Use casesUGC ads

UGC ads your AI agent reads in full, and launches paused.

Know which creator hook sold, and brief the next creator from it. Your AI agent reads and transcribes every video before spend.

7 days free with every feature. Cancel anytime.

Try asking
  • Which creator hooks bought more than their share of spend?
  • Tell me what each new creator video opens with.
  • Brief three creators on the hook that sold.
A four-frame Reel from Saltmilk, a fictional brand, in a phone, with the agent's work around it: Values tagged (Simple nightly routine, Skincare minimalist), a winning creative, a launch built paused and a prompt waiting for the owner to activate it. Four-frame vertical reel for Saltmilk, a fictional skincare brand, following one woman through a four-step night routine.

Why is UGC hard to learn from?

A folder of creator videos tells you little. The lesson lives in the hook, talent and format, and none is in the file name.

  • “final_v3_FINAL.mp4”

    Creators name files their own way, so nobody can filter by hook or by persona. The ad name carries the label, and the label drifts.

  • “It got views, but did it sell?”

    A video can hold attention and still not move the buyer. The opening, the hold, the click and the conversion need to be read side by side.

  • “We briefed the next creator from a hunch.”

    Briefs get written from whichever ad looked good last week. The creator never hears which hook actually bought more than its share of spend.

Your agent reads the whole video, not a thumbnail.

Drop creator files into Library and each video is queued to be read and transcribed. The reading happens on the video itself, so the hook is the real first line.

  • A video under 18 MB is sent to the model whole. A larger file is read from its poster frame, and the record says which one it was.
  • meta_transcribe_creative writes what is said, with the second each line starts. It runs on every Library video, and your agent can call it to read one now.
  • Spoken words only. On-screen text is not transcribed, and a video with no speech is stored as having none.
A creator video from Saltmilk, a fictional skincare brand, read in full. Its transcript starts, Here is my night routine, in four steps, and the agent files it under the hook Four-step routine, the persona Skincare minimalist and the format Reel.

Hook, talent and format become Values you can compare.

Win & Scale files each video under Values, such as its hook, talent, format, persona and angle. They are read from the creative, so a lesson survives a rename.

  • meta_get_creative returns one creative with its Values, how sure the reading was, and its transcript.
  • meta_tag_creatives corrects many videos at once by Value name. A person’s correction always wins over the model’s reading.
  • meta_get_cross_analysis crosses two Dimensions, such as persona by hook. A missing cell is a combination you never tried.
A Library creative, Saltmilk's Reel (a fictional brand), a video read in full: its Values are the angle Simple nightly routine, the persona Skincare minimalist, the format Reel and the hook Four-step routine reel, with the line the agent read, “Here is my night routine, in four steps.”, and each trait marked Winner, In line or Low delivery, inconclusive.

Find the hooks that bought more than their share.

Your agent ranks each video’s hook and hold against the other ads in the same view, then judges winners on the goal your account buys. Scores are ranks within the view, never benchmarks.

  • meta_get_history_winners names the hooks, talent and formats that bought more than their share of spend.
  • meta_get_report_view returns hook, watch, click and convert scores from 0 to 100. A video under the volume floor gets no score, never a low one.
  • meta_get_breakdowns with retention shows the share of viewers still watching at each second.
The result of meta_get_history_winners on an example account for Saltmilk, a fictional brand, goal purchase. Selfie caption, nightly reset is a winner over 90 days and average over 30. Skin, simplified hero shot is the reverse. Quote card with first-name attribution is unproven. Next test: A fresh nightly-routine selfie for the skincare minimalist, before the next restock ramps up.

Brief the next creator. Launch the batch paused.

Your agent writes the brief from what sold, with a timed script and the deliverables, then builds the launch with nothing spending. Activating is a separate call that waits for you.

  • meta_save_brief writes the concept, hook, timed script lines and deliverables. It quotes only numbers its evidence carries.
  • meta_launch_ads builds the campaign, ad set and ads paused, and joins a 9:16 and a 4:5 cut into one ad.
  • meta_activate is the one call that spends, and your agent asks you first.
An agent calls meta_launch_ads for Saltmilk, a fictional beauty brand, and gets back a paused launch: the campaign, the ad set and three ads, all Paused, with Activate waiting for the owner.

What can you ask your AI agent about UGC?

Ask the way you would ask a media buyer who has watched every video. Your agent checks the evidence before it answers.

  • Evidence checkWere all our creator videos read in full? Don’t change anything.
  • TranscriptsTell me what each new creator video opens with.
  • Which hooksWhich creator hooks bought more than their share of spend over the last 90 days?
  • Where to workAnswered in the conversation: Why isn’t the night-routine reel getting clicks?
  • Brief a creatorBrief three creators on the hook that sold, in our brand voice.
  • Launch pausedLaunch the three approved cuts as one test. Leave it paused.
Claude Code · winandscale connected

Why isn’t the night-routine reel getting clicks?

  1. meta_get_evidence_healthsynced 3h ago
  2. meta_get_report_viewgrouped by creative · 90 days
  3. meta_get_breakdownsretention curve

People watch the reel but don’t act. Its hook and watch scores are high in this view and its click score is low, so the ending is where to work. These are ranks within the view, not benchmarks, and I changed nothing.

Questions, answered

Everything else is in the docs.

Can an AI agent analyze UGC ads?
Yes, with Win & Scale your AI agent reads each UGC video in full, files it under Values such as hook, talent and format, and compares them on the goal your account buys.
Does Win & Scale transcribe UGC videos?
Yes, every video in your Library is transcribed automatically after it is uploaded, launched or imported. The transcript holds spoken words only, with the second each line starts.
What is a hook score?
A hook score is a 0 to 100 rank of a video’s hook rate against the other ads in the same view. It is not a benchmark against other accounts, and a video under the volume floor gets no score.
What if a video is too large to read?
A video over 18 MB is read from its poster frame instead of the whole file, and the record says so. Win & Scale refuses to transcribe a file it cannot read rather than guess.
Can my agent launch UGC ads without asking?
No, Win & Scale creates every campaign, ad set and ad paused. Turning one on is a separate call, meta_activate, and your agent asks you in the conversation first.
How much does Win & Scale cost?
Every plan starts with a 7-day free trial that includes every feature. After it, each ad account pays a flat monthly price set by its ad spend, starting at $49 a month for up to $10k of monthly spend.

Know which creator hook sold, then brief the next one.

Start a 7-day free trial with your website. Every batch your agent launches stays paused until you approve it.