Optimize Episode 039: Alex Ross on Programmatic SEO, User-Generated Content, and AI-Generated Content

Learn More About Alex Ross

LinkedIn: https://www.linkedin.com/in/areteross/ 

Twitter: https://twitter.com/AreteRoss 

Website: https://greg.app/ 

Episode Transcript

Nate Matherson (Speaking) Alex Ross (Speaking)

00:00

When I think about programmatic, but I think the biggest question is, do you have a database that has original information in it? Cause that's where we start, is we have for our product, a plant database. Imagine we've got like 50,000 plant species in it. So that database that we have of plants that are relevant to people is where this all begins. So if I was thinking about a new programmatic strategy, I would ask, do I have a database of information that nobody else has, right? Cause that's kind of the value we provide to Google is we understand more about how plants relate to each other and how people relate to their plants. And so the whole game for us is just figuring out how do we package that information in a way that is useful for the users of Google, right? Because then they search and they find it.

‍

Nate Matherson (Speaking)

00:48

Hi and welcome to the Optimize Podcast. My name is Nate Matherson and I'm your host. On this weekly podcast, we sit down with some of the smartest minds in content marketing and SEO. Our goal is to give you perspective and insights on what's moving the needle in organic search. Today, I'm thrilled to sit down with Alex Ross, CEO and co-founder at Greg, an app that helps people grow healthy plants in home and in the garden. Think AI, but for plant care. Before that, Alex was the director of engineering at Tinder and the co-founder and CTO of Enplug,
a startup which he grew to over $5 million in ARR and later exited. Today's episode is a fun one, and I'd argue one of our most unique episodes. We chat programmatic SEO, AI-generated content, UGC, and we unpack all of the amazing work that Alex and his team are doing on the SEO side of things at Greg.

‍

Ad Spot:

01:45

This week's episode of the Optimize Podcast is brought to you by Positional. My name is Nate and I'm one of the co-founders of Positional. We've been working on Positional for about 10 months and we've built a handful of what I think are pretty awesome tools, including we've launched Content Analytics. Content Analytics is kind of like a heat mapping tool, but for a content marketing and SEO team. We provide really granular insights into where users are dropping off within your pages. We've actually just launched a couple of new capabilities too. We've launched click mapping and click tracking to give you better insights into where your users are clicking and converting. We've also launched a more general heat mapping view too, alongside our read maps. We'd love for you to check out our entire tool set at www.positional.com. 

‍

Nate Matherson (Speaking)

Alex, thank you so much for coming on the Optimize podcast!

‍

Alex Ross (Speaking)

Thank you, Nate. I'm a big fan of the pod. I think I've listened to most of the episodes at this point. I got in and then I went back and back and back and back. And it's been a huge inspiration for what we're doing. So I'm excited to be here.

‍

Nate Matherson (Speaking)

Do you have a favorite episode?

‍

Alex Ross (Speaking)

I do. There was one episode, I don't remember the guy's name, I'm terrible with names, but it was the episode where he said that seven internal links, on the median, is the optimal number for inbound to an article. So I just remember that guest was just so knowledgeable and he was very good at explaining the concept. So maybe you can put in the show notes which one it was, but it was fantastic.

‍

Nate Matherson (Speaking)

Yeah, that would have been Ethan Smith from Graphite. And yeah, they had that amazing report that they put together on the impact of internal linking. Everyone that listens to this podcast knows that internal linking is important and hopefully something we should all be doing more of. But thank you so much for coming on. This is actually the first episode where the first question I asked was not, how did you get into the world of content marketing and SEO? So tell me Alex, how did you get into SEO? Because I know it didn't start with Greg, but you've done some amazing work at Greg so far. 

‍

Alex Ross (Speaking)

For sure. I've been floating around SEO as a mediocre practitioner for, I think it's maybe 13 years or something. But in 2010, I got my first job as a web developer and I joined a commodities future brokerage. So these are the guys that trade corn and oil and beef and stuff like that. It was a boutique firm and they had a website, canontrading.com, probably still around. And I had to get them ranking for like futures trading. It was a short tail term. So I got into it and started doing SEO for them and got them ranking,
got some organic search traffic. And I think that was around the time that like Panda came out. I remember it totally knocked like so many things off the rails. And yeah, so I got exposed there. And then my company, Enplug, we also got in SEO and we produced a big digital menu series because our product was all about like digital menus and restaurants and stuff like that. And so that worked pretty well. And then really, Greg is where I'd say I've gotten the deepest into like technical SEO, large scale, trying to get to like millions of visitors, that type of thing.


