Either works for admin, content and handouts, so choose by the job and by the data. Use whichever draft you edit least after a two-week test with the same tasks in both, keep client identifiers out unless a business plan and agreement cover them, and never let either tool produce a meal plan without your own calculation and checks.
Search results for this question are mostly prompt lists, which is the wrong starting point. A prompt works or fails depending on what you give the tool, what you do with its answer and whether the data should have been there at all. This tutorial goes the other way round: it sorts your weekly jobs by risk, sets rules for what may be entered, and then compares ChatGPT and Claude on the jobs that are safe to automate. The tools change monthly, so the comparison is a method you can repeat rather than a verdict.
Sort your jobs by data before you pick a tool
The question that decides everything is what you type in. Draw three columns on a sheet of paper and put each regular task in one.
| Column | What goes in | Examples for a freelance dietitian | Rule |
|---|---|---|---|
| Green: no client data | General content, your own words, public information | Newsletter drafts, recipe ideas, social posts, handout wording, FAQ pages, cancellation policy wording | Any plan, including free |
| Amber: anonymised | Details with identifiers removed and no rare combination that points to a person | "Adult with type 2 diabetes who skips breakfast" for a handout structure; rewriting a generic care summary | Paid or business plan, training switched off, you check the anonymisation |
| Red: identifiable | Names, dates of birth, contact details, records, lab results with context | Session notes, referral letters, anything from a client file | Only on a business-grade plan with a signed data agreement, or not at all |
Most of what a freelance dietitian wants from AI sits in green. Start there. A tool that never sees client data cannot leak it, and you will still save hours a week on content and admin.
The data rules that differ by plan
Consumer plans and business plans treat your content differently, and the differences matter more than the model names.
- Consumer ChatGPT and Claude plans let you switch off use of your chats for training. In ChatGPT the switch is "Improve the model for everyone" under Data controls, and OpenAI says it is on by default for Free, Plus and Pro accounts; in Claude it is "Model Improvement" in Privacy Settings. Switch it off before you use either for anything beyond green work.
- Retention differs. Claude's privacy centre says a deleted chat leaves its back-end systems within 30 days, and that with Model Improvement on, chats may be kept in de-identified form for up to five years for training. ChatGPT's Temporary Chat is not used for training but may be kept for up to 30 days. Both products offer memory across chats (Claude lists it on every plan, including Free; ChatGPT Free has limited memory), so switch memory off where you do not want client context carried from one chat to the next.
- Business plans (ChatGPT Business, Claude Team) do not train on your content by default. ChatGPT Business starts at two users and Claude Team is sold for teams of 2 to 150, so a one-person practice pays for two (roughly $40 to $50 a month) or uses an individual plan.
- Agreements. A business associate or data-processing agreement is a separate thing from a switch in the settings. Ask each vendor for the current position in writing before you put identifiable data in.
The general method for checking terms is in what to check in an AI tool's privacy policy and terms, and the related question of what you may enter at all is covered in can dietitians use AI for admin and content, and what stays human.

The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowWhat each plan costs, and what to buy first
| Option | Monthly cost | Good for |
|---|---|---|
| ChatGPT Free or Claude Free | $0 | Green tasks with a usage cap; ChatGPT Free may show ads and allows 3 file uploads a day |
| ChatGPT Go | $8 | More uploads and memory than Free; OpenAI says this plan may also include ads |
| ChatGPT Plus or Claude Pro | $20 each ($17 for Claude billed annually) | Regular green and amber work, more usage |
| Higher individual tiers | $100 and up | Heavy use; rarely needed for a solo practice |
| ChatGPT Business or Claude Team | About $25 per seat monthly ($20 annual), two seats minimum | Client-adjacent work under business terms |
My suggestion for a new freelancer: begin on Free for both, run the test below, and pay for one $20 plan only after the free limits block you more than twice in a fortnight. A second tool is rarely worth the money for a solo practice. Moving to a business plan makes sense when you add staff or put anything identifiable into a tool.
Job by job: where the two tools differ in practice
I will not claim one is better at everything, because both are strong at drafting and both change with each release. What differs is workflow. The honest way to choose is to run the same task through both and look at the edit time.
Handouts at the right reading level
A common task: turn clinical wording into something a client can follow. Paste your own text (green data) and ask for a specific reading age and a layout. Then compare both outputs for accuracy and tone.
