50 ChatGPT prompts for hospital marketers (with review rules)

The most useful ChatGPT prompts for healthcare marketing ask AI to structure, rewrite and analyse, not to supply clinical facts. Use anonymised inputs only, never patient data. Give the model a role, audience, constraints and approved facts, and ask it to flag uncertain statements. Every output then needs doctor review for clinical accuracy and a check against NMC, ASCI and ad platform rules.

Most hospital marketing teams in India now use ChatGPT or a similar assistant every day, usually without any shared rules. One person pastes in an enquiry transcript with a patient’s name in it; another publishes a symptom article with an invented statistic; a third writes ad copy promising “painless” treatment. None of them meant to break anything. They just did not have a list of good prompts or a checklist for what happens after the prompt.

New to the topic? Start here: AI in healthcare operations sets out the reading order.

This article gives you both. First, the rules your team should follow before using any public AI tool. Then 50 copy-ready prompts grouped by the work marketing teams actually do, each with placeholders in square brackets. Finally, how to write better prompts yourself and a review checklist for anything AI helps produce. If you are building this into a shared library, my guide on setting up a prompt library for a healthcare marketing team covers ownership, versioning and access.

What rules should come before any prompt?

A prompt list without rules is how hospitals end up with privacy incidents and regulatory complaints. These are the five I would put on the wall.

1. No patient data goes into public AI tools

No names, phone numbers, UHIDs, reports, photographs, call recordings or chat transcripts that could identify a patient. That includes “just the first name and the condition”. On consumer ChatGPT, OpenAI states that content may be used to train its models unless you switch off the relevant setting, while business, enterprise and API products are not used for training by default. See OpenAI’s explanation of how data is used. Even with training switched off, the safest rule is simple: if it could identify a patient, it does not go in. Anonymise and aggregate first.

2. Treat DPDP as already in force

The Digital Personal Data Protection Act, 2023 and the DPDP Rules notified in November 2025 govern how personal data is processed, including by tools you use. Patients did not consent to their data being processed by an AI tool for marketing. The substantive obligations phase in by May 2027, but processes built now should already meet that standard. My guide to DPDP consent for hospital marketing explains the practical steps.

3. Every clinical statement is reviewed by a doctor

AI writes fluent, confident text, and it can be wrong about dosage, eligibility, risk and recovery. Anything that states a clinical fact must be reviewed and approved by a qualified doctor in that specialty before publishing, with the reviewer’s name recorded. The process is set out in how to review AI-generated doctor content.

4. Advertising and professional codes still apply

The Indian Medical Council (Professional Conduct, Etiquette and Ethics) Regulations, 2002 and the ASCI Code apply to the output, whoever or whatever wrote the first draft. No superlatives, no guarantees, no fear-based copy, no unsubstantiated claims. See NMC and ASCI rules for hospital advertising.

5. Label synthetic media where ASCI requires it

ASCI’s guidelines on labelling synthetically generated content, issued in September 2026, require disclosure where AI-generated or AI-altered visuals or audio could materially influence a consumer decision, such as AI avatars of doctors or cloned voices. Text drafted with AI help and routine edits do not need a label, but fabricated testimonials or results are not acceptable even with one. Read the ASCI guidelines and my summary of ASCI AI labelling rules.

Which marketing tasks suit AI, and which do not?

50 ChatGPT prompts for hospital marketers

Copy a prompt, replace everything in [square brackets], and add the context the model needs. Placeholders in double curly braces are for your messaging platform’s template variables and should be left as they are. All prompts assume anonymised inputs.

SEO and content briefs (prompts 1 to 10)

These work best with a clear page purpose and a doctor who will supply facts. For the bigger picture of planning, see healthcare content marketing.

This connects directly to What ChatGPT ads mean for hospital marketing in India, GA4 setup for hospital websites: events, conversions and consent and AI in hospital operations in India: a non-clinical risk framework.