‍

Nate Matherson (Speaking)

Futures trading, that's a really tough keyword and also a very broad one. I'm curious, did you ever get to the first page for futures trading back in the day? 

‍

Alex Ross (Speaking)

We were super proud. Yeah, we got up into number eight or nine. So it wasn't high enough. You know, we probably got like 2% of the voice, but we got a lot of long tail terms. So there's a lot of like,
you can imagine corn futures and stuff like that. And the further and further you go, the more specific we got and did pretty well. 

‍

Nate Matherson (Speaking)

And this was like a, like an investment bank or a trading firm that wanted to like acquire potential customers to trade futures with them? 

‍

Alex Ross (Speaking)

That's right, it was a brokerage.
And so there was 10 brokers, totally Wall Street style guys hustling on the phone trying to get you to like, oh, there's a great deal on 100,000 corn futures expiring this date and the weather is changing in this way and so the price is going up and we read the candlestick graphs, all that kind of stuff. And so they wanted to acquire leads.
It was a lead generation SEO strategy and they wanted to acquire leads to that trade and collect commission on those trades. 

‍

Nate Matherson (Speaking) 

Well, you were probably their favorite customer acquisition channel or one of them. I'd imagine if they were doing a lot of outbound sales on the phone. I want to talk about Greg. You know, you and I have gotten to know each other pretty well over the last few months. Thank you again for being one of our earliest customers in our private beta at Positional.
And the growth that you're seeing has just been so impressive. You know, from my view, it looks like the website, greg.app, has five to six X organic search traffic over the last year. But recently it looks like organic search traffic has two to three X in the last three or so months. Is that fairly accurate from the outside looking in? 

‍

Alex Ross (Speaking)

It is, yeah. We had two inflection points.
One was this last summer around June. I can explain what caused that. It wasn't actually direct SEO work, but it was interesting. And then most recently, we put three months head down into building a new programmatic system, and that is accurate. It has led to roughly a three to four X growth in organic click-throughs and growing. So we're like hand on throttle going faster. 

‍

Nate Matherson (Speaking)

I wanna dig into both sides of that coin. So tell me what you did over the summer that wasn't like an SEO initiative, but had a pretty positive impact on organic search rankings.

‍

Alex Ross (Speaking)

Yeah. So I don't know if this is correlation or causation, so I'll just call it out up front. Our first successful growth channel has been partnering with plant retailers. So most people buy plants from Home Depot, Lowe's, Walmart, places like that. So we went out and partnered with the companies that grow their plants, because Home Depot does not grow plants, but they do sell hundreds of millions of plants per year. So we partner with the farmers who grow the plants and then we put our little QR code on their labels so that when you're in a Home Depot, you buy a plant, you bring it home, you're trying to figure out where do I put it, when do I water it, you scan our QR code, it takes you to a web landing page and that web landing page takes you to the App Store. So what we noticed is when we rolled out this product, we started getting like 10,000 visits a day roughly, and our search traffic, like pretty much same time took a step function up. And so my perception is that this has something to do with Google knowing about that traffic in some way. I don't know how exactly, but they then kind of interpreted that as like brand traffic.
And so I think they kind of think that, okay, this app Greg, kind of has this external validation of like people are finding it and using it through these other channels. So therefore it must be more trustworthy. And so I think it almost serves the same function as a backlink, for example. It's basically a trust signal to Google. 

‍

Nate Matherson (Speaking)

Yeah, I'd agree.
I think Google's gotten very good at being able to identify brand and then also like traffic from other channels that isn't organic search as being a really positive ranking factor to then correlate off of. And you might remember one of the earliest guests that we had on the podcast mentioned that every time they would go viral on TikTok, which is a channel that's not organic search, they saw a directly positive impact then
on their organic search rankings. And so I don't know how Google does it, but yeah, I would agree with you. They've gotten very good at being able to identify that traffic and then see your brand as hopefully a pretty good source of information about other related topic areas. But that's such a fun hack. You got your little card stuck in the plants. Is that the strategy you're still using today? 

