How-To Guide
How should a brand approve the exact image and caption that will be posted?
Approve one image and one caption as a single record with a checksum, an approver, a version and a destination check.
By the Superdirector editorial team. The approval-workflow pages and Meta's developer documentation cited here were read November 14, 2026; the Superdirector pricing page was read November 14, 2026. Every worked example in this guide is a labelled hypothetical, not a customer account.
At a glance
How should a brand approve the exact image and caption that will be posted?
Approve one image and one caption as a single record, not as an email thread. The record names the exact image file with a checksum, the exact caption text, the alt text, the platform and the account, and it says who approved it, when and which version. An approval that binds to a version turns the approval into proof: if someone edits the post after sign-off, the approval visibly breaks and the post goes back to review. Before the post goes out, verify the queued copy against the record; after it goes out, compare the live post with the record once more, because platforms process uploads in documented ways, such as cropping carousel images to the first image. That last check is the destination comparison, and it is what the record is for.
- The record fields that make an approval exact: image file and checksum, caption text, alt text, platform, account, approver, approval time and version.
- Approval binds to a version, the rule Kontentino states outright: edits after approval break the approval and route the post back to review.
- The destination comparison reads the live post's own fields, caption, alt_text, permalink and timestamp, against the record before anyone calls the post done.
- In an AI-run account, the same record discipline applies, and Superdirector (ours) publishes eligible Instagram carousels and single X text posts under the account's standing posting permission, with the approval step on your calendar, not on each post.
The scope of one approval, with the approved post at the centre. Original diagram, drawn from the approval-record field list in this guide.
Make it work for your brand
Find it. Remake it. Post it. Superdirector turns the viral formats in your niche into your brand's carousels, X posts and video scripts, and keeps the calendar full.
Approve one image and one caption as a single record
Most posts are not approved as one thing. The caption gets a yes in a Slack thread, the image gets a heart reaction, the alt text gets nothing, and the publish date lives in a third place. When the post then goes out wrong, there is no record to check it against, so the dispute settles on memory. Predis's approval guide names the root problem plainly: the content is created once, but the review keeps happening in fragments, and the schedule slips while the loop repeats.
The fix is to approve the post as one object. A single record carries every field the destination will show, and the approval is that record being signed. Fill it out like this:
- Image: the exported file, stored where the record lives, with a checksum written into the record.
- Caption: the exact submitted text, including any mandatory lines, hashtags and the link.
- Alt text: written at approval time, because Meta's media reference treats alt_text as its own readback field for accessibility.
- Surface: the platform and the account it publishes from, because a multi-account brand has more than one destination for the same idea.
- Approval: who approved it, when, and which version of the draft got the approval.
This is the shape Kontentino's approval stage documentation makes machine-enforced: it says the stage "is locked. Copy, visual, and date are frozen, and the approval is recorded: who, when, which version." The fields above are the manual-team equivalent, and they give you something an inbox never can: one object to open when someone asks which version shipped.
Name what the approval covers and what it does not
An approval is a decision about a post, not about a topic. Say what the yes covers so no one reuses it. One approval covers one post: the exact image and caption in the record, on one platform and account, inside the stated publish window. It does not cover a later image swap, a re-written hook, a different account, or next month's reuse of the same asset.
Hootsuite's workflow guide calls the failure mode out by name: "Post-approval edits that bypass the workflow are one of the biggest risks", and its stated remedy is routing any change made after sign-off back through the chain. A scope line in the record is what lets you tell a legitimate correction from a bypass.
Two scope rules worth writing down:
- Edits to the image or caption after approval require a new approval, because the thing approved no longer exists.
- Metadata that does not change what the reader sees, such as a corrected internal note or a changed internal tracking field, does not reopen the approval.
Bind the record to the exact image bytes
The word "exact" lives or dies at the file level. Keep the exported image file with the record and write a checksum into it, so the record names bytes rather than intentions. Any tool that computes checksums works; the point is that a changed image is detectable instead of silent.
Two honest limits belong next to the file line in the record:
- The platform reprocesses the upload by design. Meta's publishing documentation states JPEG is the only supported image format for API publishing and that carousel images are all cropped based on the first image in the carousel, with a 1:1 default. The bytes that land are the platform's render of the approved file, not a byte-identical copy.
- The API is one-way about the visual. A published media object's caption and cover cannot be swapped through the API afterwards; the realistic post-publish fix is an in-app edit or delete-and-repost, decided by correction rules.
Both limits mean the record's checksum is a guard on what you submit, not a promise about what the destination serves. That is why the record and the destination comparison are separate checks.
Record the approver, the version and the account
An approval without a named approver and a version is a mood, not a control. Kontentino's version-tracking section states the rule this article adopts: "Approval binds to a version", so an edit after approval visibly breaks the approval and the post routes back to review. Without that rule, approved posts drift, and "approved" stops meaning anything.
Name one approver per post. Kontentino's reviewer-versus-approver split is deliberate: committees comment, individuals approve. For client work, AgentMio's practice of granting the client a limited reviewer role that shows only their own queued content, with approve, reject and comment buttons, keeps the approval power where the accountability is without opening the whole calendar.
Keep the who-when-which-version line for three reasons Hootsuite's guide attributes to audit trails: agencies need it as dispute insurance when a client says the approval never happened, in-house teams need it as institutional memory, and regulated teams need it as the record they produce in an examination. An inbox is not that record, as Kontentino puts it, because emails get deleted and "Looks good!" does not reference a version.
