How Glasp Grew to 3.5 Million Users Organically
From YouTube Summary breakout to AEO experiments
Dear GTM Strategist,
AI didn’t just change how we create - it’s changing how we learn.
Few teams understood this as early, or as deeply, as Kei and Kazuki, the founders of Glasp.
From a hospital bed in Japan to 3.5M users worldwide, their story is a masterclass in what organic growth means in the age of AI:
Building for curiosity, not just clicks.
Letting small experiments go viral.
Turning human knowledge into something that lasts.
From building one of the first ChatGPT-based Chrome extensions to figuring out AI search (AEO) - everything based on a sound business strategy and understanding users’ needs.
That’s why I invited them to present their case in this week’s GTM Strategist newsletter.
Welcome to the AEO goldrush!
Based on the survey we did with Kyle Poyar, a stunning 51% of GTM practitioners plan to increase their investments in AI engine/search optimization (AEO) in 2026 (full report coming next week!).
Ahrefs is leading this transition with Brand Radar - a tool to help you track your brand’s visibility in AI search in ChatGPT, Perplexity, Google’s AI Overviews, and other major chatbots.
P.S.: Glasp is their happy customer, too.
Now, let’s hear it from Kei and Kazuki 👇
We’re Kei and Kazuki, co-founders of Glasp - a social learning platform that helps people capture, organize, and share what they learn online.
We also built YouTube Summary, which uses AI to transcribe and summarize videos so anyone can learn faster and more effectively.
Over the past few years, Glasp grew to over 3.5 million users, entirely through organic discovery — no ads, no affiliate deals, no paid campaigns.
In this post, we’ll share how it happened:
How we anticipated the coming AI wave before ChatGPT’s breakout.
How YouTube Summary went viral and created a new category.
How we evolved from summaries to AI Clones, making human knowledge interactive and lasting.
How we scaled with Programmatic SEO, and finally,
How we began exploring AEO (Answer Engine Optimization) — the emerging frontier where models, not humans, become your audience.
Each phase wasn’t a “growth hack.” It was a small, continuous experiment that changed how we understood what “organic” really means in the age of AI.
1. How Glasp Started
At 20 years old, Kazuki, a co-founder of Glasp, faced a near-death experience.
A sudden subdural hematoma paralyzed the left side of his body, and doctors told him he could go into cardiac arrest at any moment. Lying in that hospital bed, he wasn’t thinking about success or ambition — he was haunted by a single question:
Had I left anything behind that truly mattered?
Surviving that moment changed everything. He realized that knowledge is one of the most valuable things we accumulate in life, yet it often disappears when we do. The books we read, the insights we gain, the lessons we learn — most of it fades into private notebooks, never to be shared, never to benefit others.
Later, after co-founding his first startup and experiencing a painful split, Kazuki turned to books for guidance. He read over a hundred works from history’s greatest thinkers and realized something profound: while technology evolves, human nature doesn’t.
Across centuries, people have wanted the same thing — to leave something meaningful behind.
So he began to wonder:
What if we could democratize knowledge — make it visible, interconnected, and built upon?
That question became the foundation of Glasp — short for Greatest Legacy Accumulated as Shared Proof.
We built Glasp with a simple but profound mission: to democratize access to the learnings and experiences people collect throughout their lives — turning them into a utilitarian legacy for others.
We wanted to take knowledge out of private silos and bring it into the open —
to help people see what others find meaningful, and to build a collective intelligence that transcends generations.
2. Reading the AI Wave — and Launching Fast
By mid-2022, we could sense something brewing. Transformer models were improving fast, developer chatter was accelerating, and it felt like conversational AI was about to cross into the mainstream.
So when ChatGPT launched in November 2022, we didn’t wait. On the very same day — or perhaps the day after — we released a tiny Chrome extension called ChatGPT Extension.
It was absurdly simple: about 20 lines of code. All it did was float ChatGPT on top of any webpage using an iframe. You could read an article, highlight a phrase you didn’t understand, and instantly ask ChatGPT about it — without ever switching tabs.
Despite its simplicity, it went viral. Within a week, it hit 100K–200K installs, becoming the #2 most popular ChatGPT-related extension in the world.
That experience taught us two lessons we still live by today:
Move fast, even with imperfect ideas. The wave itself can carry you.
Tiny experiments can punch far above their weight.
The success of ChatGPT Extension didn’t just give us users — it gave us momentum, visibility, and infrastructure we could reuse. When we later spotted a new opportunity around YouTube transcripts, that foundation allowed us to act instantly.
3. The Breakout: YouTube Summary
After ChatGPT’s debut, the ecosystem exploded overnight with “ChatGPT × X” products — PDF summarizers, Gmail assistants, WhatsApp and Twitter bots, Google Search overlays — everything imaginable.
We stepped back and asked a simple question:
What’s the underlying pattern here?
At that time, ChatGPT only supported text — image input wouldn’t arrive until GPT-4 in March 2023 — so the opportunity space was limited to what could be done with text alone.
And fundamentally, GPT could do two things with text:
Expansion — generate, rephrase, elaborate.
Contraction — summarize, shorten, extract key points.