1. Keyword cluster for a service line. Act as an SEO strategist for a [multi-specialty hospital / single-specialty clinic] in [city]. List 30 search queries patients in India type about [specialty or procedure], grouped by intent: symptom research, treatment options, cost, doctor or hospital choice, and post-treatment. For each, suggest the page type that should answer it. Do not invent search volumes.
2. Content brief for a procedure page. Write a content brief for a page on [procedure] at [hospital name], [city]. Include: target query, search intent, suggested H2 and H3 questions, the facts the treating doctor must supply (eligibility, preparation, duration, recovery, risks, fee range), internal pages to link, and a list of claims that must not be made. Leave every clinical fact as a placeholder for the doctor to fill.
3. Outline from patient questions. Here are questions our front desk hears about [condition]: [paste anonymised list]. Turn them into an article outline with question-led H2s, ordered from most common to least. Flag any question that needs a doctor’s answer rather than general information.
4. FAQ set for a doctor profile. Draft 8 FAQ questions patients ask before booking with a [specialty] consultant: timings, languages, fees, what to bring, follow-up, teleconsultation, insurance. Write questions only; we will fill answers from the hospital’s verified information.
5. Title and meta options. Suggest 10 page titles (under 60 characters) and 10 meta descriptions (140 to 155 characters) for a page about [topic] in [city]. Avoid superlatives such as best, top, number one, guaranteed and painless. Use British spelling.
6. Rewrite for readability. Rewrite this paragraph for a patient with no medical background, at roughly a Class 8 reading level, keeping every clinical fact exactly the same and adding nothing new: [paste approved text]. List any terms you simplified so the doctor can check them.
7. Local page differentiation. We have location pages for [area 1], [area 2] and [area 3]. Here is the factual information for each: [paste facilities, timings, doctors, directions]. Suggest how to make each page genuinely different and useful, without duplicating text or creating thin content.
8. Internal linking map. Here is a list of our page URLs and titles: [paste]. Suggest internal links from each page to 3 to 5 relevant pages, with natural anchor text. Point out orphan pages and pages that compete for the same query.
9. Content gap check. Here are the H2 headings from our page on [topic]: [paste]. Here are headings from three pages ranking above us: [paste]. List the questions they answer that we do not, and the ones we answer better. Do not suggest copying their wording.
10. Schema field list. List the structured data properties appropriate for a [Hospital / MedicalClinic / Physician] page, and which facts from our page should populate each. Output as a table. Do not generate values we have not supplied.

Social media and video scripts (prompts 11 to 18)

Short video works when a real doctor speaks plainly. Use AI to structure, not to replace the doctor’s voice. The free list of 50 reel hooks for doctors pairs well with prompt 11.

11. Reel hooks. Write 15 opening lines (under 10 words each) for short videos where Dr [name], a [specialty] consultant, answers common patient questions about [topic]. No fear, no shame, no promises of results.
12. 60-second explainer script. Write a 60-second script for Dr [name] explaining [procedure or condition] for patients: what it is, who it is for, what to expect, and when to see a doctor. Mark every clinical statement with [VERIFY] so the doctor can confirm it before recording.
13. Myth versus fact post. Draft a carousel of 6 slides on common misconceptions about [topic]. For each myth, leave the fact as a placeholder [doctor to provide], and suggest a neutral headline. Avoid ridiculing anyone who believes the myth.
14. Awareness day calendar. List health awareness days in India between [month] and [month] relevant to [specialties]. For each, suggest one post idea and one doctor video idea. Mark dates I should verify against the official source.
15. YouTube description and chapters. Here is the transcript of our video: [paste]. Write a YouTube description (150 words), 5 chapter timestamps, and 10 tags. Do not add any claim that is not in the transcript.
16. Repurpose a long video. Here is a 12-minute doctor talk transcript: [paste]. Suggest 5 clips of 30 to 60 seconds that stand alone, with the start and end lines for each and a caption for Instagram.
17. Caption tone check. Review these 5 captions for a hospital Instagram account: [paste]. Flag anything that could read as fear-based, body-shaming, a guarantee, a superlative, or a call to self-diagnose. Suggest calmer rewrites.
18. Regional language adaptation. Adapt this approved English caption into natural [Hindi / Telugu / Tamil / Marathi] for a general audience, not a word-for-word translation: [paste]. Keep medical terms in English where patients commonly use them. Note anything a native speaker should check.

