Published on

25 ChatGPT Prompts for Virtual Assistants (2026)

Authors
  • avatar
    Name
    PromptShelf Editorial
    Twitter

A good virtual assistant lives or dies by turnaround time. The faster you draft the email, build the schedule, or pull the research together, the more clients you can carry without dropping a ball. That is exactly where the right ChatGPT prompts for virtual assistants pay off: not by replacing your judgment, but by cutting the blank-page time on the repetitive work you do twenty times a week.

This is a working list. Every prompt below is copy-paste ready, built around a role, a task, real constraints, and a defined output. They are written for the freelance VA who juggles several clients, not a single corporate desk. Pick the ones that match your client mix and save them as templates.

How to use these ChatGPT prompts for virtual assistants

Swap the bracketed parts for your real details before you hit enter. The more specific you get with names, dates, tone, and word limits, the less editing you do on the other side.

Two rules keep you safe. First, never paste a client's private data into the free tier: no full email threads with personal addresses, no signed contracts, no card or bank details. Strip identifying information first. Second, read every output before it goes out under your name or your client's. ChatGPT invents dates, numbers, and policies with total confidence, so you are the fact-check step. Treat each result as a fast first draft, not a finished deliverable.

Email and inbox management

VAs spend more time in other people's inboxes than anywhere else. These get the routine replies out the door.

1. Draft a reply in the client's voice

Prompt: "You are an executive assistant writing on behalf of [client name], a [role/industry]. Draft a reply to the email below. Tone: [warm and brief / formal / friendly]. Keep it under 120 words, no exclamation marks, and end with a clear next step. Email to reply to: [paste the message]."

Set the tone tag once per client and reuse it. The word limit stops ChatGPT from padding a two-line answer into five.

2. Turn a rough note into a polished email

Prompt: "You are a virtual assistant. Turn the rough notes below into a clear, professional email. Keep my meaning exactly, do not add new commitments, and flag anything that reads as a promise I should confirm. Notes: [paste your bullet points]."

Useful when the client sends you three words and expects a full message. The flag instruction catches accidental commitments before they ship.

3. Write a polite decline or reschedule

Prompt: "You are an assistant declining a request on behalf of a busy client. Write a short, gracious email that says no to [request] without giving a detailed reason, keeps the door open for the future, and stays under 90 words. No apologies longer than one sentence."

Saying no well is a VA superpower. This keeps it kind and short instead of over-explaining.

4. Summarize a long thread before the client reads it

Prompt: "You are an assistant. Summarize the email thread below into: (1) what is being asked, (2) what has been decided, (3) what is still open, (4) the one action my client needs to take. Keep each section to one or two lines. Thread: [paste]."

Hand your client the four-line version so they decide in ten seconds instead of scrolling. Strip names and addresses first.

5. Build a reusable canned-response library

Prompt: "You are a virtual assistant. Write 6 short canned email responses for these recurring situations: [list situations, e.g. 'meeting confirmation', 'invoice received', 'out of office routing']. Each under 60 words, neutral-professional tone, with a [bracketed] slot for the detail I fill in."

Run this once per client and paste the results into your text-expander tool. The bracketed slots make each one reusable.

Calendar and scheduling

Scheduling is where small mistakes cost the most. These keep the day realistic.

6. Build a time-blocked daily plan

Prompt: "You are a productivity assistant. Build a time-blocked schedule for my day from the task list below. Work hours are 9:00am to 5:00pm with a 30-minute lunch. Put deep-focus work in the morning, batch the small admin tasks, and leave a 15-minute buffer between blocks. Output as a table with columns Time, Task, Notes. Do not exceed the available hours. Tasks: 'Draft client newsletter (90 min), reply to support inbox (45 min), book travel for CEO (30 min), update CRM records (60 min), team check-in call at 2:00pm (30 min, fixed), expense report (40 min)'."

This is the one I tested below. The fixed-time call plus a strict 8-hour window makes it easy to check whether the model actually respected the constraints.

7. Turn messy availability into options

Prompt: "You are a scheduling assistant. From the availability notes below, propose 3 meeting time options that work for all parties, in [time zone]. Flag any conflict you spot. Notes: [paste availability for each person]."

Cross-time-zone scheduling is error-prone by hand. Always double-check the zone math the model returns.

8. Draft meeting agendas from a goal

Prompt: "You are an assistant preparing a meeting agenda. The meeting is [length] and the goal is [goal]. Build an agenda with time allocations per item that fit the total length, a clear desired outcome for each item, and a named owner slot. Output as a numbered list."

Forcing time allocations that sum to the meeting length stops the 30-minute agenda that needs 80 minutes.

9. Write calendar event descriptions

Prompt: "You are an assistant. Write a clear calendar event description for [event]. Include the purpose in one line, a prep checklist of 2 to 3 items, the dial-in or location placeholder, and who to contact if something changes. Keep it scannable."

A good event description cuts the "what is this meeting again" replies. Keep it to a glance.

10. Plan a multi-stop travel day

Prompt: "You are a travel-planning assistant. Build an hour-by-hour itinerary for the trip below, with realistic buffers for security, transfers, and traffic. List what could go wrong at each leg and a backup. Verify nothing, just flag where I must confirm times myself. Trip details: [paste]."

