The old handoff for short-form advertising was easy to count. A production partner delivered a folder of finished cuts, the brand approved each file, and the media team selected a few for launch. TikTok’s latest Smart+ workflow changes that operating model. Brand-owned assets, creator content, and Symphony-generated creative can now enter a shared approval and activation flow, while recommendation, auto-add, and material-level reporting expand what can happen after the first campaign setup.
The practical consequence is not “let the platform make the ads.” It is that the unit of procurement is shifting. If the platform will choose from a changing pool, the brand needs to buy a governed pool of assets rather than a fixed number of MP4s.
What TikTok has actually announced
On October 5, 2026, TikTok said Smart+ Creative Selection brings advertiser assets, Symphony-generated creative, and creator content into one approval workflow. The company did not say that every account or market receives an identical feature set, so availability still needs to be checked inside the advertiser’s own Ads Manager.
The scale of the new container matters. TikTok’s January 2026 help documentation says an ad can function as an asset group, combining creative, copy, product sets, destinations, identity and tracking. Where available, an ad can contain up to 50 video, image or carousel assets and five texts, with multiple ads inside an ad group.
TikTok’s documentation for Recommended Creatives adds an operational detail: the pool may include generated variants, existing assets and authorized posts. Auto-add can introduce new creative during a campaign, and advertisers may require approval before those assets enter delivery. The upgraded Smart+ documentation also describes reporting grouped by creative material ID.
Those facts show that TikTok is broadening automated creative management. They do not prove that AI, creator and brand-owned assets carry the same rights, that every automatic recommendation is appropriate, or that automation causes better performance in every campaign.
Why a video count is no longer a useful scope
“Twenty videos” says very little once a single ad container can connect many videos, texts, products and destinations. Twenty edits may be locked to one product and one market. Ten well-governed assets may support far more valid tests.
The risk profile also changes by source:
- Brand-owned footage is usually easier to reuse, but product claims, packaging and prices can expire.
- Creator content may feel more native, but usage windows, identity, account authorization and paid-media rights need separate checks.
- AI-generated creative can scale quickly, but the brand still needs records of source inputs, realistic alterations, approved claims and disclosure treatment.
Automation expands the combination space available to the platform. Procurement therefore needs to define which assets may enter that space and under what conditions.
ONCE’s three-layer, six-field pool
A campaign-ready pool should carry three layers of control.
Layer 1: source class
Mark every video, image, post or generated variant as brand-owned, creator-authorized or AI-generated. Hybrid work needs more than one marker. “Live-action product plus generated background” should not be flattened into a generic brand asset because the provenance and review obligations differ.
Layer 2: permitted combinations
An asset is not approved in isolation. Specify the SKUs, claims, languages, creator or brand identity, product sets and destination URLs it can accompany. A creator license limited to one territory, or a demonstration filmed for an older model, should be excluded from other combinations.
Layer 3: release state
Use at least three states: eligible for automatic selection, approval required, and paused or expired. A paused or deleted asset should not silently re-enter through a later refresh. TikTok’s option to pair auto-add with an approval requirement gives brands an operational place to enforce this distinction.
Each asset then needs six fixed fields:
- unique asset ID and current version;
- source class and original file location;
- rights scope, account, territory and expiry;
- approved product, claim, language and market;
- permitted copy, identity, product-set and destination pairings;
- release state, last reviewer and withdrawal timestamp.
If one of those fields cannot be completed, the asset should not enter an automatically selectable pool by default.
Quote the work as four cost packages
Production still includes shooting, editing, sound and finishing. A creative-pool project adds costs that a per-video quote can hide.
The base production package captures reusable product, performance, sound and still assets. The controlled variation package creates hooks, ratios, pacing, caption and language versions. The rights and data package maps licenses, asset IDs, products and approved claims. The operating package covers additions, pauses, expiries and material-level reporting reviews.
A useful budgeting expression is:
Total pool cost = base production + controlled variants + rights and metadata governance + in-flight refresh and withdrawal
This is not an industry price benchmark. It is a way to compare two proposals that both promise “20 videos.” One may deliver only files; the other delivers assets that can be combined, measured and withdrawn without rebuilding the campaign.
Accept the pool twice
Before launch, reverse-sample the pool. Pick an asset at random and ask why it is allowed to run with the current product, copy, identity and destination. If the creator license, AI status or product claim cannot be traced, remove it from automatic selection.
After launch, read back four things:
- which creative material IDs actually entered delivery;
- which were added manually, recommended or auto-added;
- which copy, product-set and destination combinations actually ran;
- whether pauses, expiries and withdrawals took effect within the agreed response time.
Material-level reporting makes this readback more practical, but it still does not prove creative causality. Product, audience, bid, timing and landing-page changes remain confounders that need to be controlled in a test.
Creator assets also need contract governance beyond a platform access mechanism. Our guide to creator video procurement explains why a platform authorization code is not a rights agreement. For hybrid and generated assets, the AI ad provenance delivery framework shows how to keep source, change and release records connected.
When automatic selection earns a wider role
A governed pool makes sense when a brand has a stable catalog, recurring media demand, standard license templates, and a team that can review material-level reporting. Automation can then explore inside a space the brand has already approved.
Manual, fixed combinations remain the better choice for regulated or high-consequence claims, narrowly licensed creator work, and flagship storytelling that must preserve a deliberate sequence. The objective is not maximum automation. It is an auditable, reversible combination space.
The next time a brief asks for 30 TikTok videos, ask one question before pricing: will those assets enter Smart+ recommendation, automatic selection or in-flight refresh? If they will, the scope must include source, rights, combination, status, reporting and withdrawal rules.
TikTok can choose from the pool. The brand still has to decide what is allowed into it.