Ads copy and keywords (prompts 19 to 26)

Keep the focus on searches that lead to appointments; my piece on keywords that book appointments explains the logic, and compliant ad copy for doctors covers tone.

19. Search ad variations. Write 15 Google responsive search ad headlines (up to 30 characters) and 4 descriptions (up to 90 characters) for [specialty] consultations at [hospital], [area], [city]. Include location, timings, and booking options. No superlatives, no outcome claims, no prices unless I provide them: [price if any].
20. Keyword intent sort. Here is a search terms report export: [paste anonymised CSV]. Group the terms into: likely to book, research only, job seekers, wrong service, wrong location, and competitor brand. Suggest negative keywords for the irrelevant groups.
21. Negative keyword list. Suggest negative keywords for a Google Ads campaign promoting [service] at a hospital in [city], covering jobs, courses, free treatment, home remedies, government schemes we do not accept, and medicines. Explain any that are risky to exclude.
22. Meta ad primary text. Write 5 versions of primary text (under 125 characters) and headlines for a Meta ad promoting a [health check package] at [hospital]. Neutral tone, factual inclusions only, no fear of illness, no before-and-after framing. Package facts: [paste].
23. Landing page critique. Here is the text of our landing page for [service]: [paste]. Act as a patient deciding whether to book. List what is unclear, what is missing (fees, timings, doctor names, location), and where the page asks for too much information.
24. Lead form questions. Suggest the minimum lead form fields for a [consultation booking] ad that lets our team call back, and explain why each is needed. Do not ask for symptoms, diagnosis or medical history in the form.
25. Ad policy pre-check. Review these ad texts against Google and Meta healthcare advertising policies as you understand them: [paste]. Flag likely problems and say which policy area each relates to. I will confirm against the current policy pages.
26. Sitelinks and callouts. Suggest 8 sitelinks with descriptions and 10 callouts for a [hospital] Google Ads account, using only these facts: [paste facilities, timings, services, insurance tie-ups].

PR and communications (prompts 27 to 32)

AI is useful for first drafts and preparation. Journalists still need real news, real quotes and verified facts. See how to write a hospital press release journalists use.

27. Press release draft. Draft a press release about [event: new unit, new technology, doctor joining, camp]. Facts: [paste]. Use a news angle relevant to [city] readers, a quote placeholder for [spokesperson], and an editor’s note. No superlatives or firsts unless I supply proof: [proof].
28. Media pitch email. Write a 120-word pitch email to a health reporter at [publication] offering Dr [name] as a source on [topic, e.g. monsoon infections]. Explain why it is timely and what the doctor can speak about.
29. Spokesperson Q&A prep. List 15 questions a journalist might ask Dr [name] about [topic or announcement], including difficult ones about cost, access and safety. Do not write answers; we will prepare them with the doctor.
30. Holding statement. Draft a short holding statement for [situation type, e.g. a social media complaint going viral] that acknowledges concern, confirms we are looking into it, protects patient confidentiality, and gives a contact. No admission of facts we have not verified.
31. Internal announcement. Write an internal email announcing [news] to all staff, with what changes for them, who to contact, and what not to share externally yet.
32. Op-ed outline. Outline an 800-word opinion article by Dr [name] on [public health topic in India]. Suggest a thesis, three arguments, and where data is needed. Mark every statistic as [SOURCE NEEDED].

Patient communication templates for WhatsApp and email (prompts 33 to 38)

Write templates, never individual messages containing patient details. Templates go through approval once, then your system fills the variables. Only message patients who have opted in.