‍

Alex Ross (Speaking)

For sure. We love it because imagine like that's the best moment to discover a plant care app. Well, you're at home and you're like, what the heck do I do with this thing? And so it's kind of, I think with any business, you want to reach your customer at the exact moment that they most need your product. And so for us, that's completely ideal. The unit economics are also great because we just modify their existing labels. And so there's no cost to print the label. And then finally, it helps us build good relationships with these farmers so we can get good plants back to our consumers. So yeah, it makes a lot of sense for us, but it's difficult to scale.

‍

Nate Matherson (Speaking)

0:09:32

I do want to talk about programmatic SEO because so many startup founders, they come to me and they're like, Nate, we want to do programmatic SEO. And I'll say to them, I don't understand how programmatic SEO could be relevant at all to your business. So let's focus on traditional SEO, at least as a starting point. But programmatic SEO has this allure to it, where people tend to think it's this very sexy, high growth, immediate results type channel. And I guess it could be, but I would love to understand your programmatic strategy a little bit more. What work have you been doing on the programmatic SEO side of things?

‍

Alex Ross (Speaking)

Yeah, definitely. So I think, when I think about programmatic, I'd be curious if there's a counter example to this, but I think the biggest question is,
do you have a database that has original information in it? Because that's where we start, is we have, for our product, a plant database. Imagine we've got like 50,000 plant species in it. And for context, there's around 400,000 known species of plants to mankind. And then there's maybe like 10 or 20,000 discovered every year right now in like the Amazon
and places like that. And so there's efforts underway to categorize all plants. So that database that we have of plants that are relevant to people is where this all begins, right? Is we had, again, 50,000 species. So if I was thinking about a new programmatic strategy, I would ask, do I have a database of information that nobody else has, right? Because that's kind of the value we provide to Google is we understand more about how plants
relate to each other and how people relate to their plants. And so the whole game for us is just figuring out how do we package that information in a way that is useful for the users of Google, right? Cause that may search and they find it. And so our journey, we've actually had two major iterations on the programmatic SEO front. The first is we naively published a plant care page per species. So every single species in the app has one page on the web. It's like greg.apps slash plant care slash mums, for example. And that page was mostly aimed at water, light and fertilizer. So like how much do you water it? How much light does it need? And do you need to fertilize it? And if so, how much and what types of fertilizer? And so that got us to around 100,000 visits per month from Google. So it did like pretty well for sure. And
it was, you know, maybe, I don't know, 10% of our business like acquisition efforts, but it didn't really scale out very far. And pausing, we enriched those pages, not just with our like scientific information from the database, but also with user generated content. And that's really where I think programmatic starts to fly. But Greg, the app has an in-app community that we built that is like kind of like I would compare it to Reddit most, I would say a lot of question and answer. And the key is that when people post questions or answer questions, they tag the specific plant they're talking about. And that tagging of information and content is really the critical leverage point with UGC. And so we're able to say like, okay, all these images, we have millions of images of plants, all of these ones
have to do with that plant type, or all these questions have to do with that plant type. So those single pages were just these very long, very long scrollable pages that address those three topics, as well as like questions and images about that specific plant species. So that was iteration one. 

‍

Nate Matherson (Speaking)

And just so I understand it, for all of our listeners, you've got a massive data set, both in terms of like plant care and types of plants that exist in the world. And then you're enriching that data set
with your own UGC data that's created by users on your platform, similar to something like a Reddit, like you said, where your users are basically questioning and answering one another about the different care or how to take care of some of these plants. And so it's your programmatic pages, I guess, have two different data sources that they leverage to create them, is that right?

