"SEO's New King of the Grind": Pollo AI Founder Abiao on How to Do SEO, Work with Creators, and Run Paid Ads
In AI startup circles, there are a handful of extremely low-profile founders whose products have reached tens of millions of dollars in annual revenue — yet they almost never speak publicly. Abiao, this week's guest on Crossing, is one of them.
二是美国 KOL 的成本非常高。所以对我们来说，美国市场不是那种你一开始就能主攻的，而是慢慢渗透的。我们前期更多是靠自然增长，通过 SEO 等方式，先把全球用户拉起来，再逐步观察哪些国家或地区表现更好。
🚥 Ronghui
你们现在主要的市场分布大概是什么样的？
👦🏻 阿彪
现在来看，美国肯定是最大的，可能占到 30% 左右。其次是欧洲、东南亚、拉美这些地区。日本我们也做得不错，但日本市场比较特殊，需要专门的本地化运营。
🚥 Koji
你们在日本是怎么做起来的？
👦🏻 阿彪
日本其实是个意外之喜。我们一开始并没有专门针对日本做运营，但发现日本用户的付费意愿很强，而且用户粘性很高。后来我们就专门招了日语运营，做了本地化，包括支付方式、客服这些。现在日本大概能占到 10% 左右。
🚥 Ronghui
回到产品迭代的话题，你们从套壳到现在这个版本，中间有没有经历过什么特别大的挫折或者转折点？
👦🏻 阿彪
挫折肯定有的。最大的一个转折点是，我们接入的某个模型突然调整了 API 政策，限制了我们这种第三方平台的调用。那段时间我们的视频生成功能几乎瘫痪，用户投诉很多。
这个事件让我们意识到，完全依赖单一第三方模型风险太大。所以我们开始加大自研投入，虽然不自研底层大模型，但我们在上层做了很多东西，比如自己的视频理解模型、风格迁移模型，还有一些轻量级的生成模型。这样即使某个第三方模型出问题，我们也有备选方案。
另一个转折点是在用户增长上。我们早期主要靠 SEO 和自然流量，增长比较平缓。后来我们尝试在 TikTok 上做内容营销，发现 AI 视频的特效内容在 TikTok 上传播力很强。我们做了一个"AI 亲吻"的特效，在 TikTok 上爆火，那一周给我们带来了几百万的新用户。这让我们意识到，泛娱乐化的内容营销可能是我们这种全家桶产品的一个重要增长引擎。
🚥 Koji
那个"AI 亲吻"特效具体是什么？能展开讲讲吗？
👦🏻 阿彪
其实就是用户上传两张照片，AI 会生成这两个人亲吻的短视频。技术上不算特别复杂，但情感冲击力很强，很适合在社交媒体上传播。我们在 TikTok 上做了挑战赛，用户生成后分享自己的"AI 亲吻"视频，带动了大量 UGC 传播。
当然这种内容也有争议，我们后来也加强了内容审核，防止被滥用。但这件事让我们看到了 C 端传播的巨大潜力。
🚥 Ronghui
你们现在团队 90 多人，这个团队结构大概是什么样的？研发、运营、市场这些比例如何？
👦🏻 阿彪
研发大概占 60%，包括前端、后端、算法、测试这些。运营和市场大概 25%，剩下的 15% 是职能支持，比如财务、HR、客服。
我们算法团队增长很快，从最初的几个人到现在二十多人。因为我们在做一些自研的模型和工具，需要大量算法人才。
🚥 Koji
你们招人有什么特别的要求或者偏好吗？
👦🏻 阿彪
我们比较看重两点：一是对 AI 有真正的热情和好奇心，不是只是追风口；二是有出海经验或者国际化视野。因为我们做的是全球市场，很多产品决策都需要理解不同文化背景的用户。
另外我们团队分布比较散，深圳、长沙都有办公室，也有远程的同事。所以我们也很看重自驱力和沟通能力。
🚥 Ronghui
你们远程协作的比例高吗？
👦🏻 阿彪
大概 20% 左右是全职远程，还有一些是混合办公。我们工具用得比较重，飞书、Notion、Figma 这些。但核心团队还是尽量在一起，面对面沟通效率确实更高。
🚥 Koji
回到商业化的话题，你们现在千万美金级 ARR，这个收入结构大概是什么样的？订阅、广告、API 这些占比如何？
👦🏻 阿彪
主要是订阅收入，占到 80% 以上。我们采用 freemium 模式，免费用户有基础额度，付费会员解锁更多功能和更高额度。
广告收入很少，基本不做。API 收入有一些，但占比不高，大概 10% 左右。我们开放了一些 API 给开发者，但不是重点。
🚥 Ronghui
你们的付费转化率大概是什么水平？
👦🏻 阿彪
整体付费转化率在 2%-3% 左右。但不同产品差异很大，Hix AI 这种效率工具转化率高一些，能到 5% 左右；Pollo AI 这种创意工具低一些，1%-2%。因为创意工具的用户更多是来玩一下的，付费意愿没那么强。
🚥 Koji
你们有考虑做企业版或者 B 端业务吗？
👦🏻 阿彪
Hix AI 已经在做企业版了，主要是一些内容营销团队、电商团队在用。Pollo AI 目前还是聚焦 C 端，但也在探索一些 B 端场景，比如广告公司的创意生产、电商的商品视频生成这些。
我们觉得 C 端和 B 端的产品逻辑很不一样，不能简单把 C 端产品加个企业功能就卖给 B 端客户。B 端需要更强的定制化、更好的服务支持，我们还在摸索。
🚥 Ronghui
你们现在融资情况怎么样？是 bootstrapped 还是拿过机构的钱？
👦🏻 阿彪
我们拿过一轮天使投资，后面就没有再融了。现在基本是靠自我造血，收入覆盖支出还有盈余。
我们其实比较谨慎对待融资。一方面是我们现金流还可以，不需要急着拿钱；另一方面是不想让外部股东过多干预产品节奏。当然如果未来要做更大的事情，比如自研底层模型，可能还是需要融资。
🚥 Koji
你们现在最大的成本支出是什么？
👦🏻 阿彪
研发和算力。算力成本增长很快，尤其是视频生成，调用第三方 API 的费用很高。我们在想办法优化，比如做一些缓存机制、预测用户行为来预加载，减少重复生成。
另外我们也在测试一些国产模型，成本比海外模型低很多，但效果还有差距。这是个权衡。
🚥 Ronghui
你们对 Pollo AI 的最终愿景是什么？想把它做成什么样的产品？
👦🏻 阿彪
我们希望它成为每个人的 AI 创意工作室。无论你是专业设计师，还是普通用户想做个有趣的视频发朋友圈，都能在 Pollo 上找到合适的工具。
长期来看，我们认为 AI 会重塑整个创意产业的生产流程。现在的创作者需要学很多复杂软件，未来可能只需要描述你的想法，AI 就能帮你实现。Pollo 想降低这个门槛，但同时又保留足够的专业控制力，让专业创作者也能用。
🚥 Koji
这个愿景很大，但也很卷。你们怎么建立自己的护城河？
👦🏻 阿彪
几个方面吧。一是模型层的灵活性，我们不绑定单一模型，而是给用户选择权，同时自研一些差异化能力。二是工作流整合，从素材管理、生成到编辑，一站式完成，用户不用在多个工具间切换。三是社区和数据飞轮，用户创作的内容可以沉淀在社区里，反过来训练我们的模型，形成正向循环。
但说实话，这个领域变化太快，很难说有什么牢不可破的护城河。我们能做的就是保持敏锐，快速迭代，比竞争对手更早看到趋势。
🚥 Ronghui
最后一个问题，对于现在想做 AI 出海创业的年轻人，你有什么建议？
👦🏻 阿彪
几个吧。第一，不要一上来就做平台，先从单点工具切入，验证需求后再扩展。我们早期做了 20 多个产品，很多都死了，活下来的是那些找到了明确用户价值的。
第二，出海不要只盯着美国。美国市场确实大，但竞争也最激烈。东南亚、拉美、中东这些市场增长很快，门槛相对低，可以先积累经验和现金流。
第三，重视 SEO 和内容营销。我们早期靠这个冷启动，成本很低但效果持久。现在流量越来越贵，要尽早建立自己的流量渠道。
第四，控制成本，保持盈利。AI 创业烧钱很快，算力、人力都是大头。能自我造血的公司才有选择权，不用被资本牵着走。
第五，团队要精悍。我们 90 多人做了两个头部产品，很多同行团队是我们的几倍大，但效率没我们高。AI 时代，小团队用好的工具和方法论，可以做出很大的事情。
🚥 Koji
非常实在的建议。感谢阿彪今天来「十字路口」分享，也祝 Pollo AI 和 Hix AI 继续大卖。
👦🏻 阿彪
谢谢，也欢迎大家来用我们的产品，有兴趣加入团队的也可以联系我。
收听完整播客：
关于「十字路口」：
「十字路口」是由 Crossing 出品的一档关注中国科技出海、AI 创业与全球化商业的播客节目。我们邀请一线创业者、投资人和从业者，分享最真实的经验与思考。欢迎订阅关注。
Second, users have extremely high expectations for products, and if they see it's a wrapper product, they might not even give it a second look.
This isn't something money alone can fix. Even if I wanted to break into the US market — say, by throwing $100,000 or $200,000 at influencers — it's not something I can just decide to do. Theoretically, focusing on a single country for breakthrough would definitely work better.
