How winning Meta ads are built, script by script
Published 13 Nov 2026. Data read scripts and sections as stored on 2026-11-13; figures computed 2026-11-13 05:22 UTC (corrected from the 05:06 version, which counted copied scripts once per ad). Source: Meta's Ad Library (EU transparency data), read by Poseidon, unless the text names another source.
Almost every winning e-commerce ad opens on a hook: 99.6% of 2,011 unique scripts read on 13 November 2026, from 503 shops. The median script runs 10 sentences across 5 different sections. After the hook, the product appears in about 7 scripts in 10 and a call to action in about two thirds; a named problem appears in fewer than a third.
Poseidon transcribes the spoken script of winning ads and labels each sentence with its role: hook, problem, product, proof, offer and so on. About 97% of sentences are labelled right on Poseidon's hand-checked sample. Brands often run the same script on many ads, so each script is counted once here: the 3,303 winning ads with a sectioned script carry 2,011 different scripts.
Which sections winning scripts use
2,011 unique scripts from winning ads (e-commerce brands, still running after 21 days or more):
| Section | What it is | Scripts that use it | Share | Median length when used |
|---|---|---|---|---|
| Hook | The opening line | 2,003 | 99.6% | 1 sentence |
| Product | The product named and shown | 1,440 | 71.6% | 2 sentences |
| Call to action | What to do now | 1,332 | 66.2% | 1 sentence |
| Solution mechanism | How it works | 1,262 | 62.8% | 3 sentences |
| Offer | Price, discount, bundle | 1,019 | 50.7% | 2 sentences |
| Proof | Results, reviews, numbers | 1,008 | 50.1% | 3 sentences |
| Problem mechanism | Why the problem happens | 667 | 33.2% | 5 sentences |
| Problem | The problem named | 617 | 30.7% | 3 sentences |
| Projection | Life after the product | 496 | 24.7% | 1 sentence |
| Agitation | The problem made worse | 496 | 24.7% | 5 sentences |
| Discovery | How the speaker found it | 460 | 22.9% | 2 sentences |
| Objection | A doubt answered | 412 | 20.5% | 3 sentences |
- The hook is short and always first. One sentence at the median, in first position in 99.5% of scripts.
- Explaining the problem takes the most room. Why the problem happens and the problem made worse each run a median 5 sentences.
- The offer and the call to action are brief: 2 and 1 sentences.
- How long in seconds? 280 of the scripts have a timed transcript; their median length is 79 seconds. The others are counted in sentences only.
By niche and country
A market is shown when it has 30 or more unique scripts from at least 5 shops, no shop above 40% of them, and the three biggest shops at 90% at most. On 13 November, one market passes:
| Niche | Country | Unique scripts | Shops | Median script length | Different sections, median | Proof | Problem mechanism | Offer | Call to action |
|---|---|---|---|---|---|---|---|---|---|
| Menopause | France | 50 | 10 | 44 sentences | 8.5 | 92.0% | 74.0% | 82.0% | 96.0% |
Winning menopause scripts in France run long (44 sentences at the median, against 10 overall) and almost all of them carry proof, an offer and a call to action.
Not shown yet, under the bar once copied scripts count once: weight loss in France (35 unique scripts, but one shop has 17 of them), bloating in Italy (33, one shop 15), bloating in France (29), and every other niche page (under 30).
What this can't tell you
- Why a structure wins. These are the scripts of ads brands kept running, not a test of one structure against another.
- Seconds for most scripts. Only 280 have timings; for the rest, length is in sentences.
- Ads without speech. Only ads with a spoken script are cut into sections.
- Most markets. Most niche and country markets don't have 30 different scripts from enough shops yet.
How this was measured
- Source: each ad's spoken script, transcribed by Poseidon and cut into labelled sentences; section labels checked by hand on a sample (about 97% right, per Poseidon's data standard). Figures computed on 13 November 2026 at 05:22 UTC.
- Which ads: e-commerce brands that passed Poseidon's shop check, still active and live 21 days or more.
- One script, one count: two ads whose transcripts are the same text count once.
- Markets: the free niche pages' own ads, from shops checked by hand, whose words name the niche.
- Minimum sample: no figure drawn from fewer than 30 scripts.
- Correction, 13 November 05:30 UTC: the first version counted a script once per ad that carried it, so brands that copy one script to many ads weighed too much. It showed 7 markets; once each script counts once, 1 passes.
FAQ
What's the best script structure?
This page shows what winners use, not what works best for you. Ask Poseidon for the scripts of the winners in your niche and read them in full.
How long should my hook be?
In winning scripts, one sentence at the median.
Can I see a script cut into sections?
Yes. Every ad with a script in Poseidon comes back with each sentence labelled. Ask "show me the script of the longest-running weight-loss video in France".
Ask Poseidon how winners in your niche are built. askposeidon.app