33. Appointment confirmation. Write a WhatsApp appointment confirmation template under 400 characters with placeholders for {{patient first name}}, {{doctor}}, {{date}}, {{time}}, {{location}} and a map link. Include how to reschedule. No medical advice.
34. Pre-visit instructions. Draft a friendly email listing what to bring to a first [specialty] consultation: ID, previous reports, medicine list, insurance card. Leave any preparation instructions (fasting, etc.) as [doctor to confirm].
35. Missed appointment follow-up. Write a polite WhatsApp message for patients who missed an appointment, offering to rebook without making them feel judged. Under 300 characters.
36. Post-visit feedback request. Write a message asking for feedback after an OPD visit, with a link placeholder [feedback link]. It must not offer incentives for reviews or ask only happy patients.
37. Health check reminder. Write a reminder for patients who opted in to annual health check reminders, mentioning that their last check was around [month] and how to book. Include an opt-out line.
38. Regional language versions. Translate these 3 approved message templates into [language], keeping placeholders exactly as they are: [paste]. Keep the tone respectful and simple.

Analysis and reporting (prompts 39 to 44)

Paste aggregated tables, not raw lead lists. If a column could identify a person, remove it before the prompt.

39. Monthly report narrative. Here is an anonymised, aggregated summary of this month’s marketing data: [paste table of sessions, enquiries, appointments, cost by channel]. Write a 200-word summary for the management team: what changed, likely reasons to check, and three questions to investigate. Do not invent causes.
40. Funnel drop-off. Here are our monthly numbers by stage: [enquiries, contacted, appointments booked, appointments honoured, treated] for [months]. Calculate stage conversion rates and point out the biggest drop. Show your working.
41. Campaign comparison. Compare these campaigns by cost per enquiry and cost per honoured appointment: [paste aggregated table]. Flag where the cheapest enquiries are not the cheapest appointments.
42. Review theme analysis. Here are 100 public Google reviews of our hospital with names removed: [paste]. Group them into themes (waiting time, billing, doctor communication, cleanliness, parking, etc.) with counts and representative quotes.
43. Search Console query analysis. Here is a Search Console export of queries for [page]: [paste]. Identify queries with high impressions and low click-through, and suggest title or content changes for each.
44. Spreadsheet formula help. I have a Google Sheet with columns [list]. Write formulas to calculate [metric] by month and by channel, and explain each one.

Research and strategy (prompts 45 to 50)

AI is a thinking partner here, not a source. Any figure it offers must be traced to a real source before it reaches a deck.

45. Catchment questions. We are planning marketing for a [specialty] unit in [area, city]. List the questions we should answer about the catchment (population, competing facilities, travel time, insurance mix, languages) and where in India such data is usually found. Do not guess the answers.
46. Competitor content audit framework. Create a framework to compare our [specialty] content with 5 competitors on: topics covered, doctor visibility, video use, fee transparency, booking ease and update frequency. Output as a table template.
47. Patient persona draft. Draft 3 patient personas for [service] in a Tier 2 city in India based on these interview notes: [paste anonymised notes]. Keep to what the notes support and mark assumptions.
48. Launch plan skeleton. Outline a 12-week digital launch plan for a new [unit or clinic] in [city], covering website, Google Business Profile, search ads, doctor content, PR and referral outreach. Give tasks by week, not budgets.
49. Stakeholder brief. Summarise this long strategy document for a hospital board in one page: [paste]. Keep the numbers exactly as given and list decisions required.
50. Pre-mortem. We plan to [initiative]. Imagine it has failed after six months. List the 10 most likely reasons, grouped by people, process, data and compliance, and one early warning sign for each.

How do you make these prompts better?

The difference between a generic answer and a usable draft is almost always context. Six habits make the biggest difference.