But again, this isn't something I can will into existence. For now, our more feasible approach is to go broad, work with every influencer we can reach, and pick the lowest-hanging fruit first. This way we get some revenue from each country, and it adds up to decent results. Once we have enough strength, and our product can genuinely earn the respect of those more discerning, higher-standard top-tier KOLs — that's when we'll go after them.
Pollo AI's Three-Pillar Traffic Growth: SEO, Influencers, and Ads
🚥 Koji
Actually, after friends heard Abiao was coming on Crossing, a lot of them told me they especially wanted to learn about traffic and SEO from you. Because it feels like there's this legend...
🚥 Ronghui
If you want to do this stuff, you definitely need to go study under Abiao first (laughs).
🚥 Koji
Right, so I'm really eager to ask — Pollo has only been out for a few months, and it's already reached a pretty stable ~4 million monthly visits.
I'd love to hear how you achieved these traffic numbers.
👦🏻 Abiao
First off, SEO is really just one piece of the puzzle.
Our company is currently pushing SEO, influencers, and ads simultaneously. This product didn't grow purely through SEO. And even with ads, we split between Google ads and FB ads. So it's this three-pronged approach together that got it to this level.
It's also built on several years of accumulated experience going global before this. What I mean is, when you see me full from eating the tenth bun, I already ate nine buns before that. Those nine buns are probably what people didn't see.
If we're talking specifics about how it's done, I'd say it's those three dimensions I just mentioned: SEO, influencers, and ads.
From the SEO angle, if you break it down, it's actually quite simple. One is core features, two is micro-tools, three is models, four is articles. Build out all these pages first, make sure you have landing pages that can actually receive the traffic. Only with these landing pages can traffic actually arrive there.
For example, say you have 100 pages, each getting 10 visits — that's 1,000 visits a day. The logic is that simple. So you need those 100 pages first to even have the possibility of 10 visits per page. So you start by building out pages across those modules I just mentioned — that's the first thing.
Second, you probably need to do some PR and backlink promotion to give the site some authority, some credibility, so your rankings can improve.
And of course there are many other finer details. Like how to better optimize user experience, how to make things more SEO-friendly, more crawler-friendly, and so on. Those are the more granular parts. The broad framework is what I described above.
🚥 Koji
There was a WeChat public account article that got shared quite a bit. I saw it was specifically about Pollo, calling you the new SEO king of Google.
It analyzed some of your SEO methodology. I'm not sure if you saw that article — what did you think? Was it accurate?
👦🏻 Abiao
Broadly speaking, it wasn't wrong. Although we're called the new SEO king of Google, I think the real king among kings would be Monica. From what I can see, they're way more intense than us — they're doing much better on SEO.
🚥 Koji
How so?
👦🏻 Abiao
Basically every AI-related or work-scenario tool on the market, they've built out. They're more intense than us.
They write articles on pretty much everything. We haven't been that thorough, that intense, so I think they're the true king — whether it's product features or SEO, they're extremely intense.
🚥 Koji
Actually, getting to this point, I feel like what Abiao said earlier doesn't seem that special. There aren't any clever tricks that people apparently don't already know. What you're describing sounds more like fundamentals, or common knowledge you'd pick up from entry-level textbooks when you first started.
