Guide · By PostBoost team · · Markdown
Short answer: a scheduled post that never went live almost always failed for one of six reasons: the account's connection expired or was revoked, the network refused the image or video, the text broke one of the network's rules, the network's daily limit or spam filter stopped it, the post was never really scheduled, or the network had a temporary error. Open the failed post and read its error first: it names the cause. Fix that, then retry the failed account instead of creating the post again. To catch most of these before the publish time, run the post through the free social media post checker.
Start with the error message
A scheduler can only report what the network sends back, but that message is usually specific: "The submitted image with aspect ratio ... cannot be published", "The Instagram account is restricted", an expired token. Guessing wastes time; the message tells you which of the causes below you are looking at.
In PostBoost: a post that failed on any of its accounts gets the Failed status, even when the others published, and while the workspace has failed posts, Posts shows a Failed tab that lists them. Open the post and go to its Delivery tab: every account has its own line with its status, a View on platform icon link when it went live, and the network's error message in red when it failed. A post can be live on four networks and failed on one; the Delivery tab shows which. If you use webhooks, each failed account also sends a post.publishing_failed event, so a failure can reach Slack or your own system without anyone opening the dashboard.
The 6 causes, in the order to check them
| # | Cause | Typical sign | Fix |
|---|---|---|---|
| 1 | Connection expired or revoked | Every post for one account fails at once | Reconnect the account, then retry |
| 2 | Media the network refuses | One network fails, the others publish | Re-export the image or video, replace it, retry |
| 3 | Text that breaks a network rule | Only long posts or posts with certain characters fail | Shorten or edit that account's text, retry |
| 4 | Network limit or spam filter | Fails after a burst of posts, or on repeated content | Wait, spread posts out, vary the text |
| 5 | Never actually scheduled | No error at all, the post is still in drafts or waiting | Schedule it, or approve it |
| 6 | Temporary network error | Timeout or "try again" message | Retry once |
1. The account's connection expired or was revoked
A scheduler publishes with an access token the network gave it when you connected the account. When that token expires or is revoked, every post for that account fails until you connect it again. This is the cause when several posts for the same account fail together.
How long a connection lasts depends on the network:
- LinkedIn: access tokens last 60 days. Apps with refresh tokens can renew them for up to a year, after which the member must sign in again, and LinkedIn says it can revoke tokens at any time.
- Facebook Pages: Meta says long-lived Page tokens have no expiration date and expire only under certain conditions. In practice those conditions are a password change, removing the app's access, or losing your role on the Page.
- Instagram and Threads (direct login) use 60-day tokens that a scheduler must refresh before they run out.
In PostBoost: the connection health page (Tools → Connection health for workspace admins, https://postboost.co/app/<workspace>/health) lists every account with its status (healthy, failing, expiring or expired), the predicted expiry and the latest publish error, with a Reconnect button. PostBoost refreshes tokens where the network allows it and, when a connection's predicted expiry is 7, 3 and 1 days away, emails the workspace admins a reconnect link. When a network revokes access, the admins get a "We’ve Lost Connection - Action Required" email and the account shows Unauthorized with a Re-authenticate button on the Accounts page. Developers get the same as the account.token_expiring and account.disconnected webhooks, and GET /{workspaceUuid}/account-health in the API.
2. The network refused the image or video
The networks' publishing APIs are stricter than their apps. Instagram is the classic case: the app accepts a 3:4 portrait, but the publishing API takes images only between 4:5 and 1.91:1 and up to 8 MiB, and fails anything else with "The submitted image with aspect ratio ... cannot be published". Video has its own rules: Instagram's API asks for MOV or MP4.
The fix is to export the shape each network accepts. The image sizes table and resizer lists every placement with its source, and social media video specs does the same for video. For Instagram feed posts, 1080×1350 (4:5) is the safe portrait size; Instagram post size explains why 3:4 fails.