‍

Alex Ross (Speaking)

13:24

Yeah, spot on. Okay, so that was the first iteration. And you said that got you to about like 100,000 visitors a month. What's like the second iteration that's unlocked more growth? Yep, so the first iteration, like I said, was naive. Like we just said, we have these species. Well, it's just one to one publish
a page. Right. So then we said like, okay, let's, let's think harder about like, what is our competitive advantage here? And something like really is two essential insights led to growth. I think the first was that there's a lot more topics about every plant than just water, light, and fertilizer. And so I began this whole initiative where I went in and did the keyword research to figure out what are all of the different things people talk about with regards to plants. And it turns out there's like 26 different topics that we know about today. And so these include like browning leaves or leaves dropping or root rot, you know, the roots get all moldy and stuff or direct sunlight. Toxicity is a big one. Toxicity specifically to cats. I don't know why. I guess cats eat a lot of plants, but toxicity for cats with regards to plants is like four times as searched as dogs. And humans, like not even barely above the radar. So imagine we worked through all the permutations of like, wow, it turns out people are searching a lot more about plants than those specific top three topics. So that was insight number one. Insight number two was that we could rank better if we published very targeted pages just for those search terms. So I don't know how replicable this is, but I have a personal theory that the internet is just constantly moving towards more and more convenience. And so I think people like being able to search and get a very specific answer to their question. I think that's why Reddit and Quora do well, for example, is you get very hyper-targeted results. You don't have to wade through all this other content. I don't know how replicable that is, but basically we then started publishing, instead of one page for all the topics, we started publishing one page per topic per plant. So we have like a mom's toxicity to cats page, for example.

‍

Nate Matherson (Speaking)

So just so I can recap for our listeners and so that I understand, it sounds like you went from in the first iteration having like three different modifiers to like a specific type of plant to now what is like 26 different modifiers to each specific type of plant. Like you said, going from just watering to now browning leaves, toxic for cats,
sunlight, direct sunlight. And so it sounds like we've like multiplied the number of pages by maybe like eight X. More. Going from let's say like one page or three pages to now 26 pages for each variant of plant. So is that essentially what happened? You said more there.

‍

Alex Ross (Speaking)

16:05

Spot on. More in that, like I said, those core three topics, which are the most important ones, like water, light, and fertilizer do get the highest volume search terms, but basically those core three topics were all on one page. So it's not that we had a page for watering and a page for light and a page for fertilizer. We had one page and still have that one page actually. And then there's like jump links. You can jump down to the water section, jump down to the fertilizer section. And we noticed that a lot of people were using those jump links. And so it turns out people want the very specific answer, if that makes sense.

‍

Nate Matherson (Speaking)

That does make sense. And I think like over the last few years, Google has gotten very good at surfacing very long tail pages to very specific queries, which was not the case, let's say back in 2014. Like back in 2014, we used to write like these massive 6,000 word mega guides and try to rank those pages for every keyword within a category space. But today, it seems like the longer tail you can get while still avoiding cannibalization and still, I guess I should say, like still creating a page that deserves to exist, the better it would rank. Totally. And it sounds like in your case that you've seen data that would suggest that this is also a better experience for the people coming to your site. And so I think Google can probably recognize that too. And as far as content creation goes,
it sounds like you’ve created like hundreds, thousands, tens of thousands of pages, I'm not sure. So I guess the two-parter here, how many pages have you created using this programmatic strategy? And then as a second follow-up to that, how are you actually creating these content pages? 

‍

Alex Ross (Speaking)

Yeah, definitely.
So, so far we've published 30,000 articles. And so that is, I think, you know, whatever, it's a couple of thousand species, maybe like 1500 species, and then roughly 26 pages per species. So 30,000 pages total. And the other major thing that happened, and I know everyone's thinking about this,
but what made this possible, is now we have a new tool for scalably writing the content, right? And so early on, we actually really got stuck because some of these pages need to have text on them, you know, like you need to be able to describe what's going on with like water and light and stuff like that. And we had some very naive, like NLP or natural language processing attempts at this. We were doing too many to write them by hand, right? But with, you know, ChagGPT 3.5 and 4 coming out, really it was 4 that made the big difference. We now had a way to draft like anywhere we need text, we could feed it all of our context
about that plant. And we wrote a very large, I'm talking like 10,000 lines of code, prompt engine for reliably and with high quality writing, like very specific, very well-formed sections of content for each like type of topic, whether it's toxicity or leads.

