A prospective client opens ChatGPT and types, "Who offers reflexology near me, and what should I expect at a first session?" The answer may name a handful of practitioners, summarize what each offers and link to where to book. Whether your practice appears in that answer, and whether what it says about you is accurate, depends partly on things you control: your listings, your website and how clearly you describe your work.

The short answer: use ChatGPT for two jobs. First, as a writing partner for the listings, FAQs and replies that bring clients in. Second, as a mirror: ask it what it knows about your practice, then fix whatever it gets wrong at the source. The nine prompts below cover both. The same two rules as the rest of this series apply: no client personal or health details go into ChatGPT, and it never decides what your work can promise.

This guide is part of a series. If you have not chosen a tool yet, start with the overview of AI tools for wellness practitioners; for careful website writing and small builds, see the Claude prompts guide.

Before you start: one project, one setting, one rule

Create a project for your practice. In ChatGPT, a Project keeps your instructions, reference files and related chats together, so you are not re-explaining your practice in every conversation. Create one called something like "My practice – website and listings," and use it for every prompt in this guide. Prompt 1 writes its instructions for you.

Switch off model training. On personal plans (Free, Plus and Pro), ChatGPT can use your conversations to improve OpenAI's models unless you turn this off. Go to Settings → Data controls and switch off "Improve the model for everyone" (OpenAI). Business, Enterprise and Edu workspaces are not used for training by default.

Keep client data out. Most countries give health information extra legal protection, and nothing in this guide needs it. Session notes, intake answers, client names and anything that could identify a person stay out of ChatGPT and out of your project files. Turning off training does not change this rule: it limits what happens to your data, not whether you should share it in the first place.

Turn on search for prompts 2 and 9. Those prompts ask ChatGPT what it can find about your practice on the live web. If search is off, it will answer from training data alone, which may be out of date or wrong.

1. The project instructions: teach ChatGPT your practice once

Without instructions, ChatGPT fills gaps with generic wellness language – the kind that sounds like every other practitioner and drifts toward promises you would never make. Project instructions fix that for every chat in the project. If you already wrote a practice brief with the Claude guide, paste it in and ask ChatGPT to condense it. If not, this prompt builds one.

I'm setting up a ChatGPT Project for my wellness practice. Interview me, one question at a time, to write the project instructions. Cover:
- my modalities, training and professional memberships
- where I work (town or area, in person, online or both)
- who I usually work with and what they come to me for, in their words
- my website and booking link, and the platform my site runs on
- my tone of voice: three words that fit, three that don't
- what I never claim or promise, and when I refer people to a doctor
Then write project instructions under 300 words, addressed to you (ChatGPT), with these sections: Who I am, Who I help, How I write, Never do. Under "Never do," include: "Never diagnose, never promise outcomes, never suggest my work replaces medical care, never ask for or use client personal or health information." Leave [placeholders] for anything I haven't told you.

Why it's built this way. Writing the instructions to ChatGPT, in the second person, makes them work as standing rules rather than background reading. The "Never do" section is the guardrail every later prompt relies on, and the word limit keeps it within what fits comfortably in project instructions.

Use it well. Paste the result into your project's instructions and test it with a simple request, such as "write a two-line description of my practice." If anything sounds off, correct the instructions, not the individual answer.

2. The mirror check: what does ChatGPT say about your practice?

Before changing anything, find out where you stand. How ChatGPT chooses which local businesses to mention is not fully public, and independent analyses disagree on its sources – some point mainly to Bing and Bing Places (Local Falcon), others to data providers such as Foursquare and Yelp (Surface Local). You cannot see the system, but you can see its output. Run this with search turned on, ideally in a new chat outside your project so your instructions do not color the answer.

Search the web and answer as you would for a member of the public:
1. Who offers [modality] in [town or area]? List up to ten practitioners or practices and where you found each.
2. What can you tell me about [your practice name] in [town or area]? Include services, prices, location, how to book and anything people say about it. Cite your sources.
3. Is there anything about [your practice name] that looks out of date, inconsistent or unclear across the sources you found?