Is it because you think SEO is just common sense, fundamentals practiced and repeated over and over? Or are you holding back some secrets from us?
👦🏻 Abiao
Actually, everyone knows the broad framework. I think where the real gap opens up is in all the finer details.
🚥 Koji
Can you share some of those details?
👦🏻 Abiao
Sure. For example, on Pollo we made a relatively small innovation: when users arrive, they can immediately see how the tool works. They don't need to log in to access the backend. Right now, a lot of traditional tools or websites or products — their front landing page is just a showcase.
You click "Try it now," go to the backend, then use it. At that point, if a user wants to use Feature A, but has to dig around the backend to find it — that's pretty poor user experience. But with us, when users arrive, Feature A is immediately usable. Before and after login, it looks the same — there's no sharp front-end/back-end divide.
But then, to get Google to extract more information from the page, if I only had the feature without any related text description, that wouldn't be good for SEO. So before login, they see the page with the functional area on top and text description area below — or sample showcase area, or how-to content and descriptions. After login, it becomes an immersive functional area and the bottom content gets hidden.
This is what I consider a pretty significant innovation for Pollo as a product. On the SEO level, it balances both SEO and user experience.
Actually, the core of SEO is user experience. Whether you can give users a good experience, whether you can deliver a satisfying answer to them. This answer could be the information they want, or using a product — it's all a form of delivery.
🚥 Koji
Earlier when you mentioned the four parts of SEO, the third part you said was models. Could you expand on that a bit?
👦🏻 Abiao
This is really because there are many video and image models on the market — Keling AI, Runway, Sora, Luma, Hailuo AI, and so on. Users will actually search for these brand terms, so we build dedicated landing pages for them, and we do integrate their models.
If someone searches for Keling AI, they might go to Keling AI's own page, but they might also come to our page to use it. That's what I mean by model pages.
🚥 Koji
Got it. This is actually similar to a common SEO tactic of sub-feature pages, though it feels like the competition would be fiercer. With such intense competition for these big terms, how do you actually win?
👦🏻 Abiao
SEO is really kind of mystical — it's hard to say you'll definitely rank, honestly I can't guarantee it.
A lot of the time, ranking in the US market is very difficult. For example, same English language, but in other English-speaking countries like Australia, New Zealand, Singapore — I might be able to rank near the top for "Runway," but in the US I might not rank. I'm capturing that portion of traffic.
Or because Runway hasn't done multilingual — say, no Spanish, no Portuguese — when people search in those languages, I've done multilingual, so my ranking might be relatively higher there, and I can capture that multilingual traffic.
🚥 Koji
One hot topic today is using AI to write blogs to get Google SEO traffic.
What's your take on this? Have you actually tried it? Any results?
👦🏻 Abiao
First, I think our experience with AI-written blogs is relatively limited. We use AI as assistance — half AI, half human. And I think for pure blog content, users are more discerning.
There's a new trend now — writing more human, more like a person, more like a friend. Like how I'm talking to you right now, conversational and flowing. For example, if I'm writing "The Ten Best AI Video Tools," my opening would be: I've tested a lot of tools, I have a lot of experience in this area. Now let me tell you what issues they each have — writing like a friend, and that actually performs better, it's more engaging.
Why is this happening? It's like how everyone can get full from eating now, but I expect to eat better. Users are more discerning. AI-generated content is all rigid, mechanical, cold, no human warmth. When you have human warmth, that's your differentiator — that's what discerning users are looking for.
🚥 Koji
Why haven't you taken the fully AI-written approach?
👦🏻 Abiao
It definitely can't achieve that level of quality. Being a hands-off manager isn't realistic.
🚥 Ronghui
Is there anything you see others doing that you think is a widespread misconception or mistake?
👦🏻 Abiao
On the SEO front, the first misconception is thinking it never costs money — it does cost money, just indirectly.
For example, writing articles costs money, building pages and features costs money. And second, doing backlink promotion, building up authority and PR — that costs money too.
It's not pure free-riding, not milking Google for free. That's not the logic. You still have to spend money, though it is a relatively high-ROI approach.
Influencer Marketing and Ad Buying: Timing, Breakout Moments, and the Importance of Professional Teams
🚥 Koji
When doing backlinks, in various SEO WeChat groups and discussion forums, I've heard quite a few different perspectives.
Some people encourage starting with free backlinks, while others argue you should buy high-quality backlinks right away — that links worth $1,000 or more are worth buying, and anything cheaper isn't worth considering.
From your experience, how would you advise people to approach link building?
🚥 Ronghui
How do you spend this money where it actually counts?
👦🏻 Abiao
Strictly speaking, Google doesn't allow buying backlinks. So when it comes to purchased links, Google will simply nullify the ones it considers ineffective.
The general logic here is actually pretty intuitive if you think about a website's lifecycle. A brand new site that suddenly gets a highly trusted vouching for it — that kind of backlink is unnatural. When a site first launches, it's inevitably small. It might start with some mediocre backlinks, or just basic ones from places like Medium, Quora, and so on. That's completely normal. Then gradually more people start noticing it, and higher-quality backlinks begin to appear.
That said, this doesn't mean you'll get penalized for building two or three, or even five high-quality backlinks early on. Definitely not. In Google's eyes, your backlinks should follow a natural growth pattern, but there's also a tolerance range. They won't penalize you for doing a handful — otherwise sabotaging competitors would be trivial, and Google wouldn't allow that.
For example, if you start out with zero low-quality backlinks and exclusively buy high-quality ones, all $1,000+. My company has money, I'm loaded, I drop $100,000 all at once — that's when it could backfire, because it's unnatural.
When building backlinks, imagine what a normal website looks like as it grows naturally. What should the rhythm of its backlinks be in terms of velocity, quality, and frequency? Follow that rhythm. Say I build 100 backlinks this month, then do nothing the next month, nothing the third month, then 100 again in the fourth month. That's unnatural, right?
A normal progression has a steady slope — maybe twenty to thirty backlinks per month, or around ten. Gradually increase, or maintain that volume consistently. Don't have huge fluctuations, nothing one month and a flood the next. That's what looks unnatural.
🚥 Koji
I feel like SEO practitioners I know often have this sense of powerlessness — they've been at it for what seems like forever with no visible feedback. They don't know if what they're doing today will take effect in one month, two months, or never.
What's your take? What advice would you give to these friends feeling stuck?
