Search for ChatGPT prompts for social media managers and you'll find long lists: 50 caption prompts, 30 hashtag prompts, "act as a viral content strategist". They're written for someone running one brand. Most of the advice comes down to "be specific" and "fact-check the output", which is right but doesn't tell you how.
Agency work is different. Your social media manager handles eight or fifteen clients, and captions are maybe a third of the job. The rest is turning a client's rambling voice note into a brief, replying to comments in three languages, explaining a bad month and turning "make it more premium" into something a designer can act on. Those are the prompts below. For calendar planning and captions we've already published full workflows: the AI content calendar prompt and the AI caption workflow. This post covers everything else.
What makes a ChatGPT prompt work for a social media manager?
Context, not cleverness. "Act as a world-class social media expert" changes very little. What changes the output is the model knowing this client: a jeweller in Jaipur who never discounts, or a Pune café whose regulars are college students. Every prompt in our library has five parts:
- The client card, pasted first (template below).
- The task, in one or two sentences.
- The raw material: the voice note transcript, the comments, the numbers. Real input only.
- The constraints: language, length, what not to say.
- The output format: a table, a numbered list, three options. Never "write something".
And one standing rule at the end of every prompt: do not invent facts; write [CONFIRM] wherever one is needed. That line turns the model's worst habit, making up confident details, into a checklist your account manager can work through in a minute.
The client card: paste this before every prompt
Write it once per client, keep it in their folder and update it when something changes. It takes 15 minutes per client and saves that every day.
CLIENT: [brand], [what they sell, price range], [city/cities].
BUYERS: [one sentence: who buys and why]. They hesitate because of [top objections].
VOICE: [formal / warm / playful]. Language: [English / Hinglish / Hindi / regional], [emoji rule]. Sounds like: [paste 3 posts that worked].
NEVER: [banned claims, competitor names, words the owner hates, sensitive topics].
RULE: Do not invent prices, offers, dates, awards, statistics or customer quotes. Write [CONFIRM] wherever a fact is needed.
10 ChatGPT prompts for agency social media work
1. Client voice note → creative brief
Here is a transcript of the client's voice note: [paste]. Turn it into a creative brief with: objective, key message, format, must-include details, deadline. List anything the client said that is unclear or contradictory as questions I should ask them. Do not fill gaps with assumptions.
The questions list is the most useful part. It's what stops your designer from guessing.
2. Five hooks for one Reel
The Reel shows [what's on camera]. Write 5 opening lines of under 8 words each, one per angle: a question, a bold claim we can prove, a mistake, a before/after, a local reference. Mark which need a fact confirmed.
3. One post, three registers
Rewrite this caption three ways: clean English, everyday Hinglish in Roman script, and simple Hindi in Devanagari. Keep the offer and CTA identical. Don't translate word for word; write how a [city] customer actually talks.
Have a native speaker read the Hindi and any regional version before it goes out. The model's Hindi is often grammatically fine and still sounds like a government notice.
4. Repurpose one Reel into the rest of the week
From this Reel script [paste], draft: a 5-slide carousel outline, 3 Story frames with one poll, and a 2-line Google Business Profile post. Same facts only.
5. Comment and DM replies, with an escalation flag
Draft replies to these comments [paste] in the brand voice, under 25 words each. Mark any comment that is a complaint, a refund or legal question, or a health or safety issue as ESCALATE and don't draft a reply for those.
6. Reply to a negative Google review
Draft a reply to this review [paste]. Acknowledge the specific problem, don't argue, don't admit legal fault, offer a direct contact ([CONFIRM] number). Under 60 words. Give two versions.
7. Vague feedback → designer task list
The client said: "[paste feedback]". Translate it into specific, numbered changes for the designer (colour, type, layout, copy). If the feedback could mean two different things, list both readings so I can ask the client.
This one alone cuts revision rounds. More on that in handling client revisions.
8. Monthly numbers → report summary
Here are this month's numbers and last month's [paste]. Write a 5-line summary for a busy business owner: what went up, what went down, the one post that did best and what we'll change. Describe only what the numbers show. Don't guess at causes; I'll add those.
Pair it with the structure in our monthly report guide.
9. Explaining a bad month
Reach fell [x]% this month because [the real reason, from me]. Draft a short WhatsApp message to the client that states it plainly, takes responsibility for our part and gives the fix with a date. No jargon, no excuses about "the algorithm".