‍

Nate Matherson (Speaking)

You know, oftentimes I'll talk to startups and they'll go to ChatGPT and ask it simply to write 2000 words on insert keyword. But it sounds like you've got a 10,000 line prompt
that you're feeding to ChatGPT and you're using your own unique data sets to empower the creation of that content, including your UGC data set. And so I'm guessing that the output here is actually very unique. Like my problem with AI generated content is it tends to return Wikipedia pages. And that's not that helpful or engaging for most queries. But in your case, given that you're feeding the input as your own unique data source, I imagine that the output is actually quite good and quite helpful. And you mentioned that GPT-4 was the unlock there. So between GPT-3.5 and 4, that's where you noticed the biggest difference in terms of quality that you got back. It was a major difference.

‍

Alex Ross (Speaking)

Yeah, it did. It was the unlock, very binary. Like now this is possible. And specifically, it's that ChatGPT-4's reasoning abilities became better. So it was better able to wrangle the context we gave it, and better able to follow very specific instructions. I think today we're doing a lot of work with LLMs, and I actually think they are like dramatically underappreciated in terms of what they could do. It's really just that it takes a lot of time. So to your point, I actually did
start by going into ChatGPT and just drafting entire articles, and very quickly found that that does not work. And really like there's a couple of tricks I can explain, but one of them, for example, is that you need to focus it onto the smallest little piece of detail. You can't, there's something going on with the attention span of these LLMs, where when you give it too big of a task, it's like its brain gets like scrambled and it kind of like loses attention. I even noticed this. We do a lot of work with AI generated imagery as well. And it is a weird thing. But let me tell you, I will end up with these beautiful illustrations and then the face of a person in that in that whole beautiful imagery is just like scratched out. So it's like the LLM got to the point where it's trying to draw the human face and it just like gets totally frizzed out. It loses attention, it can't draw the face. Even though in the background there's all these like articulate buildings and like so much detail, but there's a point at which the LLM loses attention and that's where you get into like hallucinations, that's where you get into like just nonsensical like reasoning. And so very quickly we learned that we have to narrow it down so that it's almost like an assembly line where like the LLM is focusing on this part and then on that part and then on that part. It's almost like you know we don't actually have one LLM doing this, but rather our prompt engine is really a coordination framework where it's like, we're almost assigning work to a hundred different LLMs. And each of them is like, it's only a Ford factory. It's like, you're working on the headlight and then you're working on the glue for the door. And at the end of that, you end up with a really good product.

‍

Nate Matherson (Speaking)

21:56

And you've taken it to the extreme of prompting chat.jpg to write these pieces of content and I love it. And one of my other big complaints with the AI writing tools or AI written content more broadly is that anyone could do it. Anyone could go to a ChatGPT and ask it to write an article on SDRs. But in your case, given that you have access
to this unique data set, the content that you're creating could not be created by anyone else. I find that really, really compelling, and it sounds like you're using AI-created content in a very artful way. Given that you've published tens of thousands of these pages, how much manual review needs to go into each of these content pages? Or is that just not possible at this type of scale?

‍

Alex Ross (Speaking)

Definitely not possible. We approached this from the perspective of we need to front load that QA. And so what we did is we did do, and what most people would not have done, is hundreds, if not thousands of iterations while building the pipeline, right? So like, this is just me sitting there and just reading output after output after output. We'd run a batch, I'd read it, iterate, read it, run a batch, read it, iterate. So I think like, you know, other people have commented on this, but I think the unique thing about like this era of AI is that it puts the focus onto taste. So I think I have like a pretty refined sense of like what I like, you know? And I'm also pretty like intellectually honest where like I'm not gonna put out spammy content. Like if I, my bar is if I don't wanna read this and if I don't find myself smiling and enjoying reading this, why would anybody else? You know?
So I think a lot of people, I have seen some like spammy stuff put out by LLMs and like I read it and I'm like, this sucks. Like why would I, why would I put my name on that? Whereas these articles, what's interesting is the most surprising thing is that they've got a little bit of a humor in them. ChatGPT is funny.
And so it's like very odd, but like I'm reading these and like the puns are really good and I find myself wanting to read it. And so that's resulted in an output that is higher quality than a lot of our competitors. And I don't know if that's just the space, but I would say there's a lot of like really bad content about plants on the internet. And so our first bar was like, just be better than what's out there, right? And then eventually build up to like what really we think is amazing, right? 

