Five formats. One of them is already sitting in the vault.
Ranked by ceiling, not by fashion. Each one carries the specs to build it, the reason the category rewards it, and the fence that an investigational device draws around it.
3 video types
2 carousel/static types
7-slide standard
All 4 platforms
The five
Open any row. The verdict on the right is what the format is best at, not how good it is.
Video against carousel is the wrong argument. In 2026 the two formats are solving different problems: video wins reach because discovery runs through watch time; carousels win saves and dwell because completion is work the reader has to do. A programme needs both, in roughly a 60/40 split by volume.
This is the format that produced the highest-performing brand-owned post we found in the entire aesthetics injectable category: Galderma’s travel through epidermis, dermis and hypodermis, at 56.8 engagements per thousand followers — 4.7× the next-best brand post on the platform.
It works because it resolves a question the audience genuinely cannot answer from a before-and-after: where does the material go, and what does it do there? A face answers “did it work.” A cross-section answers “what is it.” Only one of those two questions is still open in this category.
For Conexeu it is also the only high-ceiling format that is fully compliant. An animation of designed architecture is a description of the device, not a claim about outcomes — provided the voice-over and on-screen text stay in the language of design and intent.
Build it like this
Open on the anatomy, not the product. First frame is tissue, second frame is the depletion. The viewer has to feel the absence before the scaffold arrives.
One camera move, continuous. Cutting between angles reads as a montage; a single descending push reads as a journey. Galderma’s is one move.
Label the layers on screen — epidermis, dermis, hypodermis. On-screen text is what makes the clip findable in in-app search and what an AI engine reads off the frame.
Show the scaffold as a lattice, not a fluid. The whole brand argument is that CXU™ is a structure. If it animates like a gel, the film argues against the positioning.
End on the host cells populating the lattice, then hold two beats on the empty lattice resolving into tissue. That final beat is the loop point.
Cut three lengths from one render: 12s (Reels hook), 22s (the full move), 45s (LinkedIn, with the CSO narrating). Master the 12s cut 9:16 — it is the one that travels if the TikTok question in Chapter 06 is ever answered yes.
Compliance rails
No timeline, no percentage, no “results” frame, no patient face.
Voice-over verbs stay in the design register — designed to, intended to, supports. Never restores, regenerates as an achieved outcome, proven, or works.
Investigational-device line burned into the final frame and repeated in the caption’s first 125 characters, not the last.
Animations depicting mechanism are promotional material. They go through legal review before render lock, not after.
Aspect
9:16 & 1:1
Runtime
12–22s
Hook window
First 3s
Caption
125 char, front-loaded
Audio
Score, no VO on 9:16
Cost
Highest
Carousels beat Reels on engagement in Q2 2026 — 0.50% against 0.48% — and beat them roughly 2× on saves. On LinkedIn the gap is decisive: native document posts run about 7.00% engagement, the highest of any format on the platform, against 6.00% for video.
The reason is mechanical, not aesthetic. Both platforms now rank held attention: LinkedIn calls it dwell time, Instagram calls it completion rate. A carousel is the only format where the reader has to physically work to reach the end, and 2026 data puts posts at 7–10 slides at roughly 3.5× the dwell time of a single image.
Seven is the number to build to. Below five there is not enough depth to earn a save; above twelve, completion drops off around 40%. Seven also happens to be the exact length of the argument Conexeu needs to make.
Build it like this
Slide one is a number, not a title. The hook is the statistic; the topic is the caption’s job.
One idea per slide, and the idea has to advance. If a slide restates the previous slide in new words, delete it — filler slides are what kill saves.
1080×1350. Never 1:1 for a carousel: the 4:5 frame takes ~30% more feed height and that is free dwell time.
Put the swipe cue on slides one through six and remove it on seven. The absence of the cue is how a reader knows they finished, and finishing is what gets counted.
Slide seven carries the ask and the caution line together. Never the caution line alone on its own slide — it reads as a disclaimer being hidden.
Export the same seven frames as a PDF for LinkedIn. It becomes a native document post, which is the single highest-engagement format on that platform.
