Creative briefs often begin with a demographic bundle: people aged 25–34, interested in technology, willing to pay for quality. That may help size an audience, but it does not tell a director what the audience is trying to do, which scene would earn attention, or what evidence would make a claim believable.
On October 13, 2026, Meta announced new AI-powered Facebook controls that let people add and manage interests for Feed and Reels, then explain in their own words what they want to see more or less of. Meta says recommendations can respond in real time.
The useful conclusion for brands is not “put more keywords in captions.” Meta has not promised distribution to brands that echo a user’s phrase, and its announcement does not establish identical availability for every account or market. The procurement change is upstream: replace vague audience labels with the situations, jobs, objections, and interests people would actually describe.
Separate the confirmed change from the marketing inference
Confirmed: Meta’s October 13 announcement says people can directly add and manage topics in “Your algorithm.” After selecting Interested or Not interested, they may also receive follow-up questions and use freeform language to refine recommendations. Meta says those inputs work alongside what its systems already understand from behavior.
That last point matters. Meta’s earlier explanation of how AI ranks Facebook and Instagram content says its systems combine many predictions, behavioral signals, and user feedback. The new tool adds explicit language; it does not replace the rest of ranking.
ONCE’s inference: when people can express a more nuanced interest, briefs organized around a precise job or context become more valuable than generic personas. This is a production strategy—not a disclosed ranking formula.
Still unknown: Meta has not announced a brand-facing report that reveals which sentence caused a particular post to be recommended. A marketer therefore cannot promise “natural-language match” as a delivery KPI. Measurement must stay with observable content and business signals.
Why a demographic persona cannot direct a shoot
“Outdoor enthusiast” might mean a first-time camper, a distance hiker, a parent preparing a child’s first trip, or someone who only wants to cook on a balcony. They can share age and income while needing entirely different product evidence.
A shootable brief preserves a complete expression: “I need camping gear that fits in a small trunk and that I can set up alone.” The sentence exposes the environment, the constraint, and the acceptance test. A team can now choose a car, stage a one-person action, show packed size, record the setup sequence, and decide whether an uninterrupted shot is needed to make the proof credible.
This complements ONCE’s approach to a recognizable brand video series. A series needs a stable viewing promise, but each episode needs a new question and obstacle. Audience language is not another tagging spreadsheet; it is the source of questions worth turning into episodes.
The ONCE LENS framework: turn audience language into a production unit
Before the next social-video commission, build a LENS sheet. Each row should serve one real expression of interest. Do not force ten intents into one video.
L — Language: preserve the audience’s words
Collect sentences from customer-support questions, site search, sales interviews, community discussions, comments, and return reasons. Do not immediately rewrite them as slogans. Sort them by job, environment, risk, and desired change.
“Does a robot vacuum work for a small apartment with a long-haired cat?” is more useful than “pet owner.” It gives production a hair type, a space condition, and a decision question. Brands still need to respect privacy and usage rights; a public comment is not automatically free ad copy.
E — Embody: convert the sentence into an observable scene
Every expression needs an action a camera can witness. Do not explain “good for small spaces” only in voice-over. Specify the narrow corridor, low furniture, storage location, and movement path. Do not claim “quick setup” without showing the packed state, sequence, fasteners, and final footprint.
This step also creates reuse. A real environment, one complete action, and several product-detail shots can support a hero edit, vertical cutdowns, still frames, and new openings later. A closed master file cannot do the same work.
N — Nail the proof: assign an evidence method
Interest language contains hidden standards. “Suitable,” “easy,” “stable,” and “quiet” are not shot instructions. The brief must say how the idea will be supported: show a structure, record an operation continuously, disclose conditions, or run a compliant comparison. Performance claims without verifiable conditions should not be exaggerated through acting or editing.
For brands connected to Chinese supply chains, this layer also ties creative to product truth and market constraints: model differences, materials, ports, packaging, claim conditions, talent rights, and approved territories. It determines whether the same footage can travel across markets—or whether localization requires a reshoot rather than new subtitles.
S — Series: build an interest cluster, not isolated keywords
Group related expressions into an interest cluster, then define a repeatable format. “Real problem → product action → visible result → conditions” can support episodes about tight spaces, pet hair, night use, and maintenance without reinventing the brand every week.
A series is not the same cut with new titles. Keep the question shape and audiovisual order recognizable, while changing the scene, obstacle, and proof. Production spend then buys an expandable content unit, not a stack of synonyms.
Budget by validated interest cluster, not export count
A multi-cluster project can be scoped in four layers:
- Shared foundation: language research, creative rules, product truth, common environments, and reusable capture.
- Cluster modules: the actions, props, cast, shots, and proof unique to each cluster.
- Channel and market versions: aspect ratio, duration, subtitles, voice, disclaimers, and calls to action.
- Measurement and refresh: version IDs, performance review, missing-proof capture, and retirement of weak expressions.
Cost per validated interest cluster = (allocated shared foundation + cluster production and post + localization + measurement and revision) ÷ clusters that produce a distinct content or business signal.
This is a planning denominator, not an industry price or a claim that the new control lowers cost. Twenty exports with lightly changed copy may still test one idea. Four modules with different situations and proof may answer four buying questions.
Measure without inventing attribution Meta does not provide
Give each interest cluster a stable ID and retain its script, footage, captions, and paid variants. In controlled paid tests, compare attention, completion, click behavior, landing-page actions, and qualified inquiries. In organic distribution, examine saves, shares, comments that repeat the target problem, related site search, product-page visits, and branded search movement.
None of these proves that an individual typed a sentence and then received a specific video. They answer a more defensible question: did this audience expression help the brand create a clearer piece of content and a stronger next action? Marketing should not draw a causal chain that the platform does not expose.
When to change the brief—and when not to chase the signal
LENS is most useful when purchase depends on context, niche interests, professional tasks, or specific anxieties. Outdoor products, smart hardware, home goods, beauty devices, tools, and B2B equipment can serve several jobs with the same product; the brand needs to decide which job deserves production first.
If demand is already highly branded and people search by model number, if differentiation is mostly price and availability, or if the project is a one-off brand statement, improving search, product pages, channel information, and the hero film may be more valuable than building a full audience-language library.
The decision test is simple: if “target audience” does not let a production team independently choose the scene, action, and proof, it is not yet a shootable brief. Rewrite the segment as one sentence the audience would say, then decide whether that sentence deserves a content series.
For a cross-border launch, start with a reusable product-video production system that puts audience language, product truth, observable action, and market versions on the same capture plan. Each production line item should answer a specific customer question—not just add another file to the delivery folder.
