Nutra creatives on Facebook: formats, approaches, and reading competitors
Plenty gets written about nutra creatives and almost none of it carries numbers: which format to shoot, which button to use, how many versions to make. We counted ours, 3,524,945 Facebook nutra ads as of 20 September 2026, and put together what follows from it.
What counts as a creative here
A creative is an image or a video bound to the ad's copy and its call to action. In nutra it carries more weight than in ordinary advertising: nobody went looking for the product, so the first two seconds have to convince someone they have a problem, and the next five that the answer looks plausible.
Every figure below comes from one measurement against our showcase database on 20 September 2026. The vertical is decided by a model reading the creative itself, from video frames and the text on the image, rather than from what the advertiser called their page. Each share states what it is a share of: passing off a share of the labelled portion as a share of the whole corpus is a lie.
Half the market is video
The format is known for all 3,524,945 ads, so these are shares of the full corpus.
| Format | Ads | Share |
|---|---|---|
| Video | 1,906,016 | 54% |
| Static image | 970,334 | 28% |
| Carousel | 648,595 | 18% |
The production takeaway is plain: entering nutra with static images means competing for attention against a moving frame in half of all impressions. That does not make static useless, but it does mean it has to win on the first screen rather than on the reveal.
A fifth of all ads belong to creatives with thirty versions
Near identical variants of one creative collapse into a stack, so every ad carries the number of versions its approach has. That matters more than it looks: nobody outside can see conversion, but version counts are visible, and they show an advertiser's decision to keep paying.
| Versions of the creative | Ads | Share of corpus |
|---|---|---|
| one | 1,315,399 | 37% |
| two to four | 664,499 | 19% |
| five to nine | 382,858 | 11% |
| ten to twenty-nine | 449,240 | 13% |
| thirty or more | 712,949 | 20% |
A single ad can be a hundred dollar test, and 37% of the corpus is exactly that. A stack of thirty versions cannot be a test: somebody found an approach that works and has spent months looking for its best form. If you are taking competitors apart, that 20% is where to start, not the general feed.
Which call to action they use
A call to action is recorded for 3,488,561 ads, 99% of nutra. The shares below are of that portion.
| Call to action | Ads | Share |
|---|---|---|
| Shop now | 1,268,204 | 36% |
| Learn more | 839,986 | 24% |
| WhatsApp message | 264,353 | 8% |
| See details | 229,546 | 7% |
| Order now | 213,616 | 6% |
| Message page | 164,264 | 5% |
| Download | 94,802 | 3% |
| View Instagram profile | 70,235 | 2% |
Two thirds of the market rests on two buttons, and the difference between them is substantive. Shop now promises a purchase right away and belongs where the landing page is a storefront with a price on it. Learn more promises an explanation and pulls cold traffic onto a long advertorial. Putting Shop now in front of an advertorial breaks the promise in the first second after the click.
Messenger funnels deserve their own line: WhatsApp message and Message page together account for 13%. There is no landing page at all in those, the order is assembled in conversation. It is a separate mechanic with its own demands on the creative, and lifting its copy onto a classic funnel achieves nothing.
What language nutra speaks
Language is read from the ad's own text and is known for 2,893,334 ads, 82% of nutra. Shares are of that labelled portion.
| Language | Ads | Share |
|---|---|---|
| English | 1,388,210 | 48% |
| Spanish | 365,814 | 13% |
| Portuguese | 121,129 | 4% |
| German | 110,007 | 4% |
| French | 109,692 | 4% |
| Polish | 105,783 | 4% |
| Italian | 90,445 | 3% |
| Romanian | 70,398 | 2% |
| Arabic | 61,514 | 2% |
| Turkish | 60,963 | 2% |
Language is the only honest signal of a market available from outside. Geography cannot be derived from Ad Library data: what it returns is a delivery region, and what ad spy databases often hold is the geo of the search basket, neither of which equals a campaign's targeting. Spanish sits second by a wide margin, and while half the corpus is English, that is the United States, Britain and the Philippines at once.
Where the creative actually lands
Placements are recorded for 3,524,875 ads. One ad normally runs across several, so the shares add up to more than a hundred percent.
| Placement | Ads | Share |
|---|---|---|
| 3,394,839 | 96% | |
| 3,389,688 | 96% | |
| Messenger | 2,499,083 | 71% |
| Audience Network | 2,469,458 | 70% |
| Threads | 2,248,237 | 64% |
| 570,843 | 16% |
The practical consequence is that a creative is almost never made for one placement. It will run in the Facebook feed, in Instagram stories, and in Audience Network inside somebody else's app with the sound off. Hence the requirement visible in every long-lived creative: the point has to survive muted audio and a vertical frame.
