Playbooks · 5 min read
A Story Mention Is Not a Message You Can Reply To
Somebody posts a photograph of your product to their story and sticks your handle on it. Nice moment. The obvious idea follows immediately: thank every one of those automatically, and invite the good ones into an ambassador programme.
Before designing any of that, it is worth being exact about what an automation can actually respond to, because "story mention" covers two completely different things and only one of them produces something to reply to.
What produces an event
An automated reply needs a trigger, and a trigger needs text. Somebody has to have typed words that a keyword can match. That is the mechanism, and everything else follows from it.
Produces something to answer
Produces nothing to answer
The right-hand column is not a limitation of this product. There is no text in any of it. There is nothing for a keyword to match, nothing a reply is a reply to, and no thread the platform's rules would let you open on your own initiative — nobody on the official API can message a person who has not messaged them.
If they did not type anything, there is no message, and a reply to nothing is just an unsolicited DM with better intentions.
So "never miss a story mention again" is not a promise this or anything else on the official API can keep. What you can do is design so that the people you want to hear from end up typing.
Designing for typing
This is the actual craft, and it is upstream of every setting in the builder.
- Ask on your own story, in words that require words. A question sticker produces a typed answer, and a typed answer arrives as a DM threaded under the story. "Tag us" produces a tag. "Tell us which one you got" produces a message.
- Put a word in the caption, not just an invitation to tag. "Tag us and comment
SHAREif we can repost" turns a passive mention into a comment you may legitimately answer. - Make the ask specific. "Any thoughts?" gets nothing. "Which shade did you go for?" gets answers, and the answers are worth reading afterwards.
- Accept that the quiet ones stay quiet. Some of your best customers will tag you and never type. Reaching them is a manual job — you, in the app, replying by hand. That is not a workflow gap. It is the shape of the platform.
A tag is still not permission
Since this whole area is about people posting your product, the consent point has to be said even though it has its own post.
Somebody tagging you has shared something with their followers. They have not licensed it to you for advertising, and in most places the copyright is still theirs. Reposting without asking is common and is still taking something.
If you are going to ask, ask in a fixed template rather than a generated reply. A consent request is a sentence whose wording is the record of what somebody agreed to, and a generated one phrases it slightly differently every time — which means forty people agreed to forty subtly different things and none of them is quite the sentence you meant. Templated replies are free, so here the correct choice and the free one are the same choice.
What an ambassador programme can and cannot get from this
The honest inventory, because the gap between what people imagine and what exists is wide.
There is no ambassador feature. No tiers, no points, no tracking of who posted what, no image recognition — nothing here reads pictures. No affiliate links, no discount code generation, no attribution, no analytics, no way to see how many mentions you received, and no CRM to push anybody into. Leads leave as a CSV containing a handle, the post a message came from, the words typed, and a timestamp.
What automation genuinely does for a programme is one step: the intake. Somebody types the word, and a fixed reply goes out explaining what the programme is, what it asks of them, and where to apply properly. That reply is free, arrives instantly at any hour, and is identical for everyone — which for a programme with terms attached is a feature rather than a compromise.
The two things that must never be automated here
A message implying you have already used somebody's content. "We've featured your post!" is notification, not consent, and it arrives after the fact.
A reply to somebody withdrawing permission. That is a person's job, immediately. Put remove, take down, delete, hata do, unauthorised on the negative-keyword list. Matching folds case, punctuation and Latin accents, and it matches whole words — remove will not catch removed, so write both.
Cost, and the clocks
Templated replies are free and unlimited, so an ambassador intake reply and a consent request cost nothing however many people send them. A credit is spent only when the AI writes a new reply, one credit per reply — and for this use case you actively do not want that, because the wording is the point. Free tier is 100 credits with no card; packs start at ₹499.
Official Graph API only. Comments answerable for seven days, counted from each comment. DM threads open for twenty-four hours from their last message, restarted only when they write again — so there is no follow-up chase, and a person who did not reply has given you their answer. Takeover stands the automation down the moment you reply by hand, which for an ambassador conversation is early and often.
If the wider consent question is what you came for, the UGC post is the one that owns it. For the programme itself rather than the plumbing, ambassador programmes covers the design.