👦🏻 Abiao
First, SEO is definitely relatively slow — you need three to six months to see results.
If changes in between don't get timely feedback, you don't know whether what you're doing is correct. Especially when you don't have much experience, it's quite possible to work for three to six months with nothing to show for it. Particularly if you're not a very specialized person.
Because SEO is an extremely systematic undertaking. It includes content, site structure, URL structure, backlinks, keyword direction and competitiveness. There are so many influencing factors — it's a fairly systematic project.
🚥 Koji
At this point people often think, I'll hire a senior SEO consultant, or recruit someone with SEO experience.
Whether hiring a consultant or an employee, what are the one or two key questions you'd ask to test whether someone actually knows their SEO?
👦🏻 Abiao
First, I'd look at how they built something from 0 to 1. Take a specific website as an example — what's their thought process? They need to have a rough framework in mind. I think this is extremely important. Most people don't have this framework; they only know one particular module.
Second, I'd ask about their understanding and perspective on SEO. If they only talk about finding keywords, making pages, building backlinks — I'd consider that a pretty shallow level of understanding.
🚥 Koji
What kind of answer would show deeper understanding?
👦🏻 Abiao
Look at the dimensions. If you want to go one level deeper, I think the simplest phrase is satisfying user intent.
🚥 Koji
In the SEO field, is there anything you consider common knowledge that many people either don't agree with or simply don't know about? Anything you could share?
👦🏻 Abiao
They always think SEO is a cure-all, that it can do everything. But SEO has its bottlenecks and ceiling. It can only drive so much volume — it can't carry everything.
For instance, doing $10 million or more in revenue? That's impossible. There have to be other layers involved. So we need to understand the ceiling of each marketing channel, what it can realistically achieve. It has limitations, and it's not suitable for every industry.
🚥 Koji
What's the difference between someone who does SEO well and someone who's truly top-tier?
👦🏻 Abiao
I think top-tier people have more architectural thinking, more holistic thinking.
First, they need some understanding of the product's positioning and future direction.
Second, they also need to understand marketing and SEO. How to integrate these well with their own product, making it work for both user experience and SEO. And the entire thread needs to be clear — I think that's genuinely difficult.
🚥 Ronghui
Have you interviewed anyone who said something about SEO that really impressed you, that made you think they really got it?
👦🏻 Abiao
Actually, no. Everyone we hired for SEO at our company had no prior experience. We wrote our own training documents and trained them internally.
🚥 Koji
If someone can't join your company and learn from you, and has to self-study online — what documents or tutorials would you recommend?
👦🏻 Abiao
I can share from my own experience.
When I worked at Wondershare, first I'm very grateful they gave me a platform to practice and execute SEO-related work. At the same time, one thing that set me apart from colleagues was that I read a lot of English SEO blogs — Ahrefs, Semrush, Authority Hacker, Backlinko. Their SEO articles are extremely professional and top-tier, definitely at the cutting edge. They update these materials regularly, and in my view they're the best SEO learning resources globally. Their articles are written very accessibly, with very simple vocabulary, and the visual-text pairing and layout are very pleasant.
I still remember when I worked at Wondershare, I'd be downstairs picking up my takeout waiting for delivery, reading English SEO articles on my phone. I'd look at that stuff and think, this is incredibly substantive, this is excellent. I devoured it like a sponge soaking up nutrients.
I think if you read 100 articles, you'll basically have no major gaps in theoretical knowledge. Then if you look at 50 to 100 well-executed SEO sites, you'll basically have no problems with template cases. It's actually not that hard — reading 100 articles doesn't take that much time, looking at 100 sites. Look at their architecture, their writing style, URL layout, and so on. When you've seen enough good examples, you naturally know what to do.
🚥 Ronghui
There's something of the "read three hundred Tang poems" feeling to it.
Pollo's Influencer Marketing: Why the "Hard Push"? Breaking Through Without a "Breakout Moment"
🚥 Koji
We've talked a lot about SEO. Next we'd love to hear Abiao's thoughts on influencer marketing. Influencer marketing has played a fairly significant role at Pollo.
👦🏻 Abiao
Actually, doing influencers for Pollo hasn't been that easy either — it's more of a hard push approach.
🚥 Koji
Is it because, as mentioned earlier, Pollo isn't a product with major technical or interactive innovation, so influencers don't get excited about it, don't feel drawn to it? Is that the reason?
👦🏻 Abiao
That's one reason. The second is that it lacks a breakout moment, a hook. For example, if I had my own large language model, because what you generate comes from your own technology, you're impressive, right? That's a buff, that adds points.
But we're a wrapper. Influencer marketing really comes down to having a breakout moment. Whether your product has a hook, something controversial, something interesting — that's what gets users to like, comment, and share. Because they find it fun and engaging, or controversial and debatable, it goes viral. Otherwise, dryly pushing a product is hard to make work.
Let me give a typical example. When GPT-4o released its image generation feature, its biggest highlight before that was Chat. That was a major interaction update, an innovation, and it was genuinely good. But if that was all it had, it definitely wouldn't have reached its current level of spread.
What pushed it to true viral scale was that it achieved the Ghibli style. That was fun, everyone started playing with it and found it delightful — that's what took its reach to the next level. This is about being fun, users wanting to play with it, loving what's interesting. When people scroll short videos, they're not here to hear you seriously explain how to use something — that's boring. They want fun. Another example: Heygen went viral because Guo Degang speaking English instead of Chinese was hilarious and entertaining. That's how people discovered it was a digital human company.
🚥 Koji
I actually started my own consumer brand called Tangdao, selling pillows. Tangdao achieved pretty good results. On Tmall and JD.com, we were regularly category leader in pillows, and to this day during major sales events we're still at the top. The pillow category has been sold for a hundred years.
Why could we make a new pillow and sell it as number one across all platforms? There was a lot of differentiation behind it. For instance, our pillow is called the Cat Belly Pillow. First, many people hearing this get curious and want to click to see what it is. And we differentiated significantly from common pillows in color, visuals, materials, and shape — very obvious differentiation.
Bringing this back — Abiao just mentioned that pushing influencers was a hard push because there was no breakout moment. The Cat Belly Pillow I described has clear differentiation. I'm curious, for a product like this, how do you do a hard push? If the ROI from pushing isn't good, what are your current options?
👦🏻 Abiao
I didn't know any of this when I first started Pollo. I had to pay tuition and gradually develop that intuition.
