How She Scaled an AI App to 3M Visitors/Mo Using AI SEO ft Chang Chen #ai #seo

Everyone, on today's show we have an amazing expert. We're joined by Chung Chen, who is a growth advisor. She's worked with Otter, AI Heygin, Microsoft, tons of companies. And what we're going to do today is going to be really valuable. She's going to walk you through her framework how AI can be used in marketing and growth. Like what are the core areas you should be focused on to get leverage today. And then we're going to do a deep dive on AI SEO and and she's gotten millions of visits for the company she's worked with from her AI SEO workflow. She's going to share that secret workflow with you. She's going to tell you how she does it and we're going to tell you the mistakes that you're going to want to avoid to actually get these results. We're going to do that and much more on today's show. Let's get into today's episode. Hey everyone. Welcome to Marketing against the Grain. We have a fun show today. We are talking about AI for marketing and growth and we have an amazing expert. Cheng Chen is here with us today and she's got some amazing content, especially around AI SEO that you're going to want to stick around for. One of the reasons we're really excited to have you on the show today is because you've got a pretty clear framework for how you think about the primary AI marketing and growth use cases and that you've done a ton of work. And maybe the best way to start for everybody watching the show today is for you to kind of kick us off and level set how you think about kind of the evolution of marketing and the role AI plays in it.

So we have been experimenting with all the different aspects of growth marketing and see how AI can be helpful. We tried with AI with localization. So far the translation is usable and we definitely have been seeing AI can help us to reduce the cost and help us to create content a lot faster. We also experimenting and start to see more and more traffic coming from AI native channels like AI search engines and AI answer engines. So far it's not a major channel yet. It's not anywhere close to like SEO traffic. But we do start to see an increasing traffic from them. And AI is an AI SEO definitely start to see significant success for several of the companies influencer campaigns as well. It help us save at least 80% of our costs and be able to help us to reach more micro influencers with a higher efficiency.

Where do you see Chung the earliest kind of success with AI for Growth marketing right now. What are the things that you're most excited for or spending your time on?

Right now we are spending a lot of time on localization, SEO and the content repurposing.

That's exactly how we're kind of thinking about it too. The localization has worked really, really well at HubSpot. We have a content remix feature for our customers that's in that content repurposing bucket that has seen huge adoption and a lot of success and great retention and everything. So that's a clear early use case. And when you say native channels, you're Talking about search, GPT, ChatGPT, you're talking about perplexity Claude and how they can be referral channels. You're like, hey, I don't think they're ever going to be big referral channels.

Right now we don't see them as major channels yet. So especially for the shallower content like Listicles, people may get some brand impression, but we haven't seen to become a major traffic source yet.

Kieran and you and I have gone back and forth where you're like, hey, do we think people are actually going to interact with the links in these AI answer engines or are they just going to get their answer and move on and it's going to be kind of an indirect awareness motion. What do you like, want to share your take on that with everybody? As we kind of are in the monitoring stage of these folks.

I think for the most part you can't expect to get a bunch of traffic because it just provides the answer to you. So it's really hard to see lots of people clicking on those, on those links. Like, I just don't see that actually happening that much. I don't see people clicking on the links that much. I think that for the most part the AI agent is going to do a much better job than the blue links did. And so we're probably going to see many people kind of create content and measure that for influence and impressions and actually measure it for clicks. I think that's the big change. Like, well, we used to create content for clicks and now we kind of create content for influence. And I think that's the, that's the change I see happening.

So another AI native channel we are monitoring very closely is the AI companions. So right now, so for example, that I'm using one of AI companions to help me to plan many of my trips, even though the AI companion most of the time won't provide direct links. But I started to trust more on the recommendations because companies already trained on my preference. So most of the time the recommendation is very personalized.

Yeah, exactly. Yeah. I think it's really hard to attribute. It's kind of indirect attribution and because.

That'S indirect attribution, I think one of the things that you're saying you're spending a lot of time on and seeing a lot of value is using AI for the traditional SEO motion for Google search to try to capture a higher percentage of traffic there kind of early in this transition. Given that the answer engines haven't scaled up to be massive consumer engines just yet. We still have to really focus on the Google product and what we're able to do from SEO there. And so maybe walk us through like how you think about it, what you found to be successful, what folks can do on the AI SEO front.