The "verify nothing yourself" framing is deliberate: it reminds you that flight times and gate info are yours to confirm, not the model's to guess.

Social media and content support

Many VAs run their clients' content. These speed up the production line without flattening the voice.

11. Repurpose one post into a week of content

Prompt: "You are a social media assistant. Take the blog post or note below and turn it into 5 short LinkedIn posts, each under 100 words, each making one distinct point, in a [brand voice tag] voice. No hashtags, no emojis unless I add them. Source: [paste]."

One source, five angles, a week of queue filled. Keep the brand voice tag consistent across the client.

12. Batch caption variations for testing

Prompt: "You are a content assistant. Write 4 caption variations for this post, each a different angle: question-led, result-led, story-led, and direct. Keep each under 50 words and avoid the word [banned brand word]. Post context: [describe]."

Hand the client real A/B options instead of one caption to approve. Name any word their style guide bans.

13. Draft a content calendar from themes

Prompt: "You are a content planner. Build a 2-week posting calendar for [platform] from these themes: [list]. Output a table with columns Date, Theme, Hook, Format. Mix formats across the two weeks. Do not invent statistics or product claims."

The no-fabrication line matters most here, because invented stats slipped into a caption become the client's problem.

14. Write replies to common comments

Prompt: "You are a community manager. Write short, on-brand replies for these recurring comment types: [list, e.g. 'pricing question', 'compliment', 'complaint']. Each under 30 words, friendly but not over-eager. Add a [bracket] where a specific detail is needed."

Community management eats time. A reply bank keeps your tone steady when you are answering forty comments at once.

15. Turn a testimonial into shareable copy

Prompt: "You are a marketing assistant. Rewrite the client testimonial below into 3 short social-ready quotes, keeping the customer's meaning and not adding claims they did not make. Mark anything I should get written permission to use. Testimonial: [paste]."

Never let the model embellish a real customer quote. The permission flag keeps you on the right side of the rules.

Research, data, and admin

The grind work. These cut the time without cutting the accuracy you still owe.

16. Structure raw research into a brief

Prompt: "You are a research assistant. Organize the notes below into a one-page brief with these sections: Summary, Key Findings, Open Questions, Sources I Must Verify. Do not add facts that are not in my notes. Notes: [paste]."

ChatGPT is strong at organizing what you give it and weak at sourcing what you do not. The "must verify" section keeps that line honest.

17. Compare options in a decision table

Prompt: "You are an assistant building a comparison for my client. Compare these options across the criteria listed. Output a table, one row per option, and add a one-line recommendation noting what is missing from my data. Options: [list]. Criteria: [list]."

A clean comparison table earns trust fast. The "what is missing" note stops a confident recommendation built on thin data.

18. Clean and reformat a messy list

Prompt: "You are a data assistant. Reformat the list below into a clean table with these columns: [columns]. Standardize capitalization and date formats to [format]. Do not change any values, only the formatting. Flag any row that looks incomplete. List: [paste]."

Great for tidying contact lists or expense lines. The "do not change values" guard stops silent edits to real data.

19. Draft a standard operating procedure

Prompt: "You are a documentation assistant. Turn the steps I describe below into a clear SOP a new VA could follow with no prior context. Number every step, call out where a decision is needed, and list the tools required at the top. Steps: [describe your process]."

Documenting your own process is how you eventually hand work off. This turns a brain-dump into something trainable.

20. Pull questions for a client call

Prompt: "You are an assistant preparing me for a client onboarding call. Based on the client details below, write 8 smart questions that surface their real priorities, preferred communication style, and what success looks like in 90 days. Client: [describe]."

Walking into an onboarding call with sharp questions sets the tone for the whole engagement.

Running your VA business

You are a business of one, and the admin of that business is also your job. These handle the parts clients never see.

21. Write a service proposal

Prompt: "You are a freelance virtual assistant writing a proposal. Based on the scope below, draft a one-page proposal with: what is included, what is out of scope, a clear deliverables list, and a [pricing model] line I will fill in. Confident but not pushy. Scope: [describe the work]."

A tight proposal with explicit out-of-scope lines prevents the slow creep that kills VA margins. You fill in the number.

22. Draft a rate-increase message

Prompt: "You are advising a freelancer on a rate increase. Write a short, calm email telling a long-term client my rate will rise from [old] to [new] starting [date]. State the value I deliver in one line, give notice, and do not over-apologize. Under 120 words."

Raising rates is the conversation VAs avoid most. A measured script makes it a normal business note instead of a confrontation.

23. Build a client onboarding checklist

Prompt: "You are an operations assistant. Build an onboarding checklist for a new VA client covering access and tools, communication preferences, recurring tasks, and a first-week plan. Output as a checklist grouped by category. Leave [brackets] for client-specific items."

A repeatable onboarding makes you look organized from day one. Reuse the skeleton and fill the brackets per client.

