who you’re talking to. A state legislator, a hospital administrator, and a journalist all need to hear something different about the same research.
Try a prompt like this:
I’m preparing communications for [Audience X] about [brief description of your topic or findings]. Can you help me create a short, detailed audience profile including:
Their role, influence, and who influences them
Cultural and professional context
What they care about and what they need to achieve their goals
What they have done recently related to [your topic area]
Their beliefs and values, and words or phrases that resonate with them
Please provide a concise, realistic profile that could help someone craft messages tailored to them.
The first response will probably be useful but generic. Push back:
What feels too broad? Ask for more specificity.
What’s missing? Ask it to add recent actions or concrete examples.
Does something feel off? Say why, and ask it to adjust.
Even without AI, these are useful prompts for sharpening any draft you write yourself.
Keep refining until the profile feels like a real person you could picture sitting across from you. Stop when it’s useful, not when it’s perfect.
Step 2: Draft a Tailored Message
Once you have a profile you trust, use it to draft. A simple prompt structure:
Using the audience profile below and the findings from [brief description of your study or project], help me draft a 60 to 90 second elevator pitch tailored to this audience. Lead with the conclusion and its importance, describe what the research shows with a concrete example, and end with a clear ask. Use language and values that resonate with this audience.
Audience profile: [paste profile here]
From there, keep refining:
“Make this less formal.”
“Use plainer language.”
“Make the example more concrete and specific.”
“Shorten it to 4 to 5 sentences.”
“Make the ask sound more like something [audience] could actually do.”
Using AI Responsibly
A few principles to keep in mind:
Bring the knowledge. The findings, the stakes, the context come from you. AI helps you translate and tailor; it does not replace your expertise.
Check everything. Review any claims, examples, or data points against your actual work before you use them.
Push for plain language. If the draft sounds like it was written for an academic journal, keep pushing. Ask for shorter sentences, simpler words, and a more human voice.
Correct what feels wrong. If the profile or draft misrepresents a community, contains a stereotype, or doesn’t reflect your values, say so and ask for a revision. You are always in charge of the output.
Protect what’s not public yet. Keep unpublished data, embargoed findings, and personal information out of AI tools you don’t control.
Want to Go Deeper?
AcademyHealth’s Communicating for Impact Flexible Learning Experience is a self-paced online course that walks you through the full communications process: understanding your audience, crafting messages that resonate, and choosing the right channels to reach them. You’ll come away with a defined audience profile, a tailored elevator pitch, and message versions you can adapt across platforms. The course is open through December 31, 2026. Learn more and register at academyhealth.org/events/communicating-impact-flexible-learning-experience.
Key Takeaways Graphic
At a moment when evidence is under pressure, clear communication matters more than ever.
Knowing your audience before you write is the step most researchers skip.
AI can help you build an audience profile and draft a tailored message faster.
Bias and hallucinations are real risks. Knowing how to check the output keeps you in control.
Your expertise drives everything. AI is a thinking partner, not a replacement.