How to batch create social media posts with ChatGPT

Yes, you can batch create social media posts with ChatGPT, but the batching happens while you write, not while you publish. One instruction can come back as several drafts, and each of those drafts is then approved and scheduled on its own, so a week of posts takes one sitting and still gets one decision per post.

The pages ranking for this question today are almost all design-tool tutorials: ChatGPT writes a list of quotes, a template tool turns them into pictures, and you paste each caption into each platform by hand. That covers the images and stops exactly where the slow part starts — reviewing each post and getting it scheduled to the right account. SceneSmith is a publishing layer that closes that last step from the same conversation.

What batch creation covers, and what it does not

Three things people call "batching" are not the same job, and it helps to know which one you are buying:

• A design tool with bulk creation. It turns one template into many images. Useful for quote cards and product grids; the words and the scheduling are still manual.
• A native scheduler. LinkedIn's composer, for example, queues posts from 10 minutes to three months ahead and converts the time to UTC from your device settings (LinkedIn Help, read 2026-10-13). You still write each post inside the platform.
• An AI assistant plus a publishing connector. You draft and revise the whole set in one conversation, get a separate draft per post, approve each one, and the schedule lives in one calendar. This is the combination this guide is about.

SceneSmith's own walkthrough shows the shape of it: "Turn our launch notes into posts for X, LinkedIn, and Instagram. Schedule all three for tomorrow at 9," with the drafts coming back for review before the schedule is approved — one approval before every post, as the homepage puts it today.

Step 1 — Connect the server once

1. Sign in to SceneSmith, open the workspace for the brand, and connect each destination you plan to publish to. Confirm the exact handle for every account before you draft anything.
2. Copy the production MCP endpoint from the connection guide: https://app.scenesmith.design/api/mcp.
3. In ChatGPT on the web, open Plugins, select the plus button, then Add custom MCP server. Enter a name and the server URL, configure authentication, read the risk warning, and select Create as a plugin (OpenAI, Add custom MCP server, read 2026-10-13).
4. Install the plugin, then start one clean conversation with a check: Show my SceneSmith setup status. Confirm the workspace and destination handles it reports before you ask for a batch.

Two limits on that OpenAI page are worth planning around. Custom MCP servers are added and used on ChatGPT on the web, and workspace permissions and security restrictions, Lockdown included, apply to them. The supported protocols are SSE and streaming HTTP. Never paste a bearer token or a provider secret into a client configuration; SceneSmith's MCP page says the same thing.

Step 2 — Give one instruction with the whole set

The reason batch drafting fails is usually the prompt, not the tool. Give the facts once, then describe what each post is for. A workable instruction:

Use only the facts below. Draft three posts from them.
X: the practical details, one paragraph.
LinkedIn: why the opening matters, for people who run small businesses.
Instagram: a caption for the storefront photo, two sentences.
Show each draft with its caption, media and destination.
Do not schedule anything yet.

Facts: print studio opens Saturday at 10 a.m., 18 King Street,
first 20 visitors get a free print. No results or stock claims.

Three habits keep the set usable. Keep the facts identical across all posts so the third draft cannot drift from the first. Ask for the drafts first and the scheduling later, so the review step is a separate decision. And name the destinations explicitly, because the same sentence reads differently as an X post and as a LinkedIn post.

Step 3 — Approve one draft at a time

This is where a batch differs from bulk publishing. OpenAI's documentation is explicit that write actions require confirmation by default, and that a tool without a read-only annotation is treated as a write action. You can expand each tool call and inspect the JSON payload before you approve it. So the review is not a formality: it is the last place you see the exact text, the media, the destination and the time for that one post.

OpenAI also lets you remember an approve or deny choice for a single tool for the rest of that conversation, with its own caution attached: only do that for a tool whose later write actions you are willing to accept without seeing them. In a batch session that setting is the difference between "I approved three posts" and "I approved whatever comes next". SceneSmith keeps the decision per post: the draft card shows the caption, media and destination, and nothing is published until you approve that specific draft.

Step 4 — Schedule the set and check the record

Schedule with an explicit time and time zone in the same conversation — "schedule the three for tomorrow at 9:00 a.m. America/Iqaluit" — and confirm each entry on the calendar afterwards. LinkedIn's own limits show why the platform matters here: a member post can be set from 10 minutes to three months ahead, the time is converted to UTC from your device's time zone, and some post types cannot be scheduled at all (LinkedIn Help, read 2026-10-13).