‍

Nate Matherson (Speaking)

Yes. And I was looking at one of your competitors after one of our calls the other day, who will not be named. And I think like the bar is quite low in terms of quality for some of these keywords. As far as where we go from here, like given you've created like 30,000 or so of these pages and you're using like, let's call it 26 modifiers to create pages within each plant topic space, are you planning to increase the number of modifiers? So maybe go from like 26 to 48, or are you planning to just increase the total number of like BCs of plants as the starting point for those modifiers?

‍

Alex Ross (Speaking)

All of the above. Like, you know, it's really cool because this strategy just fits into what the point of our company actually is.
We started this company with the goal to be the plant information company. So really we want exhaustive coverage of every plant on planet earth and eventually other planets. And we want every topic to be covered for every plant. So I think like mission-wise, even if there wasn't an SEO angle,
yeah, even if there's one person searching that plant per month, we want to be there, right? We want to be the go-to known reliable source of high quality, accurate information, scientific information. In plants, there's a lot of not scientific information. And so that's kind of like our distinguisher is our team. We have botanists, I'm an engineer, like, you know, it's scientific. So I think we have line of sight to like probably two to three hundred thousand articles. And then from there, we will look at, yeah, bringing in more species because there are, what's interesting about plants is that every year there's like, there's like 2000 new orchid varieties produced every year genetically engineered. Farmers are genetically engineering plants every year because that's their product, right? I like to compare plants to sneakers. There's GOAT and these other marketplaces, but people get into sneakers. Every year, there's a new Air Jordan or whatever the trendy thing is this year. Same thing with plants. We have to keep up with that. There is a new, renewing source of species that we need to keep publishing information on. I'm sure that we missed topics. Some of the topics were very surprising to me. It turns out that people are searching about the benefits that specific plants have a lot more than I would have thought. So it's like money tree benefits was like a surprising keyword, type of term that is just not served at all by the existing pages. So we were able to fit right in there and pick it up pretty quick. So I think we will also look for more, I call them dimensions or permutations of like the pages. 

‍

Ad Spot:

We're about halfway into this episode of the Optimize podcast, and I just want to bring you a special note from one of our sponsors, that being Positional. If you're anything like me, you probably love internal linking, but you probably don't do enough of it. And it probably takes a ton of time to find missing internal links throughout your old pieces of content and then internally link the new pieces you create. And that's why we built internals. We'd love for you to check out our internal linking tool set at positional.com. And that was a word from our sponsors.
Now back to this episode. 

‍

Nate Matherson (Speaking)

Well I hope there are some benefits of having a money tree. It seems like there are some obvious ones. I want to double click on the UGC side of this strategy, because we've seen some changes in the SERPs over the last few months. When you look at sites like Reddit and Quora that have 2, 3x their traffic from organic search over the last few months, it's very clear that Google is trying to prioritize or at least diversify their SERP results with additional UGC-based data sources
to provide maybe searchers a different perspective or maybe longer tails of information that they otherwise wouldn't get. And I love that UGC is a big part of your content creation strategy. How do you actually incorporate that UGC into the content that you're creating?

‍

Alex Ross (Speaking)

That's a good question. And I will say, you're the person that pointed out Reddit and Quora's traffic change to me. And for anybody who hasn't seen that graph, like I recommend pulling it up because I wish I had a friend at that company. I wanna know like what is going on? Cause that graph just jumps off a freaking page. Yeah, so for us, our UGC is closest to Quora.
Like the spirit of our community is most like Reddit. As very kind and very help oriented, but the format of the content is Quora. So it's question and answer. And so basically if there is a page about a Monstera, which is a very trendy plant, having browning leaf tips, which a lot of people don't understand, like why is that happening? It doesn't look healthy, is it? We actually have a bunch of questions specific to Monsteras with browning leaves, right? And so the simplest way we use it is we enrich those pages and we show
the top five questions about that topic and that plant on that page. It basically increases the likelihood that they're looking at something that looks exactly like their problem. The other major area is plants are very visual. And so we have millions and millions of photos. And so we add those to the pages. And that's important because it tells the user that they're in the right place. Because a lot of people don't, most people actually don't know plant names. They recognize plants visually. They look at a picture and they say, yeah, that does look like what I have in front of me. Nobody knows like Monstera Deliciosa,
what does that mean? And so that image serves a function of saying you are in the right place, right? 

