If you sell a TikTok Shop seller in Vietnam in a compliance-sensitive category such as health supplements, cosmetics or functional food, working with creators who write their own scripts, creator content is probably your largest source of orders and, quite possibly, your largest source of refunds at the same time. Traceability in Vietnam stopped being a documents exercise and became a content problem the day enforcement moved to the account rather than to the listing. A seller can have every certificate in a folder and still lose the month, because what gets judged now is what the creator said on camera. The categories that carry the heaviest documentary burden — health supplements, cosmetics, functional food and anything making a claim about the body — are also the categories where creators improvise most, because the product promises something and the natural way to sell a promise is to make it bigger. That combination is where the penalties live: commission frozen on the offending account for up to ninety days, content pulled, and the creator relationship spent on a video that cannot be re-uploaded.
Price the penalty before you decide how careful to be. A creator account with commission frozen for ninety days is three months of settlement that does not arrive, and the seller's loss is not the commission itself but everything spent to produce it: samples shipped, briefs written, negotiation time, and the campaign window that the content was supposed to fill. Take a mid-size programme: eight active creators, four samples each per quarter, plus the internal hours to brief and review. That is a quarter of creator spend producing content that cannot be posted, against a catalogue slot that now has no campaign in it. Add the part that is hardest to recover — the creator who was told their script was fine and then had their account penalised will not take the next brief at face value, and rebuilding that trust costs more than the samples did.
What the rules actually attach to
Three things get judged, and sellers usually prepare for only the first. The product must be documentable: proof of origin, the import declaration, a quality inspection report, and a label in Vietnamese — a label, not a translated sticker applied afterwards. The claim must be supportable: whatever the creator says the product does has to be something a document actually says, and 'the supplier told me' is not a document. And the content must be real: material generated by AI and presented as genuine experience is prohibited outright, which catches more than deepfakes — it catches a synthetic testimonial, a generated before-and-after, and a creator reading a script about a product they never used. The second and third are the ones a documents folder does not protect you from, because both are produced in the creator's script, and nobody in the approval chain ever reads the script.
The asymmetry is worth stating plainly. Preparing the documents is bounded work — a folder, a checklist, a week. Preparing the script is unbounded, because every creator writes a new one and every new script can introduce a sentence nobody has checked. Sellers who spend the week on documents and leave the script to the creator have protected the part that was already safe and left the exposed part untouched. The fix is not more caution, it is moving the check to where the risk is produced: the words, before they are recorded.
Step 1 — Turn the documents into a say / do-not-say list
Write the list in two columns and keep it to one page. Left column: what may be said, phrased exactly as a creator can say it — the product name, the category, the ingredients as they appear on the Vietnamese label, the usage instructions from the label, and any claim that appears verbatim in an inspection report. Right column: what may not be said, and this is the column that does the work — treating or preventing a condition, results within a stated number of days, comparisons against medicine or against a named competitor, 'clinically proven' without the report in hand, before-and-after visuals that were generated rather than shot, and any first-person experience the creator does not actually have. One page is the constraint that makes it usable. A six-page compliance document will be sent, acknowledged, and never opened again.
Write it in Vietnamese as the working version, even if your team operates in another language. The person who has to understand it at speed is the creator, in their own language, forty minutes before a shoot. A list that has to be translated in someone's head first is a list that gets paraphrased, and paraphrase is exactly how an approved claim becomes an unapproved one.
The two-column list is the artefact worth having, and it is also the thing nobody has time to draft properly. If you want a starting structure, register with DAMI and ask the team for the free creator content compliance checklist: the allowed and prohibited columns are pre-filled by category, with space to paste the exact wording from your own inspection report.
Step 2 — Attach the claim to the certificate that supports it
For every claim a creator is allowed to make, name the document that supports it, and keep them in the same place. This sounds bureaucratic until the first time you need it, which is usually not an audit but a dispute: a buyer complains, a competitor reports the video, or the platform asks, and the question is not whether you have the certificate but whether you can show which sentence it covers. A claim without a named document behind it is a claim that will be dropped from the next brief anyway, so the discipline pays for itself twice — it makes the current video defensible and it makes the next brief faster to write. One row per claim: the sentence, the document, the page or clause, and the date checked.
Watch the date column more than the document column. Certificates expire, inspection reports are issued per batch, and a report that covered the batch you were selling in March does not cover the batch you are selling in October. A claim that was defensible last quarter and is not defensible now is the most common way a compliant programme becomes non-compliant without anyone changing anything.