Your text: "Aim for consistent carbohydrate distribution across meals to support glycaemic stability."
Prompt: "Rewrite for an adult reader with no medical background, at a 10-year-old reading level, in two sentences. Do not add advice."
Illustrative output: "Try to eat a similar amount of starchy food at each meal. This helps keep your blood sugar steadier through the day."
What to check: it kept the meaning, added no new advice, and "starchy food" is a fair plain-language match for carbohydrate. If it had added "avoid bread", you would delete it.
Both tools are likely to manage this. The difference you may notice is in how much each over-adds when told not to. Count the invented lines across five handouts.
Long documents and reference material
If you paste a long guideline or a supplier PDF to summarise, file limits matter. Claude's upload help says PDFs can run to 1,000 pages, but charts and images are analysed only in PDFs of 100 pages or fewer; ChatGPT accepts documents up to 512MB, with text capped at two million tokens a file, and Free accounts get three uploads a day. Those are generous for a practice, so the limit that matters more is your own: check any summary against the source, because summaries can drop caveats such as "in adults only".
Newsletters and social posts
This is where style matters more than accuracy, and the right tool is the one whose draft sounds least like a template. Give each the same three sample paragraphs of your own writing and ask for a draft in that voice. Keep the one you edit least. A month of ideas is laid out in a month of nutrition newsletters drafted with AI.
Reusable instructions
Both products offer Projects, where you store standing instructions and reference files, so a tool remembers your house style, handout layout and disclaimers. Claude's pricing page allows up to five projects on Free, with project sharing reserved for team plans; ChatGPT has Projects on Free too, with up to five files per project. Custom GPTs are scheduled to retire on 11 December 2026, so use Projects rather than building one. Put in a project only material that is green.
Session notes and charting
This is a red-data job. A general chat assistant is a poor fit unless you are on a business plan with an agreement, and even then a purpose-built scribe may be a better choice. See AI charting for dietitians: drafting ADIME notes with a review step for a careful workflow.
Why AI should not write meal plans alone
A meal plan is a clinical output. It rests on a client's energy needs, conditions, medicines, allergies, culture, budget and food access, and numbers that must add up. A language model produces fluent text, not verified arithmetic, and it can state a nutrient figure confidently when it is wrong. The risks are specific:
- Wrong numbers. Portion weights and nutrient totals may be invented or inconsistent between meals.
- Missed contraindications. A plan can include a food that conflicts with a medicine or an allergy, because the model did not know or did not weigh it.
- False precision. "1,847 kcal" looks exact and is not.
- Overreach. A plan for a person with a medical condition may stray outside your scope.
A safe pattern: you set the targets, you choose the foods, you calculate in your nutrition software or a composition database, and the AI helps only with formatting, wording and recipe variations that you then recheck. A before-and-after shows the boundary.
Unsafe: "Write a 7-day meal plan for a 52-year-old woman with high blood pressure and type 2 diabetes, 1,600 kcal."
Safer: "Here are the day's meals I have already calculated [your list, no identifiers]. Format them as a one-page table with a shopping list. Do not change any quantities."
The second prompt keeps the clinical decisions with you and uses the tool for layout, which it does well.
A two-week test you can run alone
- Pick five green tasks: a handout rewrite, a newsletter paragraph, a reply to a new enquiry, a cancellation policy rewrite, a recipe variation.
- Run each through both tools with identical wording.
- Time your edit from first draft to the version you would send.
- Count errors: facts wrong, facts invented, tone off.
- Record the usage limit: did either stop you mid-task?
- Pick one, unless the results are a tie, in which case pick the cheaper.
Test log (illustrative)
Task ChatGPT edit time Claude edit time
Handout rewrite 4 min 3 min
Newsletter paragraph 6 min 5 min
Enquiry reply 2 min 2 min
Cancellation wording 5 min 4 min
Recipe variation 3 min 6 min
Invented facts (total) 3 1
Limit hit? Once No
In that made-up log, Claude edits faster on four tasks and invents fewer facts, but ChatGPT is quicker for recipes. A dietitian could reasonably pick Claude and use ChatGPT's free tier for recipe ideas. Your log will differ, and that is the point.