In PostBoost: an Instagram feed post with an image outside 4:5 to 1.91:1 that you create through the API or MCP comes back with an instagram_image_ratio warning in the response, and the validatePost dry run returns the same warning without saving anything. For posts made in the dashboard, the pre-publish checker agent (Tools → Agents) flags a scheduled post with that problem before its publish time and proposes holding it; every proposal waits in the Approval queue for a person. Either way you see it before the publish time, not after. PostBoost does not crop the image for you.
3. The text broke a network rule
Most schedulers stop you from scheduling a caption over a network's character limit. The failures that get through are rules a plain counter misses:
- YouTube refuses a title or description containing
<or>, so a heart typed as<3or an arrow typed as->fails the upload. See YouTube title character limit. - Threads counts emoji by their bytes, so a caption that looks like 480 characters can be over the 500 limit. See Threads character limit.
- X allows 280 characters on most accounts and counts every link as 23.
In PostBoost: the composer counts each network's limit as you type, and a scheduled post that breaks a network's requirements says "This post is scheduled, but it will not be published on ... until you fix ... issue" instead of failing silently. Two checks go further than the composer: a post created through the API or MCP with forbidden YouTube characters or a Threads caption that may be too long comes back with a youtube_forbidden_characters or threads_too_long warning, and the pre-publish checker agent flags the same problems on scheduled posts made in the dashboard. Paste any caption into the free social media character counter to see how each network counts it.
4. A network limit or spam filter stopped it
Networks limit how much an app can publish for one account. Instagram's error list includes "You reached maximum number of posts that is allowed to be published by Content Publishing API", with Meta's advice to try again the following day. Instagram also blocks posts it suspects are spam ("We restrict certain activity to protect our community") and refuses all posts from an account that is restricted until you sign in to the Instagram app and complete what it asks.
The fix is to space posts out, avoid posting the same text to the same account again and again, and clear any warning in the network's own app.
In PostBoost: for posts created through the API, MCP or an AI agent, PostBoost refuses a post that repeats the same content for an account within 24 hours and caps posts per account per UTC day (for example 25 for Instagram), so an automation loop can't trip the network's limits. The guardrails docs list the caps per network.
5. The post was never actually scheduled
Sometimes nothing failed: the post never left. Check that it is not still a draft, that the date is in the future in the timezone you meant, and that it is not waiting for someone to approve it.
In PostBoost: a post that waits for a teammate's approval shows Needs approval and is not published until someone who can approve it clicks Approve. The API reports the same with needs_approval: true when a post is scheduled or queued.
6. The network had a temporary error
Networks time out too. Instagram's own error list has "It takes too long to download the media" and "Create media fail, please try to re-create media", both with the advice to try again. If the error looks like a timeout or a server error and nothing else is wrong, one retry usually works.
Retry the failed account, not the whole post
Creating the post again publishes a duplicate on every network that already worked. Retry only the account that failed, after fixing the cause.
In PostBoost: on the post's Delivery tab, the failed account has a Retry button. It publishes that account again and leaves the accounts that already went live alone. For an X or Threads thread that stopped halfway, the button's label is Continue publishing: "Publishing continues from the part that failed. The parts already published stay as they are." From the API, the same is POST /{workspaceUuid}/posts/retry/{postUuid}/{accountUuid} (retryPostAccount); it answers 422 if the account did not fail or the post is still publishing.
Catch failures before the publish time
Most of these causes can be found before the post is due, which is how PostBoost's publishing is set up:
- Paste the caption into the social media post checker: it checks length, hashtags, media count and video length per network, with the same rules PostBoost checks before it sends a post.
- Check the image shape with the image sizes resizer.
- In PostBoost, turn on the pre-publish checker under Tools → Agents (admins and members). It looks at the posts scheduled for the next 24 hours (up to 168) and proposes holding any post whose account needs reconnecting, that fails a network's requirements or has a dead link. A person approves each proposal; the agent never changes a post on its own. The agents docs have the details.