Yeah, so for AI SEO, we are definitely seeing success. I think this is one of the first application for AI because the generative AI technology is really good for copy creation and it can be treated as the tireless writer in your team that if you ask the writer to create a thousand piece of content or ask the AI to rewrite a thousand piece of content, they will never complain and it will always deliver and the renter will be available 24 7. So we started to see a lot more success so far for one of the products we are seeing. We're able to drive the traffic to 3 million monthly visiting just six months. And even with all the Google updates, we haven't seen much traffic drop yet. And we will do 2x am month in just 6 months. And the content creation cost just has been slashed. But right now we're being very cautious to experiment and use AI only content. Because if you think about AI for content, AI for SEO there can be three categories. There's AI only content that you give AI a prompt and they will write 100% of the copy or in the traditional way, there's a human created content and we have writers and we have editors to help her to ensure to create a high quality piece of content with several steps. And then now we are experimenting and try to mix the two. So we still have human in the loop and to be able to follow almost the same writing process as if that we still have the human to help us to create the content, but we use AI to help us to streamline and to expedite some of the steps. The process that we are using right now, we still start with keyword and previously we have human to do Research and to find the top search results. But now we have AI to do the scraping and AI to help us to summarize them. After that we have the human review to distill the learnings and to identify good parts of content that we should add. And then we have AI to take the human input as well to learnings to create article brief and then we have human review again to make sure that everything looks right. And the next part we have been experimenting the most, which is how AI can learn and take some of the unique content from our product, from our brand, from our user as well as our human writer to really be able to add information gain and to create that unique content. And with all the learning information, then we still have AI to draft the initial draft, but we always have the human which is our editor and writer to polish it, to make sure that it's on brand, to make sure it's a high quality. So far we find this mix of human expertise and AI employee with the AI writer to really be able to improve the efficiency without sacrifice.

Yeah, can you maybe explain the first step? What does it do? Can you just take us through what that first step is? It summarizes the top SERP results.

They will come to Google or Bing and they will search for the keywords that we target and they will find the top results and research what are the different components the top results writes about. They will also look at the people also ask section and trying to find the relevant longer tile keywords and the longer tile questions that we can include and cover to make sure the content is comprehensive. And some of the key learning points they will also document like some of the industry benchmark market data points so that we can reference them in our article. And for the AI step that we have is we have AI connected to Google search and scrape data and to learn the key points and to memorize the key data points that we should reference so that we can have backlink so we can have reference links added to our own article.

What tools are you using to do each of these stages? So that I know that's one of the first questions we'd get in the comments is like great, I understand how I can create a lot more search content that's better with AI. But like what's tactically, how do I go and do that?

This screenshot is. This screenshot is using L Ops. So it's a workflow automation platform that can help easily to connect different components. They have a native component for Google search. Then you can add your own Prompt to TCAI how to summarize the search results.

Is this flow for blog articles landing pages? Can you give us more specificity on the types of pages and tactically, how do you think about taking a workflow like this and like doing it at scale? Because if you're going to grow millions of extra visits every month, like you obviously have to get pretty high scale and I'd love to just kind of walk through like how you do that.

This specific flow is for a informational blog. We also have a different flow for listicle articles, for how to articles and for landing pages we also have a more programmatic ways because for program analyzer pages that it has the structure like section one is talking about this topic, section two is talking about a different topic. So we are able to use one workflow and a list of keywords because each page will have a different keywords then turn into like a table, then each row will represent the content for one page. And we use webflow so we can just upload the CSV and be able to create landing pages at scale.

So you're basically for landing pages, you are taking your workflow, running that, running your landing page keywords from that and just basically spinning those up within webflow or you know, most CMS products would do something similar here. And for those landing pages, are those mostly like content focused landing pages or different derivatives of product focused landing pages so that people can kind of understand, are you trying to take different product offerings and verticalize them and specialize them or. Or are you trying to do something different?

So for this one, which is AI video app, it has all the different features and for one feature we actually identified 10 plus different keywords that we can target to create one template landing page and then use the same workflow to create different versions of copy for the same feature so you can target it for different keywords.