‍

Nate Matherson (Speaking)

Well, that's so interesting that you mentioned unique images being also important for users, and it sounds like an important part of the strategy here, because with some of the recent updates, we've seen websites that really invest
in creating unique images and videos and uniquely sourcing those elements into their pages have had a lot more success than other sites that are simply copying and pasting imagery or videos from other parts of the internet into their pages. And so it sounds like you've got this massive repository of unique imagery that you're also enriching your pages with. And I'd have to think that would be like a pretty positive signal to Google. Would you agree?

‍

Alex Ross (Speaking)

30:30

Absolutely. I personally have a hunch that Google, like if you think about it, Google needs websites to keep publishing original content. I think there's something going on right now where if you're a website like Stack Overflow and you've got a bunch of original content and you have all of these different LLM makers. It's not just ChatGPT, you have like Anthropic and Cloud, you have Mistral, you have a bunch of other like, you know, maybe HeyPi, are those guys. Basically like, it's very unclear
if you should be distributing your content for free to all of these other companies. I think Google is hungry for original images, for example. And so I think that like, again, this is just a hunch, like we don't get a ton of, our search, our images show up in search results a lot. It's not really a huge driver of our traffic, but I do believe that Google must be keeping track
of like, oh, this is useful that we're getting so much raw data that we have never seen before from this property, right? There's gotta be some sort of like novel content score that Google has for a property, especially one like ours, where we've got like hundreds of thousands of pages published. So they must be scoring, like, is that novel or not? How else would they know that ours is novel and therefore worth sending people to? On the images front, I would say like, I personally would never just like use documentaries or stuff like that. I would say the AI imagery is fascinating. And we, so when we're thinking about images for these pages, half of it is UGC.
The other half is a strategy I'm more excited about and kind of more passionate about personally. And that is like illustrations. And so we have an experiment running right now where we used, I think it was mid-journey, to generate using the context from the article, we would basically say generate an image in a specific style of a person with, for example,
a Monstera plant with brown leaves. So basically the image, the story image, then fits the context around it. And it serves a little bit of a function of like saying you're in the right place, but based on early traction, it seems like these pages are getting more traffic. And my belief is the reason is
because it's more entertaining. And that's something that I think like maybe not all websites think about, but like we're a consumer app, I come from Tinder. And so I am always thinking about what is fun, you know, like what do people actually want to like be delighted by and what's going to get them to like keep scrolling. And it doesn't always have to be information.
It could also be a story, right? So these images that we have are like, kind of like Pixar, like, not exactly that imagery, but like, it's kind of like that. The ones that work are very heavy with emotion. And this is the coolest thing, is that you see these pictures and it's like a little kid with this like really emotional, like excited, elated face, or if it's a sad article about troubleshooting, it's a sad face. And it's just so authentically sad and their plant is there and sometimes we have a pet in the picture and the dog is really happy. So these images aren't like, again, they're not functional, but they are really beautiful. That's the key. Again, quality is totally the most important thing here. And so that requires a lot of like curation and we have not been able to fully automate that yet. But I do think these like illustrations in this storybook format, it's kind of like those books you read through as a kid, Aesop's Fables and stuff, they make you feel something.
Dr. Seuss is another good example. I think there's something to that. And in our domain, like we're trying to make plant care fun and accessible, especially for kids. And so the image can also play the role of entertaining, which from a technical point of view, I do think increases like time on page, scroll rate, engagement, click through to other pages. So I kind of feel like that's something like that we're kind of on to early that I don't think people are thinking about, which is how do we create a good story with these images? You know, you've created a new bar and that is, as you mentioned before, does this piece of content make me smile?