A working week, with the time it might save
Illustrative sums for a freelance dietitian seeing 14 clients a week. The figures are invented to show how to size the benefit, not a promise.
| Weekly job | Time by hand | With AI draft and your edit | Data column |
|---|---|---|---|
| Reply to 6 new enquiries | 45 min | 20 min | Green (no details) |
| One newsletter | 90 min | 45 min | Green |
| Two social posts | 40 min | 15 min | Green |
| Adapt one handout for a new client group | 60 min | 30 min | Green or amber |
| Session notes, 14 clients | 3 hours | Do not use a general chat tool | Red |
The green and amber jobs save about 2 hours 45 minutes a week in this illustration, which is roughly 130 hours over a 48-week year. Even at a modest hourly value, that dwarfs a $20 monthly plan ($240 a year). The red job is the largest block of time and the one a chat assistant should not touch, so do not let the arithmetic push you into it. If notes are your biggest time sink, look at a purpose-built tool, as in dietitian software compared: which practice platforms have useful AI.
Teaching a tool your voice
Drafts that sound like everyone else's are a content risk, because your clients chose you for how you explain things. The cure is evidence, not adjectives. Paste three short pieces you have written yourself and ask for a description of your voice, then reuse that description.
Prompt: "Here are three paragraphs I wrote [paste]. Describe my writing style in six bullet points: sentence length, tone, words I favour, words I avoid, how I open, how I close."
Illustrative output:
- Short sentences, rarely over 18 words
- Warm but practical; no moralising about food
- Favours "try", "notice", "most people"
- Avoids "clean eating", "detox", "cheat"
- Opens with a common situation
- Closes with one small next step
Then: "Draft a 150-word post on [topic] in that style. Do not use any of the words I avoid."
Save that description in a Project so you do not retype it. It works in either tool, and the words-I-avoid list is a good guard against diet-culture language slipping in. Check that the draft keeps to your list; models sometimes drift back.
When the free limit bites
Free plans cap usage, and the caps change. ChatGPT Free describes unlimited text chats with limits on uploads, images, voice, research and memory, and Claude Free carries its own limits. When you hit one in the middle of a task, you have four options in order of cost: wait until the limit resets, switch to the other tool's free plan for that task, shorten the job by splitting it, or move to a $20 plan. Keep a note each time it happens. Three blocks in a fortnight on work that matters is a good reason to pay.
Using two tools without making a mess
Some practitioners use both. That is fine if you keep rules simple: one tool is the default for drafting, the other is used for a second opinion on wording or for a task the default does badly. Do not paste the same client information into both, and keep the same privacy settings on each. A second tool also doubles the terms you have to read, so only add it if the test log shows a clear gain.
An intake line about AI that clients will understand
If client information touches AI at any point, or AI helps write materials, tell clients. A short paragraph in the intake paperwork is enough, and it is worth adapting to your own practice and any professional body guidance you follow.
"I sometimes use AI writing tools to help prepare general handouts and administrative messages. I do not enter your name, contact details or health records into these tools. I review everything before it reaches you. If you would prefer I avoid AI for anything connected with your care, tell me and I will."
Notice what the paragraph commits you to: no identifiers, your own review, and a clear opt-out. Only write what is true of your practice. If you later start using a scribe on a business plan, update the paragraph before you do.
When to move to a business plan
Move when one of these is true: you want to put identifiable information into a tool and have a signed agreement that allows it; you add a second practitioner or an assistant who needs shared Projects; or your usage is steady enough that two seats at about $40 to $50 a month is cheaper than the time you lose to limits. Until then, an individual plan with training off and a one-page policy is usually enough.
A five-minute check before any handout goes out
Whatever tool drafted it, run the same short review before a handout, post or email reaches a client. It catches most of the errors that matter, and it is quick enough to do every time.
- Numbers. Every quantity, portion, frequency and percentage matches your source. Circle any that you cannot trace.
- Absolutes. Replace "always", "never", "cures", "guarantees" and "boosts" unless you can defend them.
- Scope. The text stays within general guidance. If it reads like advice for one person's condition, cut it or move it to an individual consultation.
- Allergens and conditions. Any recipe or suggestion flags common allergens and does not imply suitability for a medical condition.
- References. Every study named exists and says what the text says; you opened it.
- Voice. It sounds like you, with none of the words you avoid.