Basically what you're saying is like, hey, instead of the old way is either have kind of one or two product focused landing pages that are more broad or to have a person who manually goes and creates like a whole portfolio of specific landing pages. Instead you're like, hey, I'm using AI to figure out what keywords are best for this particular product and then I am going to go after those keywords by having AI through my kind of programmatic prompt and spreadsheet, generate the content and the pages automatically and then a human's going to go and review and publish. Is that correct?

Yeah. And the key is to make sure that you actually have something unique for each one of the pages, you can pull in different unique blog articles that you wrote previously, or you can pull in different user reviews, but just to make sure that has the unique piece of content so when the user review data won't see us a duplicated content.

Hey everyone, remember that episode we had with Dharmesh? Well, Dharmesh has exclusively gifted us access to agent, AI's company research agent, just for all of you. So if you're looking to spy on your competitors so you can get a deeper, deeper dive into their strategy, this is the agent for you. You'll get access to their key demographics, their web traffic trends, and you'll stay ahead of them with all of the competitor analysis tools. So if you want it, you can grab it at the link in the description below. Now let's get back to today's show.

So it takes the keywords, it runs, AI creates all of the articles. What percent of a complete page do you think the AI does versus like what how much does the human have to do? So is it 50% of the work is done by the AI and then the human has to add the 50%. And then the second question would be really interesting is like, what are the things the human is doing that is making these pages much better? Like to rank or is there some commonalities and the stuff that they add that works really well depends on the pages.

If, for example, if they're saying product feature but we want to target different keywords, then that part it's a lot easier because most of product description and most of the product positioning part can be reused. But we use AI to rewrite. But if it's a blog article like this one, then it takes a lot more human review and takes a lot more personal experience. And to have our editor's perspective, this one, I would say probably 50% or even sometimes more than 50%. But if it's for expansion, then it's more automated.

Okay, and what do they typically add to the page to make it rank really well?

Most of the time it's the unique content. I think I have a chart here because right now that we do see more and more people starting to scale their content creation with AI. But at the end of day that we are still the same Internet, we want to have information gain, we want to be able to stand out from the competition. The unique content, the unique perspective is still the key. Right now we start to identify more ways that we can have unique content sources that can be also an editor's input. The product messaging is still very important because all the SEO is for driving traffic and to have conversion and to have brand awareness. So we want to make sure that the AI will still be able to introduce our product in the proper location in the right way. Market research data is also something we started to embed and started to insert to our articles because our unique perspective of how the market that we are in. So it's a way to help to educate the potential users about how the market is progressing and how our product positioned within the market. User generated content, like a user review defined user discussion in our own community. Those can be really, really good content to help us to bring out that unique perspective and the subject matter expert insights we have on YouTube interviews or broadcast like this to bring in the subject matter experts insights as well. So those are all the unique content sources that we are experimenting to add to our product so it can run high. Other than all the traditional on page SEO stuff, the on page SEO stuff AI is already really good at. But I think the unique content piece is still something that AI cannot do by itself.

That makes a ton of sense. Have you found these like what is the performance been like for the pages that have been created by AI and humans? Has it been on a par with what humans have done? Have you found the same sort of ability to rank those pages to get traffic from those pages?

We do see most of the content we read as performing on par with the traditional human owning content. Because right now you can see that AI is not replacing anything yet. We still need a lot of human input, but it definitely helps to save time.

Yeah, so it's basically an ability for the human to do more versus.

It'S.

A productivity gain versus being able to rank things better. It's just being able to create more content.

And the most important part for the team right now, feel like AI is taking over the part that human does not like to do. No one likes to click on the top 10 links one by one and then scan each one of them one by one manually. Is that you can just do that with one click. So you're taking over the repetitive and the energy draining part so the human can actually focusing on the creating part.

Cool.

This is a very cool workflow.

And just to clarify for everybody watching, what we're doing here is you're creating the content with AI and obviously all these pages are trying to convert these visitors, but using AI to basically get the traffic and then all the calls to action, everything those are still just at the template level. So as the pages are spun up like the landing pages and everything, the forms or call to action, everything are just built into the template. You're not doing any kind of conversion personalization with AI on these pages, is that right?

We are seeing some agents on the page like chatbots, yes, that can start to do some of those, but this workflow is not that.