Step 3 — Give creators replacement lines, not just prohibitions
A list of prohibitions alone produces worse content, not safer content, because a creator told what not to say will fill the gap with something you have not reviewed. Give a replacement line for every prohibited one, written in the creator's voice and short enough to say on camera. 'Treats acne in three days' becomes 'this is the ingredient list, and here is what the label says it is for'. 'Clinically proven' becomes 'here is the inspection report, and this is the number in it'. 'I have been using it for a month' becomes 'I have been using it for four days, and this is what I have noticed so far' — which is both true and, counter-intuitively, more persuasive, because specificity reads as honesty. Hand over a script with the lines already swapped, not a document with the lines struck out.
Keep the creator's own voice in the replacement. A line a creator cannot say naturally will be delivered flat, and a flat delivery converts worse than a cautious one. Write the constraint, not the wording: 'say what the label says, in your own words' survives contact with a camera far better than a scripted sentence the creator has never said out loud.
Step 4 — Make sign-off a gate before anything posts
Make approval a gate with an owner. Every script or finished cut gets one named person who reads it against the list before it posts, and the record of that reading is kept with the creator: which version was approved, by whom, on what date, against which list version. This is the part that turns a policy into evidence. Without it, when a video is challenged you have a folder of certificates and no way to show that anyone checked the words. With it, the same folder becomes a defensible process, and the conversation with the platform is about one video instead of about your whole programme.
Budget the review time honestly. One careful read of a short script is five minutes; one careful read of a two-minute cut with on-screen text is fifteen. Multiply by the number of creators you run and decide whether that fits one person's week. If it does not, the answer is not to skip the review but to reduce the number of live claims — fewer creators making more careful statements beats more creators making statements nobody has time to check.
How DAMI keeps the list, the brief and the sign-off in one place
DAMI brings an 8-million-plus creator database, competitor and competitor-store creator search, automated outreach tasks, AI-assisted outreach, sample management, campaign management, store analytics, and multi-store team dashboards into one place. Compliance fails at the handover, not at the rule. Three steps keep the list attached to the brief and the approval attached to the record.
Step 1 — Send the list with the brief, through the creator's record
Deliver the say / do-not-say list and the brief as one package from the creator's record, with a version number. Output: every creator can show which list version they were working from, and you can, without searching messages.
Step 2 — Log the approval against the same record
Record who approved the script or cut, on what date, against which list version. Output: one line per piece of content that turns a policy into evidence when a video is challenged.
Step 3 — Keep claim and document on the SKU
Store the approved claim next to the certificate that supports it, with the date checked. Output: the next brief is written against claims you can still defend, and expired documents surface before they are quoted on camera.
Frequently asked questions
If my documents are all in order, can the creator still cause a violation?
Yes, and that is the normal case. The documents cover the product; the violation usually comes from what is said about it or from content presented as real experience that was not. A complete folder protects the listing. It does not protect the script, and the script is what gets judged.
Does the AI-generated content rule really apply to ordinary creator videos?
It applies to material generated by AI and presented as genuine, which covers synthetic testimonials, generated before-and-after visuals and scripts about products the creator never used. It is not limited to obvious deepfakes. If a viewer would reasonably believe the person on camera had the experience being described, the content has to be real.
How long does a commission freeze last, and what does it cost me?
Up to ninety days on the offending account. The seller's cost is rarely the commission itself; it is the samples, briefs, negotiation hours and campaign window spent producing content that cannot be posted, plus the trust lost with a creator who was told the script was fine.
Can I manage this with a shared document?
For the list itself, yes — one page in a shared doc is better than nothing. It stops working at the part that matters, which is proving which version each creator received and who approved what against it. A shared document has no record of who opened it, so when a video is challenged you can show the list but not the handover.
Conclusion
Every cost line in this article moves without anyone deciding to move it, and every one of them is cheaper to fix before the next campaign than after it. DAMI brings an 8-million-plus creator database, competitor and competitor-store creator search, automated outreach tasks, AI-assisted outreach, sample management, campaign management, store analytics, and multi-store team dashboards into one place.
Fifteen minutes today: open the last five creator briefs you sent in a sensitive category and count the sentences that make a claim about results, speed, or comparison with medicine or a competitor. For each one, ask whether a document says exactly that. The ones that do not get cut from the template now, before the next brief goes out with them in it.
Start with DAMI and run the next cycle from numbers you can defend.
