You paste your brand website (or a description). We surface every adjacent creator format on social media that transfers to your brand — with the mechanic that makes it work, the specific way you'd adapt it, and a comment-grounded proof of why viewers actually engage.

How is this different from a general chatbot?

General chatbots guess. We ground every answer in real scraped data — 16 parallel social searches per run, ~500 viral videos filtered down to the patterns with 2+ creators actually running them, then up to 50 real comments per top video to prove the audience behavior. Our intelligence system writes the report on top of a live retrieval index, not the other way around.

How long does a Discover run take?

About 90 seconds cold. Cached 24 hours per brand so returning to the same analysis is instant. Every card has a live progress rail showing what's happening — brainstorming phrases, scanning social feeds, reading comments — so you know when to check back.

What data sources power it?

Realtime scraping of social platforms — search + creator profiles + comments — plus a proprietary corpus of viral hooks and transcripts that grows daily. Historical trend trajectories are stored, so you get context a competitor spinning up today literally cannot.

Can I share the analysis with my team?

Yes. Every Discover run gets a public read-only share URL — /discover/share/ — with a custom OG image so it previews cleanly in Slack / DMs. Recipients don't need an account to view.

Does it also do competitor deep-dives?

Yes. The Competitor deep-dive surfaces direct rivals + structural analogues based on audience-overlap, pulls their organic, paid (Meta Ad Library), and creator-led content into one feed, and ranks by real overlap. Same brand paste, richer output.

What powers the reasoning?

A layered intelligence system — a strategic reasoning engine for pattern classification, transfer writing, and conviction analysis, plus a high-throughput classifier for brand extraction and query brainstorming. Neither layer is our moat — the corpus, retrieval, and pipeline are.

What's the free plan?

Free tier includes 5 credits, one-time, no card required — enough to run one Discover on your brand and try a Post analysis. Paid plans start at $19/month for solo founders.

Is this just an AI wrapper?

No. The actual product is a continuous scraping pipeline, a corpus of 500k+ short-form videos with paired engagement outcomes, a retrieval layer, and a set of purpose-built analysis pipelines that ground every insight in real scraped data. Any single AI model is just the last-mile writer — swap it out tomorrow and the product still works.

Everyhookyourrivalsareshipping.

Paste your website. We fire parallel social searches, filter to patterns multiple creators are actually running, then ground every insight in real comments from top videos. One card per pattern — mechanic, transfer, hook to steal.

start here

How it works

From a website paste to a full pattern library in 90 seconds.

Six deterministic steps. Every claim on the report traces back to a scraped video URL, a comment thread, or a live retrieval query. No prompt magic — just a grounded pipeline.

Try Discover free

Paste your brand

Website URL or short description. We crawl the page, extract positioning, and use it to steer every downstream call so results stay on your brand's terms.

AI brainstorms search phrases

Our intelligence system generates 16 trend-hook phrases specific to your mechanic and category — the exact things creators are riffing on right now, not generic keywords.

Parallel social search

We fire every phrase across social platforms in parallel and pull ~500 candidate videos. Then we group them by phrase, dedupe, and keep only patterns with 3+ hits across 2+ creators.

Pattern classification

One card per qualifying pattern — with the mechanic (what the format IS), the transfer (specific version your brand can shoot), and a hook line to say on camera.

Read the audience

Up to 50 real comments per top video, analyzed to prove why viewers engage — tagging friends, disbelief, replay behavior — then tied back to your brand's specific transfer.

Inspiration library

10-20 patterns per brand, uncapped. Each card carries a carousel of 3-5 videos across creators, the analysis, and a share link so your team can review async.

What powers the pipeline

Four systems, one grounded report.

The reasoning layer is the last-mile writer. The moat is the corpus + retrieval + scraping infrastructure beneath it — the parts a wrapper competitor can't rebuild without also spending 6-12 months on infra.

See a real Discover run

Reasoning engineStrategy

Pattern classification, brand-specific transfer writing, and comment-grounded conviction analysis. Called once per Discover run, batched for speed and cache reuse.

Classifier layerVolume

High-throughput classification for brand extraction, query brainstorming, and comment intent tagging. Cheap enough to run continuously across the whole corpus.

Realtime ingestScrape

Live scrape across social platforms — 16 parallel searches per Discover run, plus creator profiles and up to 50 comments per hero video. Feeds the retrieval index.

HookDB corpusData

Proprietary short-form corpus — 500k+ videos with paired engagement outcomes, indexed by hook, mechanic, and format signature. Grows daily; retrospective visibility no competitor can rebuild.

Under the hood

Not another AI wrapper. A grounded pipeline.

Three specialized agents, each fed by live scrape data and a growing historical corpus. The reasoning layer writes on top; the corpus, retrieval index, and comment mining are the parts a wrapper competitor can't ship in a weekend.

Run it on your brand

Brand
Seed phrases
Social search
Pattern class.
Comment mining
Library

Seed-first search16 parallel queries

The intelligence system guesses phrases; the scraper delivers reality.

Instead of asking an AI what's viral, we ask it for 16 trend-hook phrases specific to your brand's mechanic — then fire them across social platforms in parallel. Each phrase pulls ~30 videos. We keep only the ones where 2+ different creators are already running the pattern, so every card on the report is grounded in real repeatability.

seed_brainstormparallel_searchcreator_dedupeviral_floor

Seeds are re-brainstormed per brand — no generic keyword lists. Search results are viewer-count-filtered to skip one-hit wonders that won't transfer.