So we began mapping which media formats had been tackled (PDFs, emails, web articles) — and which hadn’t. That’s when we noticed something obvious yet overlooked: YouTube transcripts.
Glasp already supported highlighting YouTube transcripts. That gave us an instant advantage. If the transcripts already existed, why not summarize them?
No one else was doing it, and the potential demand was massive. That realization became the seed for YouTube Summary.
Why We Launched It as a Separate Extension
At the time, the original Glasp extension had about 30K installs. But we knew that Chrome Web Store updates could take weeks to review — too long when the market was moving daily.
So we made a strategic call: launch YouTube Summary as a completely new extension. New extensions typically get approved in a day.
Technically, it was simple: it fetched the YouTube transcript and passed it to the user’s own ChatGPT window. Since the ChatGPT API hadn’t been released yet (it arrived later in March 2023), building on the web interface was the only practical option.
We didn’t use the GPT API directly, and that decision came with three major upsides:
Speed: No backend integration meant we could ship instantly.
Zero cost: No API fees — critical since we weren’t monetizing.
Low friction: Free access made it appealing to anyone.
The tradeoff? We couldn’t store user context or chat history — something we’d later realize had strategic value for personalization. But at that moment, speed > perfection.
From Zero to One Million Installs
The results were immediate. YouTube Summary took off even faster than ChatGPT Extension — hitting tens of thousands of installs within days.
By February 2024, just 14 months after launch, it had crossed 1 million installs — all with zero ad spend.
How did it spread?
Social virality: Influencers on X and LinkedIn with hundreds of thousands of followers shared it organically.
YouTube amplification: Creators with 200K–2M subscribers showcased it in their videos.
Press coverage: Forbes, Microsoft News, Business Insider, and Lifehacker amplified awareness even further.
Compared to the year it had taken to reach 30K users pre-ChatGPT, the difference was night and day. With YouTube Summary, it felt like undeniable product-market fit — growth was happening on its own.
We’re still deeply grateful to everyone who helped spread it 🙏
4. From YouTube Summary to AI Clones — Making Knowledge Interactive
After the breakout of YouTube Summary, we started asking a deeper question.
If Glasp could help people capture what they learn, and if AI could use that knowledge — then could we make people’s learning itself accessible, interactive, and enduring? That question led to our next big bet: AI Clones.
What an AI Clone Is
An AI Clone is your personal chat agent, grounded entirely in your Glasp highlights — from the web, PDFs, and YouTube. You can ask it questions, and it answers with citations to the exact passages you’ve highlighted.
If you choose, you can even make your clone public — letting others query your knowledge directly. For example:
“Hey Kei, what’s the best article you’ve read on product management?”
“I’m stuck on a startup issue — what would you do?”
We released the first version of AI Clones in December 2022, about a month after ChatGPT went live.
The timing mattered. Human expertise is scarce; calendars are even scarcer.
A public clone lets millions ask questions in parallel, pushing our mission of democratizing access to human knowledge one step further.
Why It Matters
Most of humanity’s knowledge remains locked in people’s heads and disappears when they pass away. With nearly eight billion people alive today — and almost none around a century from now — we risk losing vast reservoirs of insight. Many of tomorrow’s challenges have already been solved by someone, somewhere, yet that wisdom stays isolated.
Glasp exists to change that. Highlights signal what truly mattered to someone, and AI Clones turn those signals into interactive, lasting resources. By circulating knowledge instead of letting it vanish, we aim to help individuals and society move forward faster.
On a personal level, knowing that your experiences contribute to others’ growth gives life deeper meaning. And on a broader level, we’re curious about what experts actually read and learn from.
If you can’t ask Bill Gates or Elon Musk directly, could you still learn from the sources they’d highlight? AI Clones make that possible — transforming private learning into shared, living knowledge.
5. Scaling With Programmatic SEO
When YouTube Summary took off, it didn’t just become a feature — it created a new category. But as novelty faded and copycats appeared, we faced a challenge: how to sustain growth, defend our position, and expand the market — on a near-zero marketing budget.
That constraint forced us to think in two stages:
Media reach to raise awareness and Domain Authority (DA),
Programmatic SEO to compound visibility across evergreen, learning-focused topics.
From Media Reach to Compounding Visibility
At first, we focused on media. Outlets like Forbes and Business Insider were eager to cover new, useful AI tools, and that coverage gave Glasp both credibility and backlinks — a foundation for long-term visibility.
But once the “novelty wave” faded, we needed a scalable engine. That’s when we turned to Programmatic SEO, systematically transforming YouTube transcripts into searchable, indexable Q&A-style articles.
The System
Identify evergreen, educational YouTube channels (TED, Stanford, etc.).
Fetch video metadata and transcripts.
Use GPT-3.5 to summarize and convert them into Q&A articles.
Embed FAQ/QAPage schema (JSON-LD) for structure.
Add everything to our sitemap for automated indexing.
The Q&A structure wasn’t random — we’d noticed both from ChatGPT usage and Quora’s Poe launch that question-answer formats were inherently more discoverable, both for humans and for LLMs.
Results & Learnings
Traffic: Cloudflare logs showed over 2M+ weekly requests from AI crawlers (Gemini, OpenAI, Anthropic).
Search Console: ~2.2M impressions / ~14.7K clicks in 28 days.
Media mentions: Our auto-generated Q&A articles were cited by Wikipedia and Big Think — without any outreach.
While human CTR was modest (reflecting GPT-3.5-era content quality), crawler activity was massive — an early sign of what we’d later call LLM SEO.
In hindsight, the decision to embed Q&A formats early paid off. It positioned Glasp for the next frontier: optimizing not just for people, but for models that learn from people.
6. From SEO to AEO — The Memory Layer
By 2025, AEO (Answer Engine Optimization) — sometimes called GEO (Generative Engine Optimization) — was starting to gain attention. It builds on SEO’s foundation: clean structure, high-quality content, clear schemas.
But within AEO, we noticed a hidden layer that few were talking about —a growth surface quietly emerging inside LLMs themselves.
We call it the Memory Layer.
While most teams focus on how to appear in AI answers, we started asking:
What if the real opportunity isn’t just being cited — but being remembered?
If LLMs increasingly retain context about users and the sources they engage with,
then optimizing for model memory could become a new kind of growth advantage — one that compounds invisibly over time.
The “Memory Hack” Hypothesis
When ChatGPT introduced its Memory feature in early 2024, we kept an eye on how it might reshape discovery. By mid-2025, we decided to test it ourselves.
If enough users naturally encountered “Glasp” in relevant prompts — say, “YouTube summary Glasp” or “highlight with Glasp” — would the model start remembering us and re-suggesting our content later?
To explore that, we ran a small, ethical experiment through our ChatGPT Extension used by YouTube Summary users.
For free users only, we lightly modified the default prompt to:
Check if a Glasp summary already existed.
If yes, cite it; If not, generate a new summary and add a “Learn more on Glasp” link.
Here’s the actual prompt we used:
All prompts were opt-out, reversible, and clearly aligned with user intent. We weren’t trying to drive traffic — we wanted to understand how models learn from repeated brand mentions.
What We Observed
Human traffic: Almost unchanged. Most users finished inside ChatGPT — the summary was “good enough.”
AI crawler traffic: Surged dramatically. Cloudflare logs showed millions of hits from OpenAI, Anthropic, and Gemini fetching Glasp pages.
Attribution: Ambiguous, but suggestive. The rise in crawler activity mirrored the experiment timeline.
We didn’t see an immediate growth unlock — but we did glimpse a new kind of retention. If SEO is about being found, AEO is about being recalled.
And inside that, the Memory Layer may be the most under-explored growth surface today.
Measuring “AI Visibility”
To dig deeper, we built a small internal simulator — sending thousands of synthetic prompts to ChatGPT and Gemini to estimate how often Glasp appeared in responses.
The takeaway:
SEO visibility clusters around head keywords.
AEO visibility expands through long-tail conversations and follow-ups.
This mirrors how LLMs actually behave — reasoning across context, not just ranking by relevance.
Guardrails & Takeaways
We ran the Memory Hack carefully: transparent, reversible, and user-aligned.
A few users noticed the prompt tweak; most didn’t.
But the insight stuck with us: the competition for attention is moving from the search index to the model’s memory.
Most people see AEO as a buzzword. For us, it’s a living experiment — and the Memory Layer might be its most powerful (and least understood) growth lever yet.
The next frontier of growth isn’t just about visibility. It’s about persistence.
Looking back, each phase didn’t replace the previous one — it built on it. YouTube Summary unlocked accessibility, AI Clones brought persistence, Programmatic SEOscaled discovery, and AEO opened an entirely new layer. Together, they formed a continuous loop: learn, publish, and be remembered.
7. Lessons Learned — What Compounds Over Time
Across every phase of Glasp’s journey — from viral moments to quiet experiments — a few principles kept proving true.
Speed matters. Platform shifts reward those who ship fast, not those who ship perfect. Momentum compounds when you learn in public and iterate in real time.
Stay curious and pattern-match. Keep researching across fields that genuinely interest you. Spot where energy is shifting — and be ready to place a bet when a new wave forms.
Think long-term. There will be times you need to move fast for short-term signals, but always anchor your direction in the long arc of where the world is heading. Identify the domains worth investing in for years, not weeks.
Invest in what compounds. Growth happens when your effort keeps working after you stop. For us, that means putting resources into content, systems, and people — the assets that build flywheels, not funnels.
In the end, we learned that sustainable growth isn’t about chasing hacks — it’s about building the things that keep learning, scaling, and compounding over time.
📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.
✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.
🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.
🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.
 | A guest post by
|
 | A guest post by
|
Thank you so much for sharing the story behind the scenes, Maja!
The journey wasn’t easy, but looking back, it was truly joyful. I’m so proud of Kei, our team, and what we’ve built for our users.
We’ll continue exploring new ways to create value for people and future generations!
It was a pleasure to collaborate again with you, Maja Voje! Thank you so much for the opportunity! Hope our story and growth experiments are helpful for GTM Strategist readers!