Got it. So if you're a company out there and you have kind of a limit, a limited footprint on Google especially for like kind of blog content or you're you have a pretty wide product offering but aren't getting a lot of product traffic, don't have a wide like portfolio of product pages, you would use this workflow to basically increase the traffic. And then what you're saying is the number one recommendation for converting that traffic would be to have some type of AI kind of chat agent to that folks can interact with, that can qualify and either route to a human if it's a high consideration purchase or kind of facilitate an E commerce transaction if it was like a lower considered purchase.

I will say we do start to see the conversion gain if you apply a personal assistant or a personal agent to help the shopper to explore your site and to help a user if you are a SaaS product, to understand more about the free product offering and to handhold some of the early questions. The AI agent works better than the traditional tutorial, than a traditional rule based surveys, than a traditional rule based onboarding flow.

I do think that's one of the most important points that you've made today. So that I want to emphasize it for everybody kind of at the top. You walked us through that. We've went from this kind of like broad mass market marketing to this second generation of data and segmentation focused marketing. And what's really happening with AI is that all that work we used to do to create segmentation is becoming way, way, way less important. Because AI is going is able to do that on a one to one basis instead of a broader segmentation basis. And whether you're talking about in the case of today's show, AI SEO and how you're trying to basically have a lot of variations of landing pages or a higher density of blog articles to speak to more detailed versions of your core kind of buyer, that is really one of the things that AI is great at facilitating and gives us the next generation of segmentation. Am I getting all that right?

Yeah. So I have been spending most of my growth career in creating segmentation to bring out that personalization. But now I do see the manual segmentation was actually created based on experience, based on some data modeling. And we also have been iterating like many of the times when we first create a segment, it may or not be accurate. So now that we're actually just skipping that step to rely more on the AI agent to understand the user on the one on one level and provide that proactive as well as super personalizing experience and then be able to engage customers in the exact way that they want to be engaged.

To kind of close out today's show, what would be super helpful is somebody's watching this and they're like look, I've been doing a bunch of stuff with AI, but I haven't been able to get millions of more visits or drastically change the curve of growth for my business. Like what would you recommend for them as like you gave them the framework, but what should they go to like go and do to practically get started to go from kind of zero to one on this type of work. When it comes to AI SEO, we.

Do see more and more people creating tons of pages. They can be pages of a blog, they can be pages for landing pages without getting much traffic gain. Most common mistake we notice is the ignorance of the keywords and the user intent behind the keywords. Why I say that because many people that they will start with looking at the keyword that they can the keywords from their competitors and the keywords that they potentially can go after without looking at the potential difficulty or look at the existing website situation. So if it is a new website and they are trying to rank for super high traffic and super difficult keywords, then in the very short time there's not much chance that you can. And also another common mistake that we observe is ignoring the intent. So some keywords is informational, then it's more suitable for blog and for some keywords it's actually more transactional people actually looking for do something then a landing page or SEO tool will be more suitable to meet that intent. And the third most common mistake that we have been observing is the content quality. Because AI is really powerful. You can just give AI a prompt and it will be generated a lot of text for you. The longer text doesn't mean that it's a higher quality content. The longer text doesn't mean that it's a unique content. The longer text doesn't mean that it's a content that can rank. And the AI is not a game that the more the better. Most of the time that is actually the higher quality the better. So instead of creating tons and tons of low quality content that we highly recommend to really examine your workflow, to really take a look at the quality of the content to make sure that it will actually rank before you continue.

To create more to kind of close out.

The biggest mistake to avoid when it comes to AI SEO is like don't go and create a mass amount of AI generated content with LLMs that no human is working on improving. Like the quality, unique perspective, detail of information to make it really valuable and unique.

Right?

Like it's just going to be generic LLM content. People aren't going to go to it and you're just going to be spending a bunch of time for no results slash, maybe results that might even get you penalized in a search engine like Google. Is that what everybody should take away?

And I always advise don't create content when you don't have anything unique to add.

I love that. That is great advice. This framework around kind of the core use cases of AI and then how to actually get started and drive real impact with AI SEO has been awesome. If you have any questions, any follow ups on this, please put them down in the YouTube comments below and we'll see you all real soon. On the next episode. Marketing Gets the Green this data is.

Wrong every freaking time.

Have you heard of HubSpot? HubSpot is a CRM platform where everything is fully integrated. Whoa.

I can see the client's whole history. Calls, support tickets, easy emails, and here's a task from 3 days ago I totally missed HubSpot grow better.