Back then, we also did hard pushes without overthinking it — just had the team do mechanical work, everyone grinding away at it. The ROI definitely wasn't good. Even now I don't really know what the ROI is, because it's hard to measure.
So how do we push now? We mainly push effects, fun effects. But fun effects have their own problem: users think it's fun, they give it a like, okay, and then they bounce — no payment.
🚥 Koji
Actually, I've noticed some KOLs introducing Pollo really emphasize that you're an all-in-one product, a one-stop shop for the latest video foundation models.
👦🏻 Abiao
That's something we're trying now, because it genuinely is a highlight. But whether this approach works well or not — same as before, it's hard to measure, or we haven't gotten granular enough to measure it.
🚥 Ronghui
How do you measure results for this direction you're trying?
👦🏻 Abiao
We mainly measure by CPM. Actually, we don't really measure results — we measure by budget, calculate by CPM. If the CPM is low, I'll do it. Whether it's good or bad depends on views, on the final CPM.
🚥 Koji
Do you think Pollo is entirely marketing-driven right now?
Because from what I'm hearing, at least currently, it's still a wrapper product, an all-in-one N-in-1 product. So when we talk about SEO, influencers, ads, and what you've achieved today — how much of that is marketing, and how much is product functionality and product innovation?
👦🏻 Abiao
First, timing. Timing matters a lot. We started relatively early — that's one.
Second, I'd say marketing probably accounts for half. Even though it's a wrapper product, the interaction design is pretty decent.
Including our next version's interaction — there will be a major innovation. I'll keep that under wraps for now, you can experience it when it comes out. We're gradually treating it as a real product, not just a wrapper tool.
Revisiting Product Success and Failure: Timing, Core Value, and Lessons From "Tuition"
🚥 Koji
Timing is crucial in entrepreneurship, especially in the AI era. Do you have any techniques for getting it right? You've made dozens of products, whether Hix or Pollo — looking back, your timing was pretty good. Everyone says speed is everything, but beyond being fast, what else have you learned about timing?
👦🏻 Abiao
First, I think Hix definitely didn't get the timing right — we started too late. It launched in May 2023, by then ChatGPT had already blown up several times. Monica was the company that really nailed the timing, they captured that wave. We definitely didn't.
As for how to find these new things, or get the timing right — I think you have to constantly pay attention to the latest products, tech developments, or open-source solutions. Then think about how to apply them to your product, or what new use cases there might be. You have to keep thinking about this stuff continuously.
For us, I might be different from other companies in that we make many products — I'm not that focused.
🚥 Koji
Do you think that's good or bad?
👦🏻 Abiao
This approach has its pros and cons.
For someone like me, having made so many products, none of them seem to have gone from 10 to 100 — that's the downside. Maybe because I'm not focused enough, not deep enough. In terms of my overall personality, I probably wouldn't pin my hopes on just one or two products. Mainly, I just enjoy making new things.
🚥 Ronghui
You prefer 0 to 1?
👦🏻 Abiao
Yeah, I find 0 to 1 exciting. I'm naturally drawn to making new things, exploring, trying things out.
If I'm staring at the same thing for half a month, a month or two with nothing new to do or show for it, I find it boring. Sure, theoretically I could do existing things better and deeper, I'm just not that single-minded.
🚥 Koji
Do you have any regrets? In the past two years, you haven't focused on one thing — you've spread your energy around.
👦🏻 Abiao
No, from where I stand now, I think it was the right decision.
🚥 Koji
Why?
👦🏻 Abiao
Because if I had only focused on one thing, I don't think I would have come up with more new good product ideas or actualized them. And if I had just gone vertical on one thing, it might not have performed that well either.
🚥 Ronghui
That's pretty interesting. We had a guest before, Patrick, who said looking back, he wished he'd been more focused. In his podcast episode he talked about how he also tried a lot of things early on. Looking back, he felt he had underestimated his product's potential — should have concentrated more energy on it.
👦🏻 Abiao
I think it really depends on the specific situation.
Some more B2B-oriented products, you can go very deep on. Some more consumer-oriented products, there's honestly not much to do — after a year, the product is basically done. So it depends on two things: first, the product's nature; second, whether it genuinely has potential, whether the ceiling is high enough to justify dedicated investment.
🚥 Ronghui
If you're starting a company, what goals do you set for yourself?
👦🏻 Abiao
I don't have clear goals. Entrepreneurship is my life. If I'm not starting companies, not making these products, I have nothing else to do.
If there had to be a goal, getting some validation, having a sense of achievement, pride, self-satisfaction — that would be nice. Also getting recognition from peers.
🚥 Koji
Actually, hearing you talk about your entrepreneurial goals, you seem pretty relaxed. But wouldn't some VCs hear this and think, this guy's too relaxed, he doesn't seem very ambitious.
So I'm curious, small question: do you have any entrepreneurs you particularly admire, or role models you'd want to become?
👦🏻 Abiao
That kind of ambition depends on the occasion and who you're talking to.
I definitely wouldn't say in this kind of setting that I want to build a multi-billion dollar company. I'm not that kind of person, it's not my style. I prefer to keep some things inside, silently guard them. The entrepreneur I admire is Lei Jun — at the height of his success, he's still on the front lines making cars.
🚥 Koji
Did you buy Lei Jun's recent Xiaomi Auto?
👦🏻 Abiao
I don't have a car, no plans to buy one. If I were planning to buy, I'd actually quite like the SU7 Ultra, especially the one with the two big nostrils on the front.
🚥 Koji
Why choose not to buy a car?
👦🏻 Abiao
Because I don't need one. I usually just take taxis, I walk to and from work, don't need to travel for business or anything — no use for a car.
Paid Advertising Secrets: Is Hiring a "Senior Buyer" Enough?
🚥 Koji
We just talked with Abiao about two of the marketing triad: SEO and influencer marketing. Next let's talk about paid advertising.
When it comes to doing paid advertising well, what insights do you have?
👦🏻 Abiao
My insight is: hire a senior ad buyer. That's it.
🚥 Koji
How do you identify one?
👦🏻 Abiao
Find a friend or a friend of a friend, pay them to interview for you.
Take me — I've never done paid advertising, I'm not good at it. Ask me for techniques and experience, I don't know. What I can do is hire someone good, let them handle it. Don't ever think, I don't understand this area, so I'll hire someone cheap — you'll only lose more. I'm increasingly deeply convinced: in areas you're not good at, never hire mid-level, always hire senior.
🚥 Ronghui
You have to be willing to spend on this.
👦🏻 Abiao
Right, whether it's product, marketing, advertising, tech, or some specific technical area.
Say you're Android or iOS, frontend or backend, Java or whatever. As long as it's not your strength, you feel you can't handle it, completely blank — hire senior first. Never hire some hack, they'll screw you. The tuition they pay, you end up paying for. When you hire senior, the previous companies already paid that tuition.
🚥 Koji
Among the dozens of projects you've done, there must have been very successful ones and very unsuccessful ones? I guess some made no splash at all. What do you think determined success or failure across these products?
👦🏻 Abiao
Great question. I think there are a few keys to success or failure. First is timing for choosing to build a product — when you build it matters. Is now a good entry point? Second is: what is this product's most important point?
For example, a failed case: we did e-signatures (overseas version). What do users care about for this product? Brand trust, compliance — there's a huge trust cost. These to me were the key points for this product. As a startup, we couldn't really deliver on these key points, so when we chose to do this, it was basically destined to fail.
But we only realized this after doing it. And timing-wise, e-signatures wasn't good either. This space had been around forever, DocuSign was already public, plus Adobe Sign, HelloSign, etc. — lots of big players. Our timing was off, the product had no differentiated entry point. The key success factors I mentioned — brand trust, compliance — we couldn't do. However you looked at it, it was guaranteed to fail.
So now, when it comes to building a new product, there are a few key things I consider: first, timing; second, the entry point; third, at the product level, what exactly is the critical factor? Is it product strength, or trust, or brand, or marketing, or advertising, or something else? I think you need to figure that out first.
🚥 Ronghui
You mentioned earlier that a lot of these lessons came with tuition fees. Can you talk about those tuition fees? Looking back now, what's the most memorable or most painful one?
👦🏻 Abiao
It's that e-signature software I just mentioned. We worked on it for more than half a year, lost several million, and spent quite a bit on marketing too.
Right, on top of that we also developed an online PDF editing tool, but that wasn't good timing either — no differentiated entry point, and extremely fierce competition.
Of course, there's something I should clarify here. The premise was this: before AI emerged, there were almost no new scenarios. Everything was old products, old use cases. It was very hard to find something new to build, so it was difficult to have any timing advantage. You could only look for differentiated product entry points.
Now it's different. Because AI has emerged, the technology is evolving by the day, and new things keep surfacing. This creates some good timing windows to choose suitable product directions.
🚥 Koji
How do you go about finding new product inspiration? You've built so many things, and you mentioned you really love making new things, so you must spend a lot of time every day thinking about what to build next.
So I want to know: how do you spend your days, and what methods do you use to find new product inspiration?
👦🏻 Abiao
Actually, I think China's AI self-media does a pretty good job — they're quick to sync new tech updates.
Theoretically, I should do something more highbrow, like browsing GitHub for open-source, trending new projects. But I'm not from a technical background, so it always feels like there's a veil between me and that stuff.
In my spare time I also watch videos and follow some AI bloggers. They're very diligent, update quickly and keep things current. New AI-related stuff, whether from GitHub, Twitter, or big company news — they curate it all, doing a layer of processing for us. It's secondhand, but when you're scrolling on your phone in your free time, you might come across something interesting, make a mental note, and dig deeper when you're back at work. That's what I usually do.
🚥 Koji
When Pollo first came out, it attracted enormous attention. But at the same time, I also heard a voice saying: Pollo isn't anything special, it seems like I could build it too. Some people felt there was no moat.
From your perspective, Abiao, what non-consensus thing did you see at the time? Or what did you see that others couldn't do, but you and your team could?
👦🏻 Abiao
Not really, at this point. Others can indeed build it too, and we can't pull that far ahead — that's something we worry about.
But we should iterate going forward, and maybe we can create some distance. Products like ours without technical barriers — theoretically anyone can build them. The products we build in the future will definitely be like this too. So I don't pay too much attention to this; it won't be a key factor for me. What I need to do is build the product well, push forward as quickly as possible, relatively fast enough that it's somewhat taxing for others to catch up. On marketing, we also need to go all out.
🚥 Koji
Actually, with Pollo up to now, I think many people consider it quite miraculous. From zero to 4 million in traffic in just a few short months — it seemed like it happened overnight.
From the outside, it looks like 3 million every two months, then 4 million. But from the inside, do you feel there were clear traffic growth milestones? That is, were there specific events that brought explosive growth to Pollo?
👦🏻 Abiao
It did grow quite fast. Speaking of milestone events — there weren't really any. Internally it was fairly stable growth, without any sudden steep growth curve. I think it was more the compound effect of three moves working together, which produced relatively good growth.
🚥 Ronghui
Earlier you mentioned you also tried AI plus music, but that didn't take off. Why?
👦🏻 Abiao
First, our API costs were quite expensive — we were using MiniMax and Tiangong at the time. Second, users weren't paying much either. Though I know some other players are doing quite well in this space. Also, the user persona wasn't very clear. Sometimes it still comes down to luck.
🚥 Ronghui
So do you think the main reason has more to do with you, or with the music industry itself? Is there something particularly difficult about this industry?
👦🏻 Abiao
I think both. First, when it comes to user purchases, they basically just buy the cheapest option — they won't buy the premium version.
Second, our API costs were indeed quite high. Compared to other competitors like Suno, we had no advantage whatsoever. Suno's product is better and cheaper than ours. For us doing this API wrapper kind of thing, it's very hard — Suno is just too well-known. You need to have one edge: either cheaper than them, or better than them. But you can't be better than them. So you're just left scooping up small fry, catching dead fish.
🚥 Koji
What do you mean by "catching dead fish"?
👦🏻 Abiao
A blind cat bumping into a dead rat — that's a common trait of C-end products. Some people just don't know about Suno; they happen to know about you. Isn't that a blind cat bumping into a dead rat? (laughs)
🚥 Koji
There are so many AI video products now — many are already very successful, and new ones keep launching. It feels like there's a new AI video product almost every day. Among all these products, which one or ones do you personally admire most?
👦🏻 Abiao
CapCut, I guess. Though it doesn't need my admiration — it's already incredibly powerful.
Its biggest thing is that whether it's features, operations, or growth, it's all very systematic, unbeatable. A team like that could crush it in any product.