‍

Nate Matherson (Speaking)

34:13

Yeah. We had Mike Haney from Levels on the podcast a couple months back, and he mentioned that his bar is what is your content adding to the internet? And so now we've got two bars here that we're looking at with Greg. Does this piece of content add something new to the internet, which it sounds like it does. And there are a few different ways that you do that with your pages. And imagery is certainly one with all of the unique images that you collect from the users on your platform.And then two, the unique illustrations, which I have seen on your site. They are very engaging. And for all of our listeners, we're going to include a couple backlinks back to the Greg site in the show notes. So you'll be able to check some of these out. They're very engaging. They're very emotional, like you said. And I totally agree that the imagery, the unique images that you're creating, whether
with AI or from your UGC base, will certainly keep people on the site longer, get them to engage more, get them to keep reading other articles. And speaking of making us smile, I've seen a lot of emojis on your site and not a lot of websites use emojis within their content pages. What was the thought process here? That is, what was the thought process for using emojis within the content too?

‍

Alex Ross (Speaking)

It emerged authentically to our team, I think. So like the first thing I would say is you can't just like insert them if you don't use them natively, if that makes sense. But really what it comes down to is, you know, I actually learned this at Tinder, but people are visual creatures, you know, vast majority of people. I think, I think people listening to the show
and you and us, like, we are on our computers a lot. We read a lot. We write a lot. You have to keep in mind that the average person, average consumer is not that. So it depends on your audience, right? Our average audience is they're in who knows where, you know, and like, how comfortable are they with reading?
Do they read many books? Honestly, most people don't, right? Like most people are on TikTok all day. And so, I think you kind of have to, it's worth acknowledging that if you're in a consumer space, most people are visual creatures. That's why video has been dominating for the last, you know, five years. And so, emojis, I think, are fascinating because they're basically like modern day hieroglyphs, right? They're like, they can contain so much meaning if you use them appropriately. And so, we use emojis to reinforce the concept of the title or the body content visually. And so we actually are experimenting with putting them in the page titles because I noticed that Google does accept them and it does rank pages that have emojis. I wouldn't overdo this because obviously it gets spammy real quick. So our bar was that if there's an emoji that's like a slam dunk where like a Gen Z or a Generation Alpha
person would look at that, would say, yeah, that is the right emoji to use there. That emoji has rizz. You could say it that way. That then it's, then it works, right? Cause it's like, oh, that's value add. That has helped my brain understand what it's reading more effectively, right?

‍

Nate Matherson (Speaking)

37:04

For all of the spammers that are listening to this podcast, please don't use emojis. Alex is right. Google is actually pulling in the emojis into the title tags in the SERP results. And I would have to think that this is having a really positive impact on the click-through rate for your pages in the SERPs, which is probably having a really positive impact on the rankings and performance of these pages. And so for example, I'm looking at one of your title tags in the SERPs now, and you've
got this beautiful giant red apple next to the title tag, which is just like screaming at me to click on it. So this strategy might die in like the next like 12 months as like the, you know, the spammers of the internet figure this out. I might actually go and add emojis to all of our title tags after this. The key is that the emoji is like really apt, you know what I mean? Because if you see an emoji and it's like not connected to the title enough, then it is a downside. And I actually wanna say, we try to do this in our app store listing and Apple does not allow emojis in the titles. 

‍

Nate Matherson (Speaking)

So I do think you're right, that like the trajectory of this is like, do it right and it'll last as long as it lasts. This has gotta be a short lived phenomenon. I can see like in the SERPs, like coming to you soon,
like someone Googling personal loans and you've got like the money bag emoji next to the title tag or like, you know, best online casinos, you've got the little slot machine emoji and the title tag. So I'm excited to see, uh, you know, I can't imagine that this lasts forever, but at least for right now, like to everyone that's listening, maybe we should be adding some emojis to our title tags. But in all seriousness, I think you do a great job of tailoring the emojis to the context of the article. So if you've got a section about fungus, you have a mushroom emoji. If you've got a section about insects, you've got an insect emoji. If you have a section about sunlight, you've got the sunlight emojis.
And so it's very clear that you've been extremely thoughtful in the placement of these emojis, and you're not using it just, let's say, as a means to manipulