Trial reels manual vs auto graduation comes down to who shares the Reel to your followers: you, from the Instagram app (manual), or Instagram itself if the Reel performs with non-followers (auto, sent as SS_PERFORMANCE). Creator OS exposes both through one field on the post, so an AI agent can choose per Reel.
Trial Reel posted to non-followers only
|
+---- manual / true ----> YOU open Instagram app,
| tap Share to followers
|
+---- auto / SS_PERFORMANCE --> INSTAGRAM decides
after non-follower test
Either path: the API returns no "graduated" field.
Track it outside Instagram (posts:list, analytics).What manual vs auto graduation means for Trial Reels
A Trial Reel is shown to non-followers first, and graduation is the step that puts it in front of your followers. Manual graduation means the Reel stays in trial until you share it yourself from the Instagram app. Auto graduation means Instagram watches how non-followers respond and shares it to your followers if it does well.
Only Reels go through this. On the Creator OS side, the request is a normal post with post_type set to short_video plus a trial_reel field. That field accepts "manual", "auto", or true (which means manual). In the advanced form you send trialParams.graduationStrategy as MANUAL or SS_PERFORMANCE. The MCP tool create_post takes the same trial_reel field.
One limit to design around: Instagram only permits Trial Reels on accounts with 1,000 or more followers. Below that, Creator OS refuses the post up front with the error trial_reel_needs_followers and suggests a regular Reel instead. An agent that hard codes trial Reels will hit that wall on newer accounts, so check follower stats before you build the branch.
How manual graduation works: you share to followers from the Instagram app
With manual graduation the Reel never moves to followers unless a human acts in the Instagram app. Your agent keeps posting test Reels at volume while a person decides which ones graduate.
That makes manual the right default for most agent setups. Your agent’s job ends when the Reel is published. The decision stays with someone who can see the comments, the tone and the timing. Here is the shape of the call through the CLI:
creatoros posts:create \
--platforms instagram \
--media https://cdn.example.com/hook-test-04.mp4 \
--text "The 3 second rule nobody tells you about Reels" \
--title "Trial Reel 04"For the trial flag itself, use the REST API (POST /v1/posts) or the MCP create_post tool with trial_reel: "manual". The docs at the post options page list the exact field placement.
One funnel detail worth knowing: the comment-to-DM keyword automation works on Trial Reels. Commenters are non-followers, so the DM arrives as a message request. If a trial Reel is driving keyword replies, your automation should be bound to that specific post.
How auto graduation (SS_PERFORMANCE) works: Instagram shares if it performs with non-followers
Auto graduation hands the decision to Instagram: if the Reel performs well with the non-follower audience, Instagram shares it to your followers without anyone opening the app. Send trial_reel: "auto" or trialParams.graduationStrategy: "SS_PERFORMANCE".
This is the setting for volume. An agent that posts fifteen test Reels a week cannot rely on a human remembering to graduate one. With auto, the Reels that find a non-follower audience reach your followers on their own. The weak ones stay in the trial and fade.
The tradeoff is control. Instagram’s definition of “performs well” is Instagram’s, not yours. A Reel that picks up cheap views from the wrong audience can graduate into your feed. That is a brand safety question, not a technical one, and you answer it with what you feed the agent, not with the field.
Why Instagram exposes no field saying whether or when a trial was shared
Instagram returns no field telling you whether a Trial Reel was shared to followers, or when. Plan around that instead of trying to read it back.
This matters most for auto. Your agent can post, read analytics and reply to comments, but it cannot ask Instagram “did this graduate yet?” and get an answer. Any dashboard that claims to show graduation status for an auto trial is inferring it, not reading it.
So the agent has to hold its own state. Practical options:
- Record the intent at post time (manual or auto) in your own database or post title.
- Pull per-post analytics with
creatoros analytics:posts --platform instagramand watch the reach curve. A jump from non-follower views toward follower views is a hint, not proof. - For manual trials, ask the human to confirm after they share from the app. There is no API shortcut for that step.
- For auto trials, treat analytics as the source of truth and never claim a Reel “graduated”.
Can an AI agent detect graduation? What the API does and does not return
An agent cannot detect graduation directly, but it can read the signals that move after graduation: watch time, skip rate and total reach. Creator OS returns Instagram Reel watch stats per post from GET /v1/analytics/posts, which the MCP tool get_post_analytics also serves.
The three fields that matter:
| Field | What it tells you | Unit |
|---|---|---|
reelsSkipRate |
Share of views that swiped away in the first 3 seconds | percent |
igReelsAvgWatchTime |
Average watch time per view | milliseconds |
igReelsVideoViewTotalTime |
Total watch time across all views | milliseconds |
Meta reports these for Reels only. They do not say “graduated”. They tell you whether the Reel held attention, which is what you actually wanted to know. A low skip rate and high average watch time on a trial Reel is the signal to make another one in that style.