• Give a role and an audience. “Act as an editor for a hospital website whose readers are first-time patients in a Tier 2 city” produces different text from a bare request.
• Supply the facts, ask for the structure. Paste the doctor-approved facts and ask the model to organise and simplify them. Never ask it to supply clinical facts.
• State constraints explicitly. Word limits, character limits, British spelling, banned words (best, guaranteed, painless, cure), reading level, tone.
• Ask for flags, not just answers. “Mark any statement you are unsure of with [VERIFY]” or “list assumptions” makes review faster and safer.
• Show an example. Paste one approved post or ad in your house style and ask for more in the same style.
• Iterate in the same thread. Ask for a shorter version, a warmer tone, or a version for a different audience rather than starting again.

Save the prompts that work, with notes on what changed, in a shared library. Over a few months the library becomes the team’s real asset. For how this fits into a wider AI plan, see generative AI in healthcare and the notes on keeping accuracy when producing medical content at scale.

What should the review checklist include?

Every AI-assisted output should pass this before it is published or sent.

1. Privacy: no patient-identifiable data was used in the prompt, and none appears in the output.
2. Clinical accuracy: every clinical statement has been checked and approved by a named doctor in the relevant specialty.
3. No invented facts: every statistic, date, award, ranking or study has a real source you have opened yourself.
4. No superlatives or guarantees: no best, top, number one, guaranteed, painless, permanent, risk-free.
5. No fear or shame: no copy that plays on anxiety about illness, ageing or body image.
6. Hospital facts correct: doctor names, qualifications, timings, fees, addresses and phone numbers checked against the master record.
7. Tone and language: plain, respectful, British spelling, regional language versions checked by a native speaker.
8. Platform policy: ad text checked against current Google and Meta healthcare policies.
9. Labelling: AI-generated or AI-altered visuals or audio labelled where ASCI guidelines require it, and fabricated content removed rather than labelled.
10. Originality: no text copied from competitor pages; a plagiarism check on long-form content.
11. Record kept: who prompted, who reviewed, and the date of approval, stored with the content.

Who reviews what?

Mistakes to avoid

• Pasting real enquiries to “save time”. Anonymise every time, or do not use the tool for that task.
• Trusting confident numbers. AI tools can produce plausible statistics that do not exist. If you cannot find the source, delete the number.
• Publishing on the same day. Build review time into the calendar. A 48-hour doctor review window is realistic for most teams.
• Letting AI pick the topic. Topics should come from patient questions, search demand and service priorities, not from what a model suggests.
• One prompt per person. Without a shared library, every marketer reinvents prompts and quality varies widely.
• Using AI images of patients or doctors. Synthetic “patients” and doctor avatars carry ASCI, trust and consent risks out of proportion to their value.
• Skipping regional review. Machine-generated regional language text can be grammatically fine and culturally off. A native speaker should read it.

How to measure whether AI is helping

Track a small set of measures for a quarter, comparing AI-assisted work with your earlier baseline:

• Turnaround time from brief to approved draft.
• Revision rounds needed before doctor approval. If AI drafts need more rounds than human drafts, your prompts or inputs need work.
• Error rate at review: clinical corrections, fact corrections and compliance flags per piece.
• Output performance: organic clicks, ad click-through rate, enquiry rate from pages, on the same basis as before.
• Privacy incidents: any instance of patient data entered into a public tool. The target is zero, and every incident should be logged and reviewed.

What good looks like: a shared, versioned prompt library; a written rule on patient data that everyone has read; a named doctor reviewer for each specialty; a content calendar such as the free healthcare content calendar template with review time built in; and drafts that reach approval in fewer rounds than before.

Where to start this week

Pick three prompts from this list that match work your team does every week, run them with anonymised inputs, and compare the drafts with what you would have written. Keep the ones that save time without adding review effort. Share the rules section with everyone who uses AI tools, including your agency. Prompts are easy to copy; the rules and the review habit are what keep a hospital’s name out of trouble.

Frequently asked questions