Pattern classificationOne card per pattern

Every card names a repeatable format, not a single creator.

For each qualifying seed, the intelligence system writes a 4-part card: mechanic (what the format IS across creators), transfer (specific brand version), hook line (verbatim spoken opening), and rank. Duplicate patterns get merged by URL-overlap + name-similarity dedup, so the library reads as distinct formats.

pattern_namebrand_transferhook_lineclose_dedup

No hardcoded cap — you get as many cards as have real transfers. Typical Discover run returns 10-20.

Comment-grounded convictionup to 50 comments/card

Why viewers engage — from real comments, not vibes.

See comparison

Built for how content teams actually research

Skip the manual scrolling. Ship a research doc in 90 seconds.

Run Discover on your brand

~90s

Cold-run Discover, cached 24h

500k+

Short-form videos indexed

10-20

Grounded patterns per brand

What people are shipping

Content teams who switched.

Founders, growth marketers, and content agencies use ugcs.farm to research patterns and analyze competitors. Here's what they say after using it in their weekly workflow.

Try Discover free

We replaced our whole Monday creative-research call with this. I paste the Shopify URL, get 15+ patterns with real social examples in 90 seconds, and drop the share link in Slack before standup. Our best-performing short-form ad this quarter came straight out of it.

MWMara WhitfieldFounder, Lumea Skincare

The comment-grounded conviction is what sold me. Other tools tell me 'this format works.' This one tells me viewers are tagging their partners with '@name bet' — with an actual sample size shown. That's the difference between vibes and evidence.

DPDevon ParkPerformance Marketer, Northbeam

Honestly didn't expect this level of support. Slack ping about a competitor deep-dive at 11pm and Archisman, one of the founders, was on it. Product is great, but that made me stay.

PNPriya NairOwner, Halftone Studio

Solo creator here, I used to lose four hours a week scrolling social feeds for reference clips. Now I paste my landing page and get every format that transfers to my niche with the hook line already written. Freed up a whole afternoon.

JOJamie OkaforIndependent Creator

The share link is quietly the killer feature. I run Discover on a client's brand, drop the /discover/share/ URL in their weekly report, they see 15 patterns with videos and conviction reads. Sold two retainers on that alone.

SRSofia ReyesHead of Growth, Tend SaaS

Ran it on a supplement launch last week. Fifteen patterns surfaced, three we hadn't considered, one obvious steal we've already shot. What used to be a two-week competitive audit is now a Tuesday-morning coffee.

CBCaleb BrennanEcommerce Lead, Forma Goods

FAQ

Questions, answered.

Still curious? Start free

What does it do?: You paste your brand website (or a description). We surface every adjacent creator format on social media that transfers to your brand — with the mechanic that makes it work, the specific way you'd adapt it, and a comment-grounded proof of why viewers actually engage.
How is this different from a general chatbot?: General chatbots guess. We ground every answer in real scraped data — 16 parallel social searches per run, ~500 viral videos filtered down to the patterns with 2+ creators actually running them, then up to 50 real comments per top video to prove the audience behavior. Our intelligence system writes the report on top of a live retrieval index, not the other way around.
How long does a Discover run take?: About 90 seconds cold. Cached 24 hours per brand so returning to the same analysis is instant. Every card has a live progress rail showing what's happening — brainstorming phrases, scanning social feeds, reading comments — so you know when to check back.
What data sources power it?: Realtime scraping of social platforms — search + creator profiles + comments — plus a proprietary corpus of viral hooks and transcripts that grows daily. Historical trend trajectories are stored, so you get context a competitor spinning up today literally cannot.
Can I share the analysis with my team?: Yes. Every Discover run gets a public read-only share URL — /discover/share/<brand> — with a custom OG image so it previews cleanly in Slack / DMs. Recipients don't need an account to view.
Does it also do competitor deep-dives?: Yes. The Competitor deep-dive surfaces direct rivals + structural analogues based on audience-overlap, pulls their organic, paid (Meta Ad Library), and creator-led content into one feed, and ranks by real overlap. Same brand paste, richer output.
What powers the reasoning?: A layered intelligence system — a strategic reasoning engine for pattern classification, transfer writing, and conviction analysis, plus a high-throughput classifier for brand extraction and query brainstorming. Neither layer is our moat — the corpus, retrieval, and pipeline are.
What's the free plan?: Free tier includes 5 credits, one-time, no card required — enough to run one Discover on your brand and try a Post analysis. Paid plans start at $19/month for solo founders.
Is this just an AI wrapper?: No. The actual product is a continuous scraping pipeline, a corpus of 500k+ short-form videos with paired engagement outcomes, a retrieval layer, and a set of purpose-built analysis pipelines that ground every insight in real scraped data. Any single AI model is just the last-mile writer — swap it out tomorrow and the product still works.

Get started

Know who's actually fighting for your buyer.

One paste of your website surfaces the direct rivals, the structural analogues you can steal from, and every viral post they've run in the last 30 days — all ranked by real audience overlap.

Open Discover Talk to founders