The approaches that come up most
What follows is observation rather than measurement: we cannot classify three and a half million ads by approach, so this is what regular watching of the feed shows. No claims about conversion rates appear here. Nobody outside can see conversion, and anyone quoting you a figure for a competitor's creative invented it.
| Approach | What it looks like | What not to break |
|---|---|---|
| Person to camera | Shot on a phone, an ordinary room, product in hand, first person speech | The moment studio lighting appears, the format loses the trust it was built on |
| Before and after | A split frame or a sequence, the result shown visually | Facebook moderation removes direct promises of results; implication works, assertion does not |
| Problem on screen | The opening seconds show a discomfort recognisable without words | Recognition has to land before the product appears |
| Ingredient breakdown | Components one at a time, text on frame, calm delivery | A long format that depends on retention, so the opening seconds carry everything |
| News frame | Laid out to resemble an article or a report | The riskiest for moderation; impersonating a news outlet gets accounts pulled |
Whether an approach holds can be checked by version count. If the same device shows up across a dozen unrelated advertisers and each of them has a stack behind it, that is a stable form rather than one team's lucky guess.
How to take a competitor's creative apart
Reading a competitor's creative is not "look at it and copy it". A copy made without understanding reproduces the shape and loses the reason the shape worked. An order that makes sense:
- Find the offer, not the ad. One ad tells you nothing; the set of ads behind one product shows what the advertiser tried and where they settled.
- Check how many versions the approach has. One version is a hypothesis, thirty is a conclusion.
- Read the ad copy, not just the image. The copy carries the promise, and it is the copy that has to match the landing page's first screen.
- Look at the call to action and work out the funnel behind it: storefront, advertorial, or conversation.
- Open the landing page. A creative without its landing page is half a funnel, and half a funnel assembles into nothing.
- Compare against neighbours in the category. If five unrelated teams arrived at the same device, that is a property of the market rather than one team's find.
The hard part here is technical. An advertiser switches an ad off and within days neither the video nor the landing page can be found. So what a tool keeps on its own servers matters as much as what it shows you.
What to look at it with
Several tools fit the job, each with its own strength. Honestly, as of September 2026:
| Tool | Strength | Limitation |
|---|---|---|
| Facebook Ad Library | The primary source, free, shows everything currently active | No history, no filtering by offer or format, search only by advertiser and text |
| AdHeart | Large database, familiar interface, strong on CIS markets | Keyword driven search, shallow vertical labelling |
| AdSpy | Deep history and filters, many sources | Expensive, heavy interface, nutra is not broken out |
| BigSpy | Wide platform coverage, cheap entry | Noisy, media frequently unavailable |
| Minea | Good on dropshipping and physical products | Nutra is a side vertical for it |
| Vizor | Built around one vertical: the offer is read from the creative, duplicates collapse, video and landing pages are kept in HD and stay available after the ad comes down | Nutra only, Facebook and Instagram only, no other verticals or sources |
The limitation on our own row is real rather than decorative. If you need native, push or TikTok, we are the wrong tool and AdPlexity or Anstrex are worth looking at instead. We trade breadth for depth inside one vertical on purpose.
Why the exercise often produces nothing
- The frame gets copied instead of the structure. Repeating someone's image is pointless: it has already burned out on the audience it ran against.
- The creative is taken without its landing page. The promise in the ad and the first screen of the landing page are one message, and splitting it costs you the meaning of both.
- A bridge page is mistaken for an offer. Domains carrying hundreds of different advertisers are redirects, not products; the creatives on them belong to different teams and different goods.
- A lone ad gets studied. Thirty-seven percent of the corpus is single versions, and some of them simply failed.
- Language is ignored. A device built on wordplay falls apart in translation, leaving a picture with no reason behind it.
What we deliberately do not count
Two figures you will meet in other articles are absent here, and that is on purpose.
Creative lifespan. Facebook returns the snapshot date as an ad's end date rather than the day it stopped: among stopped ads, the most common end dates coincide with crawler run days. An average lifespan computed from that column measures a data collection schedule, not a market.
Distribution across countries. The Ad Library export carries a delivery region, and ad spy databases often carry the geo of the basket the ad was found in. Neither equals a campaign's targeting, and the sentence "N ads are running in Poland" is not something this data supports.
Both could be computed in a minute. Not printing them is the only way to avoid lying.
Common questions
- Which creative format dominates nutra on Facebook?
- Video. Measured on 20 September 2026 across 3,524,945 nutra ads: video 54%, static image 28%, carousel 18%. The format is known for every ad in the corpus, so these are shares of the full set.
- Which call to action is used most in nutra?
- Shop now, on 36% of ads that record a call to action, and Learn more on 24%. Together those two buttons account for two thirds of the market. Another 13% goes to messenger conversations, where there is no landing page at all and the order is assembled in a chat.
- How many versions of one creative do advertisers usually run?
- The spread is wide: 37% of ads belong to approaches that exist in a single version, while 20% belong to approaches with thirty versions or more. Version count is the only signal visible from outside that an advertiser decided to keep paying.
- What language is most Facebook nutra advertising in?
- English, at 48% of the 2,893,334 ads whose language is known, which is 82% of nutra. Spanish follows at 13%, then Portuguese, German, French and Polish at roughly 4% each.
- Can you tell which countries a creative targets?
- Not from Ad Library data. What is available is a delivery region in the export, and in ad spy databases often the geo of the search basket the ad was found in. Neither is the same as a campaign's targeting settings, which is why the language of the ad itself is the more honest signal.