Approval binds to a version: the approved version publishes, while an edit sends the post back to review. Original diagram, based on the version rule attributed in the body.
Send AI drafts through the same approval review
When the first draft is written by an AI, the approval question does not change; it sharpens. AgentMio's guide is emphatic about where AI belongs: AI drafts fit the draft stage, replacing the blank page, while a human reviewer still checks voice, accuracy and brand fit before anything moves forward. Hootsuite says the same from its side: AI-generated content goes through the same approval process as human-created content.
Predis's video guide builds the same idea into a chain of named review layers, from the initial AI-generated draft through script review, visual and design review, compliance or legal review and final publishing approval, each layer with one owner. For image posts the equivalent chain is shorter but the order is the same: the AI supplies the draft pair, the reviewer checks it against the brand and the product, the approver signs the record, and nothing publishes ahead of the signature.
Two AI-specific lines that belong in the review:
- Disclosure where the platform requires it: Meta's API supports a self-disclosure of AI usage on a media container through the is_ai_generated parameter, and the label travels with the post, so it is a pre-publication decision, not a cleanup.
- Product fidelity: an AI-generated image is checked against the real product during the review, not after publication; our accuracy checklist for AI product images is the pre-publication pass that review step performs.
Compare the post that landed with the approved record
The approval is done when the record is signed. The post is done when the published object matches it. Those are different checks, and the second one is the destination comparison. Meta's media reference lists the exact fields to read back on an Instagram post: caption, alt_text for the image's accessibility text, media_url and permalink for what and where, and the timestamp. Put each field next to its line in the record.
The example below is a labelled hypothetical, no real brand and no real post, showing the comparison as a filled-out check. Fields marked mismatch are the ones a correction decision would weigh:
| Readback field | Approved record says | Live post says | Verdict |
|---|---|---|---|
| Caption | Exact submitted text, ending with the link | Same text, link present | Match |
| Alt text | Written at approval, 12 words | Same alt text read back | Match |
| Image | JPEG export, checksum e3...c9 | JPEG served, cropped to the first carousel image, 1:1 by platform rule | Match, processed |
| Account | Instagram @labelled-hypothetical | Published under the same handle | Match |
| Publish time | Window opens Friday 10:00 CT | Timestamp Saturday 10:00 CT | Mismatch: reschedule check |
The two Meta-documented platform effects sit inside the comparison rather than surprising you. First, processing: the API guide crops carousel images to the first image and accepts JPEG only, so a rendered-differently image is usually the platform's documented behaviour, a match with processing noted, not an error. Second, readback caution: the media reference treats media_url as an optional field and instructs apps to fall back to the permalink or thumbnail when it is absent, so a missing media_url in the readback is a field the API omitted, not proof the image is wrong.
One more caution the reference itself raises: caption readbacks can leave out the @ symbol unless the reading credential can perform admin-equivalent tasks on the page. A character-level diff between record and readback needs that context before anyone calls it a mismatch. Rate-limit aware teams also keep the read light: Instagram accounts take up to 100 API-published posts in a 24-hour moving period, and a verification read is not a publish, but the comparison still reads fields from the same rate-limited service.
When a field genuinely disagrees with the record, the move is the same one Hootsuite prescribes for post-approval changes: the change goes back through a decision, in this case a correction decision, before anything else. Our correction-rules guide owns that branch: pause authority, the replacement, the closeout.
The destination comparison as a matrix, one row per read-back field. Original diagram, based on the Meta media fields cited in the body.
Where Superdirector fits and where it does not
Superdirector (ours) sits upstream and downstream of the approval work this guide describes. It produces the draft pair you would review: it watches what is working in your niche, makes branded carousels and X posts, and fills the calendar. On the publishing side, the pricing page is specific: Growth is $199/month with two accounts, and on Growth it "publishes eligible Instagram carousels and single X text posts by themselves once posting is turned on for an account." Starter at $29/month is the hands-on creation tier, with a 7-day free trial that needs a card at checkout.
That publishing is account-level, with no per-post approval step in between, and the same page says exactly where the standing decisions live: "Posting starts only when you turn it on for an account, after one warning", and "You make the important calls: what posts, what changes and what gets paid." In other words, the account-level permission is the control surface, and the per-post records, the quality bar and the destination checks stay with the people running the brand, exactly the split the AI-in-approval section above preserves: AI drafts the pair, the brand's review signs it, and publication waits for its permission on the account, not for a nod on each post.
What Superdirector is not in this workflow: the correction engine and the approval record are yours. "You stay the publisher," the pricing page says; it also reminds you to label AI content where a platform requires it and never to publish invented reviews or results. Brands that want a separate-approval queue on every post as a product feature can pair Superdirector's creation with an approval-workflow tool and keep the record discipline of this article; brands that run the AI account directly use the calendar and the account-level permission as their control, and apply the same three checks after the fact: the record, the queue state, and the destination comparison.
The honest way to decide: if your risk is that the wrong image ships to a high-stakes account, the record-and-compare discipline above costs an afternoon and works with any publisher. If your risk is an empty calendar, the account-level operation is what fills it, and the same discipline tells you afterwards that it shipped what it should have.