Then, in your project, paste the answers and ask:

Here is what ChatGPT search said about my practice. Compare it with my project instructions. Make a table of: what it got right, what it got wrong or out of date, what it left out, and the source it seems to have used for each. Then list the five fixes most likely to improve how I'm described, starting with the listing or page that needs changing.

Why it's built this way. Asking as a member of the public shows you roughly what a prospective client sees. Asking for sources tells you where to make each fix: if the wrong opening hours come from an old directory listing, editing your website will not help. Separating the audit from your project keeps your own instructions from shaping the answer.

Use it well. Save the first result as a baseline. Answers vary between runs and change over time, so treat this as a snapshot rather than a ranking, and do not chase small differences. If you do not appear at all, start with your listings before your website – prompts 3 to 5 deal with them.

3. The Bing Places prompt: the listing most practitioners skip

Many practitioners set up a Google Business Profile and stop there. Bing is easy to overlook, but it matters more than it used to, because several analyses link Bing's results to what ChatGPT says about local businesses. Bing Places is free, and if you already have a Google Business Profile you can import it rather than starting from scratch – listings imported from a verified Google profile can be verified at the same time (Semrush). The import does not copy everything, though, and changes in Google do not always carry over, so the listing still needs checking and finishing by hand.

I'm setting up my Bing Places for Business listing. Using my project instructions:
1. Write a business description for Bing in two lengths: a short version (under 250 characters) and a full version (under 750 characters). Say what I offer, where and for whom. No outcome claims.
2. Suggest the most accurate business category, and tell me if the honest choice is a general one (such as "wellness center" or "massage therapist") because no exact category exists for my modality. Never suggest a medical category I'm not qualified for.
3. List the details I should check by hand after importing from Google: name, address or service area, phone, website, booking link, hours, photos and category.
4. Write three short answers to questions a first-time client might ask, which I can reuse across my listings and website.

Use [placeholders] for anything not in my instructions.

Why it's built this way. Two description lengths cover the places a listing gets shown, from search results to the full profile, and save you writing it twice. The category question is the one practitioners most often get wrong: picking a more "medical-sounding" category to rank better can mislead people and may breach the platform's guidelines. The checklist exists because imported listings tend to look complete while quietly missing fields.

Use it well. Search for your practice on Bing and Bing Maps before you start – if a listing already exists, claim it rather than creating a duplicate. Make sure the name, address and phone match your Google profile exactly, character for character.

4. The Apple Maps prompt: your listing on every iPhone

When someone asks Siri for a practitioner nearby or searches Apple Maps, the result comes from Apple's own place data, not Google's. You manage that listing – Apple calls it a place card – through Apple Business, which replaced Apple Business Connect in April 2026 and is free to use for claiming and managing a location (Apple). A place card can show your photos, hours, contact details and action buttons such as "Book," and the same information appears across Apple Maps, Safari and Spotlight. Some features vary by country.

I'm claiming my Apple Maps place card through Apple Business. Using my project instructions:
1. Write a short description of my practice (under 500 characters) for the place card. Plain language, what I offer, where and for whom, no outcome claims.
2. Suggest the most accurate category and up to two secondary categories, with the same rule as before: honest, not medical unless I'm qualified.
3. Write the label and destination for one action button (for example "Book" linking to my booking page) and explain in one line which page it should go to and why.
4. Draft two short Showcase posts (under 200 characters each) I could use for a seasonal update or a new service, with no discounts or claims I haven't given you.
5. Write alt-style descriptions for the photos I should add: the treatment room, the entrance, and me at work.

Why it's built this way. The action button matters most, because it turns a Maps search into a booking with one tap – so it should point straight at your booking page, not your homepage. Photo descriptions help you choose photos that answer a nervous first-timer's real question: what does the room look like, and how do I find the door? Showcases are optional, but an empty place card gives a searcher less reason to choose you.

Use it well. Claim the location first; Apple verifies that you represent the business before you can edit it. Then search your practice on an iPhone in Apple Maps and through Siri to check the pin, hours and button work as expected. Reviews shown on Apple place cards generally come from third-party sources rather than Apple, so keep your other listings in good shape too.