24. Write a polite chase for an overdue invoice

Prompt: "You are a freelancer following up on an unpaid invoice. Write a firm but friendly reminder for invoice [number], [amount], now [X days] overdue. Keep it under 80 words, state the amount and original due date, and give one clear payment next step. No threats, no long apology."

Getting paid is part of the job. This keeps the chase professional so the relationship survives it.

25. Plan your own week across clients

Prompt: "You are a productivity coach for a freelancer with multiple clients. From the commitments below, build a weekly plan that protects [number] focus hours, spreads client work to avoid same-day pileups, and leaves room for admin and rest. Flag where I am overcommitted. Commitments: [list]."

Turning the lens on your own week is how VAs avoid burnout. Let the model tell you when you have said yes too many times.

What this prompt actually returns (tested 2026-06-06)

We ran Prompt 6 (the time-blocked daily plan) on free ChatGPT with the example task list baked in ('Draft client newsletter (90 min), reply to support inbox (45 min), book travel for CEO (30 min), update CRM records (60 min), team check-in call at 2:00pm (30 min, fixed), expense report (40 min)'). Reproduced verbatim:

TimeTaskNotes
9:00–10:30Draft client newsletterDeep-focus work; tackle the most cognitively demanding task first.
10:30–10:45BufferReview progress, stretch, prepare for next block.
10:45–11:45Update CRM recordsFocused admin task while energy is still high.
11:45–12:00BufferClear inbox notifications and transition.
12:00–12:45Reply to support inboxBatch email/support responses.
12:45–1:15Lunch30-minute lunch break.
1:15–1:30BufferPrepare for afternoon tasks and call.
1:30–2:00Book travel for CEOComplete before the scheduled meeting.
2:00–2:30Team check-in callFixed meeting time.
2:30–2:45BufferCapture action items from the call.
2:45–3:25Expense reportComplete and submit report.
3:25–5:00Open work blockExtra time for follow-ups, overflow, priorities that emerge, or additional project work.

This schedule uses 4 hours 55 minutes of planned task work, includes all required tasks, preserves a 30-minute lunch, adds 15-minute buffers between work blocks, and stays within the 9:00am–5:00pm workday.

Editor's notes. The arithmetic is the win here: every task kept its exact requested duration, the fixed 2:00pm call landed at 2:00pm, and the "4 hours 55 minutes" total the model reports is correct. It quietly dropped the 15-minute buffer before that fixed call, running the CEO travel booking from 1:30 to 2:00 straight into the meeting, which is the right trade-off but means it bent your own buffer rule to make the fixed time fit. It also did not actually batch the admin work as the prompt asked: the CRM update, support inbox, and expense report are scattered across morning, midday, and afternoon rather than grouped, so regroup them if batching matters to you. The 3:25 to 5:00 "Open work block" is padding the model added to fill the day; it was not in your task list, so cut it if you wanted a tight schedule. As always, the plan is only as good as your real energy and priorities, so treat it as a starting grid, not a fixed law.

Tips for getting better results

Lead with the role. "You are a scheduling assistant" steers the model harder than any polite framing. State the output format you want, a table, a numbered list, a word count, because an unspecified format defaults to something you will reformat anyway. Give one example of the tone or voice you want rather than three adjectives. And iterate: the first answer is a draft, so reply with "tighten the second paragraph" or "cut it to 80 words" instead of starting over.

The biggest time saver is building a personal prompt library. Every time a prompt works, save it with the bracketed slots intact. Within a month you will have a tested set you trust, and the blank-page time on routine work drops to near zero.

FAQ

What are the best ChatGPT prompts for virtual assistants?

The most useful ones target the work you repeat: drafting client emails in a set voice, summarizing long threads, building realistic schedules, and chasing invoices. A prompt earns its place when it includes a role, a task, real constraints, and a defined output format. Save the ones that work as reusable templates.

Is it safe to use ChatGPT for client work?

Yes, with limits. Never paste private client data into the free tier: no personal contact details, signed contracts, or financial information. Strip identifying details first, and always review every output before it goes out under your or your client's name. Treat ChatGPT as a first-draft tool, not a finished-work tool.

Can ChatGPT manage a calendar or send emails for me?

Not on its own. The free chat tool drafts and plans, but it does not connect to your client's inbox or calendar unless you set up a separate integration. For VA work, the practical pattern is to draft in ChatGPT, then copy the result into the real tool yourself after a review.

How do I keep the output in my client's voice?

Give the model a short voice tag you reuse every time, like "warm, brief, no exclamation marks," and paste one real example of past writing it should match. Consistency comes from reusing the same instruction, not from rewriting it each session.

Will using ChatGPT make my VA work feel generic?

It can, if you ship the first draft. The fix is to use it for structure and speed, then add the specific details, names, and judgment only you have from working with the client. The model handles the blank page; you handle the part that makes it yours.

Keep your workflow sharp

The pattern across all 25 prompts is the same: define the role, give real constraints, demand a specific output, then edit. Start with three or four that match your daily work, save them with the brackets intact, and build from there. The goal is not to hand your job to a chatbot. It is to clear the routine drafting so you have more hours for the judgment your clients actually pay for.

Pick one prompt from the scheduling or business section and run it on a real task today. That is how the library starts.