Here is a small workflow an agent can run: post a trial Reel with trial_reel: "auto", wait three days, then call get_post_analytics and compare reelsSkipRate across the batch. If the best performing hook has a much lower skip rate than the rest, the agent’s prompt can steer the next five Reels toward that hook format, length and topic. Creator OS provides the posting and the numbers. Your prompt or workflow decides what changes. A practical starting point for the analytics loop is described in the Trial Reels loop post.
Which mode to use when an agent posts many test Reels
Use auto for batches and manual for anything you would not want appearing in your feed unannounced. That is the short rule. Here is how it plays out:
| Situation | Mode | Why |
|---|---|---|
| Agent posts 10 to 20 hook tests a week | auto (SS_PERFORMANCE) |
Nobody will graduate 20 Reels by hand |
| New offer, launch, or sensitive topic | manual (true) |
A human approves before followers see it |
| Account under 1,000 followers | regular Reel | Trial Reels are refused with trial_reel_needs_followers |
| Client work on an Agency plan | manual by default | Each client workspace has its own scoped API key and its own comfort level |
| Experimenting with format, not message | auto | Format wins are cheap to test at volume |
Do not mix the two randomly. Pick a mode per workspace and hold it, so the analytics you compare are comparable. An agency workspace running auto tests can be compared with another workspace running auto tests. One running manual tells you something different: it measures what humans chose to graduate, which is a separate question.
How to build a decision rule: volume, risk, and brand safety
Write the rule into the agent’s prompt so the agent never picks the field on vibes. It has three inputs: volume, risk and brand safety.
Volume. Count the Reels in the batch. Above roughly five test Reels a day, manual graduation stops being reliable because humans skip it. At that volume, auto.
Risk. Ask whether a surprise in your follower feed costs you anything. A product launch, a hiring post, anything touching a client’s name: manual. A hook test for a tip series: auto.
Brand safety. Auto graduation puts Instagram’s performance call in charge of your feed. If that is unacceptable, keep the account on manual and accept the lower test volume. The field is not the problem; the volume expectation is.
A single instruction in the agent prompt does the job:
Before creating a Trial Reel, decide graduation:
- batch > 5 per day OR format test -> trial_reel "auto"
- launch, client name, or sensitive topic -> trial_reel "manual"
- account followers < 1000 -> post a regular Reel
After 3 days, pull get_post_analytics and rank the batch by
reelsSkipRate. Report the best hook. Never claim graduation.That last line matters. Your agent should be able to say “this Reel held attention better than the others” and stop there. It cannot say “this Reel graduated”, because Instagram gives no field for it. Set that expectation in the prompt and in whatever report the agent produces.
If you run this per client, the Agency plans carry a scoped API key for each client workspace, so one client’s key cannot post into another’s account. That is the same boundary you want around graduation rules: a rule per workspace, not per agent. The MCP setup page covers the read only versus read and write connections if you want a reporting agent that can pull analytics but never publish.
FAQ
What is the difference between manual and auto graduation for Trial Reels?
Manual graduation leaves the Reel in its non-follower trial until you share it to followers yourself from the Instagram app. Auto graduation lets Instagram share the Reel to followers if it performs well with non-followers. In Creator OS the choice is the trial_reel field set to "manual" or "auto", and the advanced form calls auto SS_PERFORMANCE.
How do I know if my trial reel was shared to followers?
For manual trials you know because a person tapped share in the Instagram app. For auto trials there is no direct signal. Instagram exposes no field saying whether or when a Trial Reel was shared to followers, so your own records and the per-post analytics from get_post_analytics are what you have.
Does the Instagram API tell you when a trial reel graduates?
No. Instagram returns no graduation status and no timestamp for the share to followers. Creator OS gives you the posting call and the Reel watch stats, including reelsSkipRate, igReelsAvgWatchTime and igReelsVideoViewTotalTime, but it does not invent a graduation field Instagram does not provide.
Which Trial Reels setting should an AI social media agent use?
Use auto when the agent posts test Reels at volume and a surprise in your feed is cheap. Use manual when the Reel ties to a launch, a client or a sensitive topic and a human should approve it. If the account has fewer than 1,000 followers, neither applies, because Instagram requires 1,000+ for Trial Reels.
Can an AI agent post Trial Reels automatically?
Yes. An agent holding a Creator OS API key can send POST /v1/posts with post_type: "short_video" and trial_reel set to manual or auto, or call the MCP tool create_post with the same field. It can then read the watch stats after publishing and use them to shape the next Reel.
Get started with Creator OS
Creator OS plans start at $14.99/month for 11 platforms plus Skool, with the API, the MCP server and the CLI included. You can create an account at https://www.creatoros.ca/sign-up and start for $0 with a free trial, cancel anytime.
Connect the Instagram account, add the hosted MCP server at https://mcp.creatoros.ca/mcp, and ask your agent to post one trial Reel in manual and one in auto. Read the analytics three days later and you will know which rule fits your account. To see the build in action, watch the walkthroughs on the KevBuildsApps YouTube channel and the launch video.
Get started for $0
Free trial for new subscribers. Cancel anytime.