5. The consistency audit: make every listing say the same thing

Search engines, map apps and AI assistants draw on many sources at once. If your website says one phone number, Google another and an old directory a third, they have less reason to trust any of them – and a client who rings the wrong number is a booking lost. This prompt turns a tedious comparison into a ten-minute job.

Below are my business details as they currently appear on each platform. Compare them field by field: business name, address or service area, phone, website, booking link, hours, category and short description.

My website: [paste]
Google Business Profile: [paste]
Bing Places: [paste]
Apple Maps: [paste]
Gyfts profile: [paste]
Other directories: [paste any others, such as a professional body's register or Yelp]
Give me:
1. A table with one row per field and one column per platform, marking any mismatch.
2. A single "correct" version of each field, based on my project instructions. Where my instructions don't say, ask me rather than guessing.
3. A short to-do list, platform by platform, of exactly what to change.

Flag tiny differences too, such as "St." versus "Street," a missing unit number or an old phone number.

Why it's built this way. A table makes mismatches visible at a glance, which a paragraph of notes does not. Asking ChatGPT to check your instructions – and to ask you when they are silent – stops it from "correcting" a listing to a version it invented. The platform-by-platform to-do list is what you actually work through.

Use it well. Copy each listing's details as they appear publicly, not as you remember them. Keep the agreed "correct" version in your project files so the next audit has a reference. Your Gyfts profile belongs in this check like any other listing.

6. The FAQ prompt: answers AI tools can lift

When someone asks ChatGPT "does reflexology hurt?" or "what should I wear to a Reiki session?", the reply draws on whatever pages it can find and understand. A short FAQ on each service page – real questions, direct answers – gives AI tools and search engines something clear to quote, and gives a nervous first-timer the reassurance they came for.

I want to add an FAQ section to my [service or modality] page. Here are questions clients and prospective clients actually ask me (no names or personal details):
[list 6–12 questions]
1. Pick the six most useful for someone deciding whether to book. Rewrite each question the way a person would type or say it.
2. Answer each in 40–80 words. Put the direct answer in the first sentence, then add one or two sentences of context. Describe what happens in a session and what clients often report; never promise results.
3. Where an answer touches on a health condition, add one line encouraging the reader to keep working with their doctor.
4. Produce FAQPage structured data (JSON-LD) that matches the answers word for word, and tell me in one line where to add it on [WordPress/Wix/Squarespace].

Mark anything you're unsure of with [CHECK]. Don't add statistics or cite studies.

Why it's built this way. Answering in the first sentence is what makes an FAQ quotable, whether by a search engine snippet or an AI assistant. The word range keeps answers long enough to be useful and short enough to be read. Structured data must match the visible text exactly; search engines' guidelines treat hidden or mismatched FAQ markup as a problem, so the prompt insists on a word-for-word match.

Use it well. Start with the service that brings in most new clients. Run the finished FAQ through the claims check in the Claude guide – any of the three tools can run it – before publishing. For the kind of plain, answer-first writing that works well here, see Gyfts' guide to what to expect at your first holistic session.

7. The inquiry reply: turn a question into a booking

Being found is only half the job. When someone messages to ask "Is this right for me?" or "How much is it?", the reply decides whether they book. Practitioners often answer these late in the evening, from memory, in slightly different words each time. A small set of reply templates – written once, adapted in seconds – makes every answer clear, warm and consistent.

Using my project instructions, write reply templates for the inquiries I get most often. Use placeholders like {first_name} and {booking_link}. Never include or ask for health details beyond what's needed to book safely.
Write one template for each:
1. "How much is a session and how long is it?"
2. "Is [modality] right for me?" – explain who I usually work with, what a first session involves, and that I can't advise on a health condition by message; suggest a short call or that they check with their doctor if they have one.
3. "Do you have availability this week?"
4. "I've never tried this before and I'm nervous."
5. A polite reply when the request is outside what I offer, pointing them to a more suitable kind of professional.

Keep each under 120 words, end with one clear next step, and give an email version and a shorter direct-message version.

Why it's built this way. Each template ends with one next step, because a reply that ends with "let me know" tends to end the conversation. Template 2 is the careful one: it explains your work without assessing anyone's condition by message, which protects both the client and you. Template 5 matters too – turning someone away well, with a useful pointer, is part of a good reputation.