Then verify. A scheduled post is waiting to be sent; a delivery result is proof it was accepted. Check the result for each post in the batch rather than assuming that three scheduled posts are three published posts. The drafting side of this workflow is covered in more depth in how to automate social media posts and scheduling a LinkedIn post from ChatGPT.

What you cannot batch away

The pictures. An assistant writes captions well; SceneSmith publishes the media you supply. An Instagram feed post needs media, and bulk text-on-image templates remain a design tool's job. Pair the two: build the images in the design tool, draft and schedule the words from the conversation.

The approval. Every publish or schedule is a write action, and write actions require confirmation by default. A batch gives you one conversation and one calendar, not a standing permission to post.

The platform conditions. Instagram publishing through the API works for professional accounts, and Meta's guide adds a Page Publishing Authorization step for some Pages. Meta's publishing page also states two different 24-hour limits for API-published posts — 100 in its rate-limit section and 50 in its carousel section, with carousels counting as a single post (Meta, Content Publishing, read 2026-10-13). Plan for the lower figure and check the app. Meta's own guidance on the same page is aimed at apps that "allow app users to schedule posts to be published in the future", which is the role SceneSmith plays: the API has no scheduling parameter, so the publishing layer holds the post and fires it at your slot.

If you publish to one account and write the words there, a native scheduler is enough. If you want more of the batch-drafting background, the blog has the Codex route for posting to Instagram from Codex and the free option for scheduling LinkedIn posts.

Questions readers ask

Can you batch create social media posts with ChatGPT?

Yes, in the drafting sense. One instruction can return a set of drafts — SceneSmith's own example turns one launch note into posts for X, LinkedIn and Instagram — and each draft is then reviewed and scheduled separately. The batching saves the switching between apps, not the judgement about what goes out.

Can I schedule a whole batch without approving each post?

No, not with a connector that follows OpenAI's defaults. Write actions require confirmation, and any tool without a read-only annotation is treated as a write action. ChatGPT can remember an approval for one tool for the rest of a conversation, so use that only when you accept later posts going out without you reading them.

Can ChatGPT create the images for the batch too?

It can help you plan the visuals and write the captions, but the media itself has to come from somewhere — your own photos, a stock library, or a template tool. SceneSmith publishes the media you supply, and an Instagram feed post needs media, so a bulk image tool and this workflow fit together rather than replace each other.

How many posts can I schedule in one batch?

There is no single number, because the platform caps decide. Meta's publishing guide limits Instagram API publishing to 100 posts in a 24-hour period in one section and 50 in another, with carousels counting as one post, so plan for the lower figure. LinkedIn accepts scheduled posts from 10 minutes to three months ahead, which is comfortably more than a week of posts.

Which clients can connect to SceneSmith today?

The connection guide lists ChatGPT, Codex, Cursor, VS Code and Devin, plus other MCP-compatible clients. The client must support MCP initialization, tool discovery and OAuth 2.1 with PKCE, and you authorize your own SceneSmith account and destinations. A client the guide does not name should be treated as untested.

Does a batch need a paid ChatGPT plan?

The connection is not gated by SceneSmith, but it can be gated by your workspace. Custom MCP servers are added on ChatGPT on the web, and workspace permissions and security settings apply to adding and using them. If your workspace has no option to add a custom MCP server, use SceneSmith's own composer instead; its free workspace includes three drafts with previews, and a 7-day trial starts when you want to schedule or publish.

Sources

• Add custom MCP server — OpenAI developer documentation: adding a custom MCP server as a plugin, read and write tools, SSE and streaming HTTP, web-only access, workspace restrictions, write-action confirmation, read-only annotations, payload inspection and remembered approvals. Retrieved 2026-10-13.
• Content Publishing — Meta Instagram Platform documentation: professional accounts, Page Publishing Authorization, media_type values including REELS, carousels up to 10 items, the 100-per-24-hour rate limit and the 50-per-24-hour carousel limit, and the recommendation that apps enforce limits when they schedule for users. Retrieved 2026-10-13.
• Schedule posts — LinkedIn Help: scheduling from 10 minutes to three months ahead, UTC conversion from device time zone, and post types that cannot be scheduled. Retrieved 2026-10-13.
• SceneSmith connection guide and MCP connections page — production endpoint, supported clients, OAuth 2.1 with PKCE, and the safe first test. Read 2026-10-13.
• SceneSmith homepage — the one-instruction-to-three-drafts demonstration, one approval before every post, live destinations, and the free workspace's three drafts with previews. Read 2026-10-13.