- 01Stack your milestone with earned media
- 02Run a standing public session in your niche
- 03Get your integration partners to publish about you
- 04Soft-launch a beta inside a niche technical community
01 Snapshot
The one-glance read on who they are and how they grow. Each point is verifiable from the receipts above.
Multimodal video-understanding APIs (Marengo for search, Pegasus for video-to-text reasoning) for enterprises in media, sports, and ad-tech, fresh off a $100M Series B closed July 1, 2026.
Paid-led: three concurrent ad libraries (Google, Meta, LinkedIn) running alongside an AWS cloud-distribution deal, not a founder-audience or community motion.
~124,752 monthly visits, 42 Google ads (33 active as of 2026-07-31, one proven winner running 363 straight days), and a roughly 321-ad LinkedIn library.
As of 2026-07-31, new Google, Meta, and LinkedIn creative pitches TwelveLabs as a QA layer that catches morphing faces and broken physics in AI-generated video, a shift beyond straight video search.
Growth runs through infrastructure and paid distribution, AWS became TwelveLabs' preferred cloud partner the same week as the Series B, while owned organic search stays almost entirely branded-navigational rather than content-driven.
02 How it unfolded
The order the channels came online. Sequence is strategy: what they did first, and what they layered on once demand existed.
03 Traffic & channel mix
Estimated demand, the channel split behind it, and the keywords and referrers doing the work. Directional modeling, not audited analytics.
TwelveLabs draws an estimated ~124,752 monthly visits across all channels, down 21.2%. This pull does not include a channel-by-channel percentage breakdown, but the available data points to branded search dominance: organic pull concentrates on the company's own name rather than category terms. No referrer or competitor-domain data is available in this pull.
04 Organic content
The specific pages earning their organic search traffic, and the pattern behind why they rank. Adapt the format, not the topic.
TwelveLabs' organic traffic is overwhelmingly branded and navigational rather than content-marketing-driven: the homepage is by far the largest organic earner, largely people searching the company name directly. Product, careers, and about-us pages round out the top tracked pages, more a byproduct of funding-news curiosity than a deliberate SEO program: there's no comparison content, how-to guide, or programmatic page among the top organic earners.
Top pages by organic traffic
05 Channel map
Not traffic share. How much weight the growth system actually puts on each channel, with a one-line read on the role it plays.
What drove growth
06 The motions doing the work
For founder-led SaaS the breakdown shifts from ads to traction: where the first users came from, how the founder grows it in the open, and the compounding organic surface.
TwelveLabs shows no single breakout launch moment; its traction built instead through a string of funding-driven press cycles and modest technical-community mentions rather than one coordinated debut.
In 2022 founder and CEO Jae Lee leaned on earned media, a VentureBeat feature and a Techstars "#GiveFirst" podcast slot in the spring, ahead of a seed-funding announcement that December, using earned media to reach investors and builders rather than paying for reach.
No Product Hunt listing appears in the record, and the eight Hacker News submissions since 2023 read as technical write-ups rather than a coordinated Show HN debut; the closest thing to a discussion spike is a 28-point 2023 post on multimodal foundation models.
Dig deeper+1▾
Each major round, the 2024 Series A, the 2024 strategic round, and the July 2026 Series B, functioned as its own relaunch moment, generating a fresh company blog post and video announcement rather than building toward one defining breakthrough.
TwelveLabs' founder presence never became a sustained public-building channel; its personal social footprint peaked at the seed stage and has stayed largely dormant since.
Founder and CEO Jae Lee holds 10,241 LinkedIn followers versus just 261 on X, making LinkedIn the larger available audience even though the evidence shows no specific LinkedIn post content to judge his posting cadence there.
Lee's X account shows only five tracked posts between 2022 and 2024, spanning a VentureBeat mention, a Techstars podcast shoutout, the seed announcement, a TechCrunch clip, and an April 2024 note on the Pegasus-1 technical report, with no visible activity since.
Dig deeper+1▾
The one Reddit appearance tied to the company is a TwelveLabs team-account post announcing the Jockey research preview, a one-off rather than a recurring build-in-public subreddit practice.
TwelveLabs runs a thin, branded-navigational organic system underneath a genuinely partner-driven distribution layer, not a content-marketing flywheel.
The site's blog is active, but nothing in its top organic pages points to a deliberate comparison, alternative-to, or how-to content program, so the SEO motion reads as maintenance around the product and company rather than demand generation.
The homepage's affiliate/partner program link, paired with joint webinars alongside video-editing platform Blackbird and AWS solutions integrator TrackIt, suggests TwelveLabs recruits technology and systems-integration partners rather than individual content affiliates, so each partner's own case study or webinar becomes a discovery surface TwelveLabs doesn't have to build itself.
Dig deeper+1▾
The Google, Meta, and LinkedIn ad libraries remain the dominant acquisition lever, and the newest LinkedIn creative repositioning the product as an AI-video-generation QA layer signals TwelveLabs is chasing new demand, since its Google spend already covers its own branded terms directly.
07 Paid ads
This founder-led SaaS also runs paid acquisition. Here are the live ads doing the work, each with the X-ray and a play you can adapt.
- Running
- 38 days (longevity = it's converting)
- Hook
- "Massive" appears on a light purple background with blurred white circles.
- CTA / offer
- Try Marengo in our playground or book your demo today.
- Audience
- Teams dealing with large video archives, indicated by "Massive video archives slow teams down" and "teams get back to creating".
Why it works. Uses one bare word as a cold open before revealing the video archive problem, aimed at teams sitting on large unsearchable libraries.
Problem-agitation hook: "Massive video archives slow teams down." (0:03-0:05) Open your video ad with one bold word naming scale or pain, and hold the explanation for the next few seconds.
- Running
- 38 days (longevity = it's converting)
- Hook
- "Three years ago" - A man with dark hair and a white shirt under a dark jacket speaks directly to the camera.
- Audience
- The technical language and product demonstration of video Q&A models and metadata extraction indicate the target audience is developers, data scientists, or businesses looking for advanced video analysis solutions.
Why it works. Anchors the Pegasus 1.5 launch to a founder-style before and after narrative, building credibility with customers who remember the earlier version.
Problem-solution framing with historical context (0:00-0:09, 0:13-0:19): The speaker explains the evolution of their product, starting with previous versions and then articulating a new customer need: "What they ask now is can we segment the video into meaningful chunks and then extract metadata per chunk." Open a product update video with a specific time marker to set up a concrete before and after story.
- Running
- 38 days (longevity = it's converting)
- Hook
- On-screen text appears: "With TwelveLabs". The background is a gradient from light gray to light green.
- Audience
- The ad targets businesses and content creators interested in monetizing video content and optimizing ad placements, evidenced by features like 'Shoppable Video Experiences' and 'Contextual Alignment Analysis'.
Why it works. Uses the brand name as a lead-in to a step-by-step shoppable video demo, tying a monetization feature directly to TwelveLabs.
Demonstrating product detail extraction from video (0:06-0:16): The video shows a cooking video with a sidebar listing ingredients and a 'Shop at Amazon' button, highlighting 'Food coloring' as a shoppable item. Structure a feature-announcement video as a clear step-by-step demo with on-screen labels calling out each capability.
Show 3 more ads
- Running
- 38 days (longevity = it's converting)
- Hook
- 0:00-0:03: "What's edgy for a vodka brand" is spoken over a white background with a black outlined box containing the text "Vodka" and "Target Audience: Young Adults" next to an image of a colorful cocktail.
- Audience
- The ad targets businesses or content creators who need granular control over content moderation, indicated by the explicit mention of "Your Brand. Your Rules." and the demonstration of custom filtering for different target audiences (vodka for young adults vs. kids app for children).
Why it works. Uses an alcohol brand as a stand-in test case to prove the moderation model tunes its 'edgy' threshold per audience and category.
Problem-agitation hook with contrasting examples (0:00-0:06): "What's edgy for a vodka brand is a violation for a kid's channel." Pick one concrete, unexpected vertical for your demo and show the exact audience or parameter card your tool applies to that case.
- Running
- 38 days (longevity = it's converting)
- Hook
- 0:00-0:03, a Mac Finder window showing many video files, with the text overlay "10,000 videos."
- Audience
- The visual of a Mac Finder window and the problem of managing '10,000 videos' with 'no tags, no descriptions, no structure' targets professionals or organizations dealing with large volumes of unorganized video content, likely in media, marketing, or content creation.
Why it works. Opens on a real Finder folder with a huge file count so the target media manager instantly recognizes their own overflowing archive.
Problem Agitation with Specific Numbers (0:00-0:03): The video opens with a Mac Finder window full of video files and the text overlay "10,000 videos." followed by "No tags, no descriptions, no structure." Screen-record your own messy folder or inbox with a real count overlay instead of a mockup, so the scale problem feels documented.
- Running
- 38 days (longevity = it's converting)
- Hook
- Visual: A video player showing a cartoon titled "Popeye the Sailor Meets Sinbad the Sailor". Text input field below the video is empty. Audio: None.
- Audience
- Targets developers, data scientists, or content managers who need to quickly extract structured information from video content, indicated by the JSON output and the technical nature of 'metadata extraction'.
Why it works. Shows a silent live demo pulling structured metadata out of a recognizable cartoon, letting the interface itself carry the value proposition.
Direct demonstration of a specific feature: The video shows a user typing a command to "Extract metadata from this video into JSON" (0:01-0:04) and then displays the generated JSON output (0:05-0:11). Record a silent screen capture of your tool processing a recognizable piece of content start to finish, with no voiceover needed.
08 The playbook
The channels are not separate. They are one system where each stage feeds the next. Here is the read, then the plays to run tomorrow.
The inferred playbook
TwelveLabs' acquisition loop starts in paid attention and enterprise cloud placement, converts through a usage-metered API, and compounds via cloud-marketplace distribution rather than an owned audience.
Three concurrent ad libraries surface pain-point messaging around compliance review, sponsorship ROI, and AI-clip QA, while Google spend simultaneously defends the company's own name in search, a belt-and-suspenders approach around a small but valuable branded-search pool.
The free-tier-plus-pay-as-you-go Developer plan, priced at $0.042 per minute of video indexing and $4 per 1,000 search queries, lets a prospect start testing without a sales call, with custom Enterprise contracts as the upsell.
Dig deeper+3▾
Jae Lee has said the Amazon Bedrock listing of TwelveLabs' models produced more than 30,000 enterprise agreements within seven weeks, a distribution velocity he called impossible to reach through direct sales alone, making the AWS partnership, deepened the same week as the $100M Series B, the real compounding engine rather than any single campaign.
There is no owned-audience or community engine underneath the paid and partnership layers. The founder's social presence went quiet after 2024 and the organic-content system stays thin, so the loop depends on continuously funded ad spend and cloud-marketplace placement rather than an asset TwelveLabs controls outright.
the free-to-paid conversion rate from Developer-tier usage into committed Enterprise contracts, the close rate on Bedrock-sourced agreements, and blended customer acquisition cost across the three concurrent ad platforms.
Stack your milestone with earned media
The proofAt the seed stage in 2022, Jae Lee leaned on earned media, landing a VentureBeat feature and a Techstars "#GiveFirst" podcast slot to reach investors and builders rather than paying for reach.
The adaptationTime your own earliest funding, launch, or paying-customer milestone to a pitch sent to 5 to 10 trade newsletters and podcasts your buyer already follows, leading with the concrete number, "$12,000 raised," "first 10 paying customers," rather than a product description, since a hard number is what gets a founder booked. Start by listing those 10 outlets and drafting one pitch line built entirely around your single best number.
Cost: $0 · Time to signal: weeks · Works pre-PMF: yes
Run a standing public session in your niche
The proofTwelveLabs' "Multimodal Weekly" webinar series has run past its 100th session, pulling outside researchers and partner companies in as guests to build category authority episode by episode.
The adaptationLaunch a standing 30 to 45 minute weekly or biweekly session in your own niche, inviting one outside practitioner or customer per episode instead of only speaking yourself, then publish the recording as a video or podcast asset. The compounding value is in the running episode count and guest network, not any single episode's polish. First step: book three guests for the next three sessions before announcing the series publicly.
Cost: $0 · Time to signal: months · Works pre-PMF: yes
Get your integration partners to publish about you
The proofTwelveLabs' joint webinars with video-editing platform Blackbird and AWS integrator TrackIt show technology partners doing unpaid distribution simply by talking about their own integration.
The adaptationIdentify two or three companies already building on top of your product or API, offer to co-host a short technical walkthrough or co-write an integration guide with them, and ask them to publish it on their own channel, since the mechanism here is the partner's audience, not yours. Start by messaging your most active existing integrator and proposing a joint 20-minute demo.
Cost: $0 · Time to signal: weeks · Works pre-PMF: conditional, only works once at least one real integration exists to showcase.
Soft-launch a beta inside a niche technical community
The proofA TwelveLabs team-account Reddit post announcing its Jockey research preview ran in a small, topic-specific subreddit rather than a general audience, putting a beta directly in front of people already discussing the exact problem it solves.
The adaptationFind the two or three niche forums or subreddits where your exact buyer already asks about the problem you solve, then post your own beta or research preview as a builder sharing progress, not as an ad, with a title that names the specific capability rather than your company. First step: find the single most active relevant forum and read its top posts before writing anything, to match its house tone.
Cost: $0 · Time to signal: days · Works pre-PMF: yes
Chase the pain a new trend just created
The proofTwelveLabs' LinkedIn ad creative shifted from straight video-search messaging to positioning the product as a QA layer that catches morphing faces and broken physics in AI-generated video, chasing a pain point created by a newer trend rather than only refining its original pitch.
The adaptationWatch for an adjacent trend creating a new, acute pain point for your same buyer, then write one new ad angle naming that specific pain in the buyer's own words, not your product's features, and test it against your existing campaigns, mirroring the shift from a general capability claim to a named, specific failure mode. Start by writing that one hook before touching your existing ad set.
Cost: under $500 · Time to signal: days · Works pre-PMF: no, this needs an existing paid campaign and budget to test against. Not transferable at an earlier stage: TwelveLabs' AWS cloud-marketplace deal and the enterprise contract volume it unlocked depend on a $100M-plus balance sheet and negotiating leverage that only comes after a company reaches this scale, so a pre-PMF reader should adapt the partner and community tactics above rather than expect a cloud-platform placement deal to replicate the result.
You just read how TwelveLabs does it. Tomorrow it is someone else.
Every morning we take one company that is actually growing and break down where its customers come from: the ads still running after a year, the channel doing the real work, and the play you can run this week.
Systemaic · directional intelligence. Traffic, spend, and reach figures are SimilarWeb-style estimates and qualitative reads of public data, not audited numbers. Built on real public receipts.
// teardowns with a similar GTM motion
sources & receipts (29)
Live ad libraries: meta ad library · linkedin ad library · google ad library
Launch archives: Hacker News: Foundation models are going multimodal · Hacker News: TwelveLabs' Pegasus 1.2 and the $100M Scam · Hacker News: Marengo 3.0: A Production-Grade Video Foundation Model from · Hacker News: Semantic Video Search Engine · Hacker News: Foundation models are going multimodal · Hacker News: What Is Multimodal AI?
Researched facts: TwelveLabs raised $100 million in Series B funding: $100M Series B, co-led by NEA and NAVE · Series B was co-led by NEA and NAVER Ventures with several other participants: Participant · Series B proceeds fund R&D and geographic expansion: Funds go to R&D, expanding San Franci · TwelveLabs was founded in 2021 by Jae Lee and co-founders who met during Korean military c · TwelveLabs raised a $30M strategic round in December 2024 from data/cloud infrastructure p · Databricks Ventures' VP praised TwelveLabs' video-understanding technology as filling a ga · TwelveLabs' Marengo and Pegasus models became available on Amazon Bedrock in mid-2025: Mar · TwelveLabs launched its most advanced video foundation model, Marengo 3.0, at AWS re:Inven · TwelveLabs uses a three-tier pricing model: Free, Developer, Enterprise: Free plan require · TwelveLabs' Embed API is priced free, while indexing/generation are metered: Embed API pri · TwelveLabs has received sustained industry recognition since its early years: Featured on
Milestone sources: 2022-12-05: Twelve Labs closed a $12M seed extension round led by Radical Ventures (partic · 2024-06-04: Twelve Labs raised a $50M Series A co-led by New Enterprise Associates (NEA) a · 2024-12-12: Twelve Labs secured a $30M strategic funding round from Databricks, SK Telecom · 2025-12-01: TwelveLabs launched Marengo 3.0, its most advanced video foundation model, in · 2026-04: At NAB Show 2026 (Las Vegas, Apr 18-22), TwelveLabs launched Pegasus 1.5 (video r · 2026-06-01: TwelveLabs announced general availability of Rodeo, its first application-laye
Community threads: Product Hunt: A developer building Tellers.AI (a video editing tool) publicly credited Twe · Company/vendor technical blog (TrackIt): AWS solutions integrator TrackIt independently ev · Independent technical benchmark blog (Mixpeek): Independent vector-search infra company Mi
Traffic, spend, and revenue figures are estimates as noted in the report; the links above are the primary public artifacts.