Use it well. Save the templates in your email tool, your phone's text shortcuts and your messaging inbox. When an inquiry includes health details, reply to the booking question only and suggest a call; don't paste the message into ChatGPT.

8. The review prompt: ask well, reply safely

Reviews appear on Google, on Apple place cards (often drawn from third-party sites) and on directories, and several analyses suggest AI assistants weigh them when deciding which businesses to mention. Two things are in your control: how you ask for reviews, and how you reply. Both carry a risk specific to health and wellness work – a public reply can reveal that someone was your client.

Using my project instructions, write:
1. A short review request I can send after a session (email and text versions, under 60 words each). Ask for honest feedback in their own words, include {review_link}, and make it easy to ignore. No incentives of any kind.
2. A reply template for a positive review: thank them warmly without confirming anything about their visit, treatment or health.
3. A reply template for a critical review: thank them, stay calm, don't argue or correct details in public, and invite them to contact me directly at {contact}. Never confirm they were a client or mention anything about a session or their health – even if their review does.
4. A one-line note to myself on when to reply publicly and when to respond only privately.

Why it's built this way. The rule against confirming client status is the core of this prompt. Confirming that someone received a health-related service can amount to disclosing their health information, even when the reviewer shared it first, so replies stay general. Asking without incentives keeps you within most platforms' review rules, and "easy to ignore" keeps the request respectful.

Use it well. Never write, edit or post reviews on a client's behalf, and never ask ChatGPT to. The same principles apply to reviews on your Gyfts profile. If a review alleges harm or misconduct, reply privately and seek advice from your professional body or insurer before saying anything in public.

9. The monthly check-in: see whether it's working

Listings drift, directories copy old details and AI answers change over time. A short monthly routine catches problems early and shows whether your work is paying off. Run the first part with search on, in a new chat, as in prompt 2.

Part 1 (new chat, search on):
Search the web and tell me, as you would a member of the public: who offers [modality] in [town or area], and what do you know about [your practice name]? Cite your sources.
Part 2 (in my project):
Here is this month's answer and last month's baseline. Compare them:
1. Did my practice appear this time, and how was it described?
2. What changed, and which source does each change seem to come from?
3. Is anything now wrong or missing?
4. Here are my numbers from my booking tool and listings dashboards: [paste inquiries, bookings, Google and Apple listing views or actions if available]. Summarize the trend in two sentences, without guessing at causes the numbers don't show.
5. Give me no more than three things to do this month, in priority order.

Why it's built this way. Comparing against a saved baseline turns a vague impression into a record of what changed. Linking each change to a source tells you where to act. Pairing what AI says with your own numbers keeps attention on what matters – inquiries and bookings – rather than on whether a chatbot mentioned you on a given day. Capping the list at three keeps the routine to a short session rather than another project.

Use it well. Put a recurring reminder in your calendar and keep each month's results in your project files. Expect gradual change: listings can take weeks to update across platforms, and AI answers vary between runs, so judge the trend over several months, not one.

Quick reference

  • •1. Project instructions – once, then whenever your practice changes. Check: nothing invented.
  • •2. Mirror check – once, as your baseline. Search on, new chat. Check: note the source of every error.
  • •3. Bing Places – set up or refresh. Check: details match Google exactly.
  • •4. Apple Maps – claim and complete your place card. Check: the action button goes to your booking page.
  • •5. Consistency audit – after any change, and quarterly. Check: one agreed version of every detail.
  • •6. FAQs – one service page at a time. Check: structured data matches the visible text word for word.
  • •7. Inquiry replies – once, then refine. Check: no health advice by message.
  • •8. Reviews – once. Check: replies never confirm someone was a client.
  • •9. Monthly check-in – monthly. Check: judge the trend, not one result.

What ChatGPT should not do for your practice

ChatGPT can help with the business around your work, not the work itself. It should not assess a client, interpret symptoms, suggest what a client should do about a health concern, or write anything that implies your work replaces medical care. It can state things confidently that turn out to be wrong, including about your own practice, so check every fact it gives you before it goes anywhere public.