Compliance rails
Every number on every slide needs a citation in the caption. Answer engines and clinicians both check.
No comparative claim against a named competitor product on a carousel slide. Category-level mechanism comparisons (“fillers occupy space; scaffolds are populated”) are fine; “versus Sculptra” is not.
Preclinical stays labelled preclinical, on the slide that shows the data, not just in the footer.
Slides
7
Canvas
1080×1350
Dwell vs. image
~3.5×
Save-rate vs. Reel
~2×
LinkedIn twin
PDF document
Cost
Low
Two of the three highest-engagement LinkedIn posts we read in this category came from individual people, not brands. Galderma’s chair writing in the first person about a symposium pulled 225 engagements; a Wyss Institute scientist introducing her own breast-reconstruction programme pulled 190 with a video. Both outperformed most of what their organisations posted.
Conexeu’s own data says the same thing in miniature. Its top three posts by rate are all video with a named person in them: the Nasdaq film, the CEO on Stocks to Watch, the CSO on a podcast. Its two worst are announcements with no face.
This is also the cheapest format in the set, and the one that scales best: Dr Chavez-Munoz and Dr Pilcher can each answer one question a week on a phone camera with no crew.
Build it like this
The question is the title, said out loud in the first two seconds. “Why does a scaffold need to be a structure and not a gel?” — then answer it.
Shoot in the lab or at the bench, not against a brand wall. Location is credential.
One question per clip. The instinct to cover three is what turns a 40-second answer into a two-minute lecture nobody finishes.
Burned-in captions, always. Reels are watched muted, and the caption text is what in-app search indexes.
Post it from the scientist’s own profile first, then have the company page comment — not reshare. Thought-leader distribution runs roughly 1.6× the engagement rate of the same content on a brand page.
Compliance rails
Scripted and reviewed, even though it looks unscripted. An off-the-cuff efficacy verb from a CSO is a regulatory event, not a blooper.
No answering treatment questions in the comments. Route every clinical question to a single reviewed reply and a link.
Nothing material and non-public. This is a Nasdaq issuer; social is a disclosure surface.
Runtime
25–45s
Aspect
9:16, 1:1 crop
Captions
Burned in
Posted from
Personal profile
Uplift vs. page
~1.6×
Cost
Lowest
The best-performing independent post in our whole Instagram read set was a reframe: “bio-stimulators aren’t about INSTANT volume, they’re about teaching your skin to act young again” — 550 engagements from an injector with no brand budget, beating two brand accounts of 31k and 40.9k followers on the same topic.
Reframes travel because they hand the viewer a new sorting rule for information they already have. That is the most shareable thing on social: not a fact, but a way to hold the facts.
Conexeu has the strongest available reframe in the category and is barely using it. The market currently sorts injectables into two bins — fillers (occupy space) and biostimulators (provoke collagen). CXU™ is neither: it is a structure the body populates. A third bin is a much bigger idea than a better product.
Build it like this
State the received wisdom first, in the audience’s own words. You cannot reframe a belief you have not first said out loud.
Give the third category a name and use the same name every time. “Structural” or “architecture” — pick one and never drift.
Three columns, not two. Two columns is a comparison and invites a competitor rebuttal; three columns is a taxonomy and invites adoption.
Works as either a Reel or a three-slide carousel. Test both; the carousel usually wins saves and the Reel usually wins reach.
Reuse the exact phrasing in the site’s H2s. A reframe that only exists on social cannot be cited by an answer engine.
Compliance rails
Describe mechanism classes, never named brands. “Hyaluronic-acid fillers occupy space” is a mechanism statement; “unlike Juvéderm” is comparative advertising for an uncleared device.
Do not imply the third category is better. Imply it is different, and let the reader draw the conclusion. Superiority claims need data Conexeu does not yet have.
This is the format with the most fair-balance exposure in the set: a mechanism comparison reads as a benefit claim the moment a product is attached to one of the columns. Keep all three unbranded.