10. A messy process → SOP checklist
Here's how we currently do [process], as described by the team [paste]. Turn it into a numbered checklist with an owner for each step and the point where the client must approve. Flag steps that seem to be missing.
Our SOP documentation guide covers what to do with the checklist once you have it.
What should social media managers never trust ChatGPT with?
The prompts above share one design choice: the facts come from you and the model shapes the words. Turn that around and things go wrong. Our hard lines:
| Task | Let AI do it? | Why |
|---|---|---|
| Prices, offers, stock, timings | No | It fills gaps with plausible numbers. One wrong price is a customer argument at the counter. |
| Festival and event dates | No | Lunar festivals move every year. Take dates from an official holiday list. |
| Customer testimonials and reviews | Never | ASCI's 2026 AI guidelines prohibit fabricated endorsements, and a label doesn't fix that. |
| Health, finance, legal claims | No | Sector rules apply whoever wrote the line; a human with the brief signs off. |
| AI images, voices, faces | Only with a label | Synthetic audio and visuals fall under India's 2026 IT Rules amendment. |
| Posting times | No | Use the client's own insights, not a general rule. |
| Approving the post | Never | The client approves. The agency is accountable. |
The rules here are clearer than most agencies realise. ASCI released its final Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising on 29 September 2026. Text uses such as writing ad copy need no AI label. Fabricated endorsements, deepfakes and unauthorised use of someone's likeness are prohibited even with a label. So an AI-written caption is fine. An AI-written "review from a happy customer" is not. We covered the IT Rules side, and why visuals rather than captions are the regulated part, in will AI replace social media managers?, and the influencer side in our ASCI influencer guide.
How do you keep client data safe when using ChatGPT?
Your clients didn't sign up to have their launch plans sitting in someone's personal chat history. Three habits cover most of the risk:
- Turn off training. In ChatGPT, go to Settings → Data Controls and switch off "Improve the model for everyone". OpenAI's help page says new conversations then won't be used to train its models.
- Use Temporary Chat for sensitive work. OpenAI says temporary chats aren't used for training and may be kept for up to 30 days for safety reasons. That's less exposure, not none.
- Paste the minimum. No passwords, no customers' names or phone numbers from DMs, nothing a contract marks confidential. Personal data in a client's comments is still personal data; see our DPDP Act guide for agencies.
Put this in writing for the team. A one-page AI policy (approved tools, what may be pasted, who reviews AI drafts) prevents most mistakes. Juniors aren't careless; they just don't know where the line is until someone shows them.
Frequently asked questions
What is the best ChatGPT prompt for social media managers?
There isn't one best prompt. What matters most is the client context you paste before the task: what the brand sells, who buys, the voice, banned claims and three posts that worked. With that card in place, short task prompts give usable drafts. Without it, even a long, clever prompt gives generic copy.
Is it safe to paste client data into ChatGPT?
Paste only what the task needs. Turn off "Improve the model for everyone" in Data Controls, or use a Temporary Chat. Never paste passwords, customers' phone numbers or names, unreleased launch details or anything the client contract calls confidential. If the client is strict about data, ask before using AI on their account.
Do AI-written captions need an "AI-generated" label in India?
Text alone does not. India's 2026 IT Rules amendment covers synthetic audio, images and video, and ASCI's 2026 guidelines say AI-written ad copy needs no label. AI-made visuals and voices are different and may need labelling. Whatever the format, a false claim stays false, so the facts in the caption still need a human check.
Should junior social media executives use ChatGPT?
Yes, but with the agency's prompt library and client cards, not their own improvised prompts. Juniors get the biggest time saving on first drafts, rewrites and summaries. They also miss invented facts most often. So their AI drafts should go through the same review as anything else before the client sees them.
Sources
- OpenAI — Data controls in ChatGPT: the "Improve the model for everyone" setting.
- OpenAI — Temporary Chat FAQ: not used for training; may be kept up to 30 days for safety.
- Fox Mandal — ASCI guidelines on labelling AI content in advertising: final guidelines released 29 September 2026; text uses need no label; fabricated endorsements prohibited.
- Mondaq — ASCI's draft AI disclosure guidelines: the risk tiers in the May 2026 draft.
AI drafts. Your team checks. The client approves.
Keep each client's brand brief, AI drafts, approvals, Instagram scheduling and monthly reports in one workspace. From ₹999/month, with a 7-day money-back guarantee. See the agency management software and the Instagram scheduler.