🚥 Koji
Do you think CapCut will end up being your biggest competitor?
👦🏻 Abiao
No way, how could we possibly compete with CapCut?
🚥 Koji
But CapCut might erode your market.
👦🏻 Abiao
I haven't thought about that. I think the market is huge.
Second, I think we'll still be different. Our user mindshare leans more AI-native; they're a traditional editing or template tool. We have no intention of competing with them, don't want to compete with them, and are afraid to compete with them.
🚥 Koji
Besides CapCut, among the new generation of AI video products, is there any that you personally admire? Regardless of its current scale?
👦🏻 Abiao
If we're talking about model companies, there are certainly many. But I don't think you can count model companies in this — like Keling AI, Hailuo AI, you can't really include those. On the application layer, there's Krea or Freepik, but do I really think they're doing well at the product level? I have my doubts.
🚥 Koji
So your admiration for them isn't about their product experience and interaction design.
👦🏻 Abiao
Rather, they've gotten quite big, moved quite early, and have foresight. But I don't buy their product interaction. Of course, different strokes for different folks — some people might like that kind of interaction.
🚥 Koji
What AI video product do you think is overhyped today?
👦🏻 Abiao
I think Krea AI is overhyped. They raised $83 million, with total ARR of just over $8 million, at a $500 million valuation.
I think there's quite a bit of froth there. Maybe there's just so much money in the US market that this bubble exists. Honestly, I can't figure out why they need that much money — how do you even spend $83 million? And how is a company with $8 million ARR worth $500 million? I think there's an overvaluation component there.
But a16z's influence as an investment institution is unquestionable. Their endorsement — maybe there are product development directions or strategies that we don't yet know about.
The Shifting Landscape and Mindset of Chinese Software Going Global: From Cognitive Iteration to "Relaxed" Management
🚥 Ronghui
Whether from friends or on Jike, quite a lot of people have sought you out to chat, especially those focused on the going-global space.
Have you observed, say from last year to now, what are the common concerns people have, the questions they ask when they find you? Over this period, how have these topics of concern changed?
👦🏻 Abiao
Most still ask things like how to do SEO, how to promote. Some also ask about good new product directions or ideas — mainly these two areas. As for changes, there don't seem to be any obvious ones.
🚥 Ronghui
Then what do you think they generally care most about, or what are they particularly worried about?
👦🏻 Abiao
The core is still how to do marketing well.
I think different products suit different marketing strategies. Many people don't clearly understand what marketing approach their product actually needs — they lack that awareness. Actually, SEO isn't a cure-all; not every industry is suited for SEO.
For example, take a product like e-signature. Want to do influencer marketing? How would you even approach that? It's a pure B2B product — it's hard to imagine what hook or angle it would have, and the content is relatively dull. If you do paid advertising, your competitors might all be heavily funded companies or publicly listed enterprises. They're all spending VC money, with deep pockets — several RMB or even several USD per click. In that situation, you simply can't compete with them.
So the moment you choose a product, its suitable marketing approach is basically already determined. If you don't choose the right marketing approach, you'll likely get half the results with double the effort and see little effect.
Another example: cross-border e-commerce isn't suited for SEO. The SHEIN founder came from an SEO background, but SHEIN's growth was mainly driven by paid advertising. So every product has its destined suitable marketing approach. Choose the right one, focus your investment, and play to its strengths. It's not that every domain suits every promotion method.
🚥 Ronghui
Do you have to say these things to basically everyone you meet?
👦🏻 Abiao
Pretty much. Actually, many people ask about SEO, and it's the same stuff over and over. So I say you might as well just ask me on WeChat, I'll tell you the answer, and you don't have to make a special trip — saves the hassle.
🚥 Koji
What advantages do you think building for global markets from China has, and what problems?
👦🏻 Abiao
The advantages are our engineer dividend, our diligence and hard work, relatively cheaper engineer salaries compared to overseas, and we earn in USD so there's an exchange rate advantage.
The downside is that for very B2B-oriented products, there's a certain disconnect when it comes to understanding user scenarios and experiences. And if we need to do localization or business development and sales over there, that's definitely not our strength. You can't just hop on a video call with someone domestically to do sales — you'd never sleep.
🚥 Koji
It's been almost five years since you started your company. Of all the decisions you've made in these five years, which do you think was the most critical?
👦🏻 Abiao
It's those times when I chose the right product, the right direction. Every product that failed or didn't take off — it was because I chose the wrong direction.
🚥 Koji
Choice matters more than hard work. What do you think is the biggest change that's happened to you personally over these five years of entrepreneurship?
👦🏻 Abiao
My perspective has definitely improved. When it comes to product selection and marketing, I'm more well-rounded now, my understanding is deeper, and my mindset is more okay too.
🚥 Koji
By that do you mean things that used to make you anxious don't anymore, or things you used to fear you don't anymore?
👦🏻 Abiao
Let me give you an example from my own experience. I used to think that if you were a good employee, you'd have higher standards for yourself and you'd work overtime a lot. Whether it was work or study, I believed work was never finished. So I used to want my colleagues to work more overtime, but I realized it's useless. Now our colleagues don't work overtime at all, especially on the marketing side — R&D does work overtime. I'm more zen about it now, my mindset is more okay.
Here's how I think about it: if I've chosen the right product direction, then whether it reaches a certain milestone a bit earlier or later doesn't make much difference. It's like whether the product launches today or tomorrow — the difference isn't that big. There's no need to pull all-nighters, to leave work at three or four in the morning. That's impossible at our company. If you've chosen the right product and the right direction, whether your marketing team works overtime or not doesn't seem to decide anything.
🚥 Koji
It's like skiing down the right slope. Faster or slower, you'll reach the destination eventually.
👦🏻 Abiao
Exactly. Whether someone hits a certain milestone a week or two earlier — if you're forcing everyone to work overtime every day to get there, I don't think that means much. Give yourself a break, give others a break. And if no one's working much overtime but I'm pushing people to do it every day, I'm exhausted and they're exhausted too. So forget it, give everyone a break, adjust the mindset.
🚥 Koji
What's the best decision you think you made in the past year?
👦🏻 Abiao
Since I came from an SEO background, I understood the ceiling and limitations of SEO pretty well, so I was really troubled by the problem of having only one skill. So the right decision was probably at the company level, building out our capabilities in influencers and advertising to round out our skill set and make us more comprehensive.
🚥 Koji
In your past work experience, including your entrepreneurial experience, whose influence on you has been the most profound?
👦🏻 Abiao
Recently, for example, the launch of Manus left a pretty deep impression on me. Because it showed me how Monica, after raising funding, brought in more talented people and produced a really excellent, very innovative product.
🚥 Koji
What's your MBTI?
👦🏻 Abiao
I've never taken it.
🚥 Koji
Aren't you curious?
👦🏻 Abiao
I don't believe in that stuff.
🚥 Koji
So you don't believe in any esotericism or psychology?
👦🏻 Abiao
I wouldn't say psychology in general. Things like horoscopes, fortune-telling, "Hunan people are like this," "Guangdong people are like that" — it's fine to chat about these things, but I don't believe in them.
I wouldn't go so far as to否定 them either, say they're worthless. I'm sure I fall into some framework or type, but I just won't do that kind of thing myself. I'm too lazy to take the test.
🚥 Koji
So are you an anti-authority kind of person? Like growing up you maybe didn't respect teachers that much, or after joining the workforce you didn't necessarily think everything your boss said was right? On this point, do you think you're pretty rebellious?
👦🏻 Abiao
No, I think I'm a pretty honest, straightforward person.
But I repeatedly emphasize one thing to my colleagues: I don't think our company has any authority, and I want them to voice different opinions. To some extent I do have a tendency to defer to authority, but at the same time I also have my own judgment, and I'll ask multiple people for their perspectives.
🚥 Koji
Actually, for the past ten years since you graduated, you've been working on products for global markets. In these ten years, has there been any particular moment that you look back on and find especially interesting? As someone who's lived through the complete journey of Chinese software going global.
👦🏻 Abiao
This is about traditional going-global versus AI going-global.
The inflection point was ChatGPT. Before that, when we were doing global products, there were very few Chinese entrepreneurs in this space — hardly anyone to talk to. It was basically the traditional players, led by Wondershare, a series of related or unrelated competitors, friendly rivals, or嫡系部队, or companies started by employees who left to do their own thing. That was more or less the market.
Then because of AI, it was like everything blooming at once — all kinds of scenarios and products emerged, and everyone started going global. And right now going global is also the prevailing trend, with many more players. So I feel this is a very typical watershed moment.
🚥 Koji
So the most interesting moment was the moment AI arrived?
👦🏻 Abiao
Because there are new scenarios and opportunities.
Before, I'd looked at tons of software product websites. Either the competition was incredibly fierce, with giants everywhere — though the ceiling was high too — or it was too small, and I was constantly frustrated at not being able to find a good direction. But AI tore open this gap, giving us more opportunities to try new products, putting us on the same starting line as some big companies.
🚥 Ronghui
Something I've felt in the past year or two — though it may just be my bias — is that many companies doing global products don't necessarily actually want to go global. They're just trying to avoid the domestic rat race.
👦🏻 Abiao
Yes, that definitely happens. But I think it's understandable. The same product, if you do it well, might earn 100,000 RMB domestically but $1 million overseas. Making money in China is really too hard.
🚥 Ronghui
Many people say entrepreneurship requires having your own DNA, your own determination, or that teams need certain specific capabilities. In this context, are these factors less important?
👦🏻 Abiao
First, I don't agree with this DNA or genes talk. I think this is determined by a founder's transformation and iterative updates. So-called genes are actually just an excuse or lie you tell yourself to avoid iterating or updating.
Take Alibaba — it didn't start out with Alibaba Cloud doing large models, and it didn't start out in e-commerce either. It iterated its way there step by step. ByteDance didn't start out doing global products either, right? It didn't have "global operations genes," so how did it do it? Through iteration. So I don't believe in so-called "genes" — this kind of talk actually limits you.
🚥 Koji
I have a small question. I saw you posted a job opening on Jike once. You mentioned that the office has a 90-minute lunch break every day, and then every afternoon from 3:40 to 4:00 there's a fixed 20-minute rest period for everyone.
This policy seems very unusual. Could you elaborate? Why specifically schedule 20 minutes in the afternoon for everyone to rest together?
👦🏻 Abiao
You're really observant, you even noticed that. It's actually because a colleague made a suggestion and we just adopted it, simple as that. We thought it was pretty nice, kind of like having break time back in school.
Now we feel this 20-minute break is genuinely good, because at that point in the day, people do get a bit tired and efficiency drops. Taking a break, people can chat and interact, go out for a stroll, walk around, exchange ideas — it feels nice.
🚥 Ronghui
Sounds like you really have become more zen, no?
👦🏻 Abiao
I don't think this is being zen. I think it's just being a bit better to employees.
🚥 Ronghui
What do you call it, having a good rhythm of tension and relaxation?
👦🏻 Abiao
I don't think I'm a zen person (laughs).
🚥 Koji
Alright, thank you again Abiao for coming on Crossing to share so much practical knowledge about marketing and SEO, and also to tell some entrepreneurial stories and insights you've probably never shared anywhere else.
I'm really looking forward to Pollo's new version in a month and a half. And I hope Abiao can come back to Crossing in six months, in a year, with updated stories, better products, and more observations and reflections. Thank you, goodbye.
👦🏻 Abiao
Great, thank you all too.
Subscribe to the "Crossing" Podcast
🚦 Crossing is Steve Jobs' metaphor for Apple — standing at the intersection of technology and the liberal arts, where great products are born. AI is bringing change to every industry. We seek out, interview, and bring together a new generation of AI entrepreneurs and active participants in the AI era. Together with them, we explore and embrace new changes, new possibilities.
👦🏻 Host Koji: I co-founded Jiepang / The Fair / Tangdao, and started AI Hacker House, a community space for a new generation of AI entrepreneurs. I believe technology, especially AI, is the greatest value-creation opportunity of our generation. Feel free to reach out to chat, bounce ideas, and connect on what's next. Koji on Jike[1], Koji's website[2]
👧🏻 Host Ronghui: I've worked at a USD VC and spent five years as a Silicon Valley correspondent, following tech development and business stories. Feel free to reach out to chat and exchange ideas. Ronghui on Jike[3]
References
[1] Koji on Jike: https://okjk.co/0JSUes
[2] Koji's website: https://koji.super.site/
[3] Ronghui on Jike: https://okjk.co/0cbnYV