Let an AI agent triage failed posts
PostBoost's failed-post fixer agent looks at the posts that failed in the last 24 hours (up to 72) and proposes one next step per failed account: reconnect the account, retry with the caption cut to the network's limit when the text was too long, or retry unchanged after a rate limit, timeout or server error. It uses no AI credits and never retries on its own; a person approves each proposal.
With the PostBoost MCP server connected to Claude or another assistant, you can run the whole loop from a chat:
Run the failed-post fixer in my workspace, list its pending proposals with the error for each one, and approve the retries for accounts that don't need reconnecting.
The assistant calls runAgent, then listAgentProposals with status: pending, and approveAgentProposal for the ones you agreed to (create the agent once under Tools → Agents if the workspace has none). With the Claude connector, the shorter route is list-posts filtered to failed posts, get-post to read each account's error, and retry-failed-post for one account. Reconnecting stays a manual step: it needs you to sign in to the network. Let an AI agent post to social media safely covers the guardrails that keep an agent from making things worse.
Do this in PostBoost
Free tools for this guide
Free tool
Social media post checker: check your post before you publish
Check a post before posting: character limits, hashtags, media count, video length and see-more cut-offs for X, Instagram, LinkedIn, TikTok and more.
Free tool
Social media image sizes, with a free resizer
Image sizes for Instagram, Facebook, LinkedIn, X, TikTok, YouTube, Pinterest and more, each with its source. Crop and resize your image in the browser.
Free tool
Social media video sizes and lengths cheat sheet
Video ratio, resolution, length, file size and formats for Reels, TikTok, Shorts, YouTube, X, LinkedIn, Pinterest and more, each with its source.
FAQ
Why didn't my scheduled post publish?
In most cases the account's connection expired or was revoked, the network refused the image or video, the text broke a network rule, the network's daily limit or spam filter stopped it, or the post was never actually scheduled (it is still a draft or waiting for approval). The error on the failed post names the cause.
Why do scheduled posts fail after a password change?
Changing your password, removing the app's access or losing your role on a Facebook Page can invalidate the access a scheduler holds. Every post for that account then fails until you reconnect it. Meta says long-lived Page tokens have no expiry date but can be invalidated.
How long does a social media connection last?
It depends on the network. LinkedIn access tokens last 60 days and refresh tokens a year, after which the member must sign in again. Meta's long-lived user tokens last about 60 days, while Page tokens have no fixed expiry. A good scheduler refreshes what it can and warns you before the rest runs out.
Why did Instagram reject my scheduled image?
Instagram's publishing API accepts images between 4:5 and 1.91:1 and up to 8 MiB, as JPEG. A taller image, such as 3:4 or 9:16, fails with an aspect ratio error even though the Instagram app accepts 3:4. Crop it to 1080×1350 (4:5) and schedule it again.
Can I retry a failed post without creating it again?
In PostBoost, yes. A failed account on a post can be published again from the post, or with the retry endpoint of the API. Fix the cause first, for example reconnect the account or replace the image, or the retry fails the same way.
How do I stop scheduled posts from failing?
Fix problems before the publish time: keep connections healthy and reconnect when warned, export images in the shapes each network accepts, and check every post against each network's rules before you schedule it. PostBoost's pre-publish checker agent does this for the next 24 hours of posts and proposes holding any that would fail.
Sources
- Meta for Developers: Instagram Platform error codes (2207009 aspect ratio, 2207004 size, 2207042 daily limit, 2207050, 2207051)
- Meta for Developers: Long-lived access tokens (user tokens about 60 days, Page tokens no expiration date)
- LinkedIn (Microsoft Learn): Refresh tokens (access tokens 60 days, refresh tokens one year)
- PostBoost docs: Connection health
- PostBoost docs: Agents (pre-publish checker, failed-post fixer)
- PostBoost docs: Agent guardrails (dry run, daily caps)
- PostBoost docs: Webhooks (post.publishing_failed, account.disconnected)