Formats
Reel or 3-slide
Runtime
18–30s
Columns
Three, never two
Reuse
Site H2s verbatim
Best metric
Shares
Cost
Low
Conexeu already owns this asset. The Oracle vault holds 31 macro product photographs from the NYC shoot and 22 from the North Carolina shoot — CXU™ drops, pokes, straight pulls and tong tests, several of them framed against comparator materials.
Material-behaviour content is the oldest reliable engine on short-form video: it holds attention with no hook, no face and no claim, because the viewer is waiting to see what the substance does. It is also the only format in this set that can ship next week with zero new production.
And it argues the positioning without saying anything. A material that holds a structure on a needle tip looks like architecture. That is the whole message, delivered pre-verbally.
Build it like this
Shoot or cut at 120fps and play at 30. Slow is the entire effect.
Silence plus one texture sound. No music bed, no voice-over. Sound design is what makes these rewatchable.
One behaviour per clip: the drop, the pull, the recovery. Anthologising them into one 60-second edit destroys the loop.
Loop it seamlessly — last frame matches first. A clean loop doubles measured watch time without a single extra second of footage.
Caption carries the entire information load, because the film carries none. This is where the mechanism sentence and the caution line live.
The comparator frames are the highest-value and highest-risk assets in the vault. Legal first, always.
Compliance rails
Comparator shots showing CXU™ beside a named competitor product are comparative promotion of an investigational device. Do not publish any of them without written legal sign-off, and expect the answer to be no until there is data behind it.
Third-party marks require the acknowledgment and non-affiliation line already running in the site footer.
A material behaving well on a bench is not a performance claim. Never caption it as one.
Runtime
6–12s
Capture
120fps → 30fps
Audio
Texture only
Loop
Seamless
Assets on hand
53 stills
Cost
Near zero
Format 05 is not a proposal. These are already in the vault.
The pierceA needle entering the bead and the bead holding its dome. This is the whole architecture argument in one frame, and it needs no voice-over.The drawContinuous extrusion. Slow it to 30fps from 120 and the loop point is the moment it stops stretching.The dropDiscrete droplet formation. The highest-retention shape in the set because the viewer waits for the release.Comparator — do not publishThe vault also holds comparator frames shot against named competitor products. These are comparative promotion of an investigational device. Legal sign-off first, and expect no.
Four of the 53 macro photographs in Oracle Brain/Brand Assets/Imagery/, downsampled here. Nothing in this set required a claim, a face or a render — which is exactly why it can ship in phase two while the mechanism film is still in storyboard.
Ceiling figures come from the live read in Chapter 03. Format-level benchmarks — carousel against Reel engagement, LinkedIn document rates, slide-count drop-off, retention checkpoints — come from published 2026 datasets (Socialinsider Q2 2026, Buffer’s 52m-post analysis, Social Media Today on document posts, and 2026 carousel completion studies). Where sources disagree, the disagreement is stated rather than averaged.
The seven slides
What a seven-slide proof carousel actually contains
Format 02, built out. This is the highest-value carousel Conexeu can post right now, because the demand data is public, the argument is compliant, and no competitor is making it. Click through the slides.
Carousel — “GLP-1 volume loss is a structural problem being treated as a cosmetic one” · 1080×1350 · 7 slides
Hook — the number
8 in 10 providers treat GLP-1 volume loss with a gel.
The tissue that was lost was not a gel.
Swipe
Slide 1 — Hook — the number The hook is a statistic, not a topic. Allergan Aesthetics’ own 2026 data identified HA fillers as the top non-surgical modality used by 81% of providers treating GLP-1 patients. Leading with someone else’s number is what makes slide one credible before the brand has any.
Stakes — the size of it
67% of facial plastic surgeons are seeing more of these patients.
A 45% rise against 2024.
Facial fat grafting up 50% — second consecutive year.
63% of them had never had a cosmetic treatment before.
Swipe
Slide 2 — Stakes — the size of it Three numbers, one direction. The third is the strategically important one: this is a new audience with no brand loyalty and no vocabulary yet. Whoever teaches them the vocabulary owns the category they end up in.