Print that list and keep it by your screen for the first month. After that the habit will carry itself, and you will spot the typical AI errors, such as a nutrient confidently attributed to the wrong food, at a glance.
A realistic mistake and how it showed up
An illustrative case. A dietitian asks an assistant to summarise the evidence on a dietary pattern for a client leaflet. The answer cites a study that sounds plausible, with authors and a year, and she pastes it into the leaflet. A colleague later fails to find the paper. The assistant had produced a citation that looked right but did not exist. Nothing in the text warned her.
The rule that prevents this: never include a reference you have not opened. Ask the tool to describe what to look for, then find the source yourself in a database. If you cannot, do not cite it. This is a general rule for AI use in health writing, and it holds for both tools.
Admin jobs both handle well
For admin, differences narrow. Both write clear replies to enquiries, appointment reminders, cancellation terms, intake form wording and invoice notes. Use your own facts and let the tool handle phrasing. A short illustration:
Prompt: "Draft a friendly reply to a new enquiry. Facts: initial consultation is 60 minutes, costs [your fee], online or in person, I will send an intake form after booking. Do not promise outcomes."
Illustrative output: "Thank you for getting in touch, [first name]. The first appointment lasts 60 minutes and costs [fee]. It can be online or in person. Once you book, I will send a short intake form so we can make the most of the time. I look forward to working with you."
What to fix: check the fee, add your booking link, and remove any phrase that promises results.
The wider list of jobs and the tools that suit them is in AI tools for dietitians by job.
One page of rules for your practice
Write a short policy, even for a solo practice, and keep it where you can show it to a client or a professional body.
AI use in my practice (draft)
1. I use [tool] for drafting handouts, newsletters and admin wording.
2. I never enter names, dates of birth, contact details or records into a consumer AI account.
3. Training on my chats is switched off. Memory is [off/limited].
4. I check every fact and figure, and I open every reference before citing it.
5. I never use AI to produce a meal plan without calculating and checking it myself.
6. Clients are told in my intake paperwork how I use AI and can opt out.
7. I review this page each quarter, because vendors change terms.
That is enough to start, and it turns "pick by job and data" from a slogan into something you can follow on a busy Tuesday.
ChatGPT, Claude and dietitian work: common questions
Can I paste client details into ChatGPT or Claude?
Not into a personal free or paid account unless you are sure the settings and terms allow it for client data. Safer: remove names, dates of birth and anything identifying, or use a business plan with a signed data agreement. Check each vendor's current terms, because they differ by plan and change.
Is a free plan enough for a freelance dietitian?
For handouts, newsletters and admin wording with no client data, often yes. Free plans have usage limits that vary, so you may hit them in busy weeks. Test for a fortnight and pay only if the limit blocks real work.
Can AI write a meal plan for a client?
It can draft a structure, but it should not produce the plan alone. Portions, energy and nutrient targets, allergies and medical conditions need your assessment and a food composition database. Treat any AI plan as a draft that you recalculate and check line by line.
Should I tell clients I use AI?
Yes if client information touches an AI tool or if AI helps produce their materials. A short line in your intake paperwork covering what you use, what you never enter and how clients can opt out builds trust and covers your obligations.
Further reads
- Is an AI Scribe Worth It for a Small Private Clinic? — When a scribe beats a chat assistant for notes.
- Starting a Nutrition Consulting Business: Admin and Where AI Helps — Admin set-up for a new practice.
- AI Tools for Chiropractors: Report of Findings, Care Plans and Handouts — Handouts and care plans in a neighbouring profession.
- Who in Your Team Actually Needs a Paid AI Licence? — Who in a team really needs a paid plan.
- Nutritionist AI Prompt: Write Consent and Cancellation Wording — Wording for consent and cancellation terms.
- AI Nutrition Advice at a Gym: What Is Safe to Offer Members? — Safe limits for AI nutrition advice.
- AI Training for Dietitians: A Practice Owner's Team Plan — Rules for a team once you hire.
- AI Tools for Nutritionist Practices: Client Check-Ins and Handouts — Client check-ins and handouts with AI tools.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: ChatGPT pricing and Business pricing pages, OpenAI help articles on data controls, ads, file uploads and custom GPT retirement; Claude pricing page, Claude privacy centre article on data retention and Claude upload help article (all read 4 November 2026).