The mismatch
A filler occupies a space. It does not rebuild what made the space.
Hyaluronic acid restores contour by taking up volume.
That is the correct tool for a fold.
It is a different job from replacing depleted architecture.
Swipe
Slide 3 — The mismatch Mechanism-level, never brand-level. This slide has to be legally boring: it describes what a class of material does, which is uncontroversial and publicly documented, and stops there.
The mechanism
What GLP-1 patients lose is subcutaneous adipose architecture.
Rapid depletion of subcutaneous fat, with skin remodelling that cannot keep pace.
The structure goes first. The surface follows.
Swipe
Slide 4 — The mechanism This is the pivot slide and the one to spend design time on — a cross-section still lifted from the mechanism film in Format 01. One asset, two formats.
The taxonomy
Three jobs, not two.
Occupy — HA fillers take up space now.
Provoke — biostimulators trigger the body to make collagen.
Structure — a scaffold gives migrating cells somewhere to go.
Swipe
Slide 5 — The taxonomy Format 04’s reframe, deployed inside the carousel. Three columns, never two. The third column is the one Conexeu is naming, and naming it is worth more than winning an argument inside the existing two.
The work
A 12-month multi-tissue preclinical programme in aesthetics met its pre-defined objectives.
CXU™ is a collagen-based ECM scaffold designed to provide a temporary structural framework that supports tissue regeneration.
Preclinical. Not predictive of clinical outcomes.
Swipe
Slide 6 — The work The only slide that mentions Conexeu, and it arrives sixth. The caution language is in the slide, in the same type size as the claim — that is what separates disclosure from a disclaimer.
The ask
The full preclinical summary, and where it goes next.
Comment ARCHITECTURE and we will send the summary.
CXU™ is a device candidate in preclinical development and is not cleared or approved for marketing in the United States or any other jurisdiction.
End
Slide 7 — The ask No swipe cue on the last slide — its absence is how completion registers. A comment trigger outperforms a link because it keeps the engagement on-platform, and it builds a first-party list that no algorithm change can take away.
Why this carousel and not a product carousel
Six of the seven slides are about the reader’s problem and someone else’s data. One is about Conexeu. That ratio is the reason it will be saved rather than scrolled — and every channel in this document now penalises the teaser-plus-link pattern that a product carousel would use.
It is also the highest-value AEO asset in the plan. Slides three, four and five are the passages an answer engine can lift when someone asks an AI “what causes Ozempic face and how is it treated?” — provided the same wording also exists as text on conexeu.com. Chapter 07 closes that loop.
What not to build
Five formats the search returned that Conexeu specifically cannot use
These dominate aesthetics social and every one of them is closed to a preclinical device candidate. Naming them now prevents the quarterly argument about why the plan looks different from what competitors post.
Before-and-after. The category’s default asset and the highest-engagement format for practices. It is an efficacy claim in image form for a device that has none established. Closed, without exception, until clearance.
Patient testimonial and POV. Revanesse’s “POV: you just had your Revanesse moment” is a compliant post for a cleared HA filler. The same post for CXU™ is unlawful promotion. It is also, notably, their worst-performing format at 2.2 per thousand — so nothing is being given up.
Trade copy aimed at injectors. “Share the legacy with your patients” earned Merz 80 engagements on 40,900 followers. Practitioner recruitment is a sales-team job, not a feed job, and the feed punishes it.
The static science card. A single frame with the mechanism in the caption: 6 engagements for Sinclair, 7 for an independent injector. The format is dead on both platforms in 2026 — if it is worth saying, it is worth animating or splitting into slides.
The brand-hashtag play. We checked the live #renuva tag on TikTok: it is occupied almost entirely by unrelated dropshipping ads, a Portuguese paper brand, and NuvaRing content. Brand hashtags in this category are not defensible. Build on problem language instead — volume loss, hollow cheeks, fat grafting, GLP-1.
Observation on the #renuva hashtag was made on the live TikTok tag page on 2026-08-20. TikTok gates per-video metrics behind a login, so no engagement figures for that platform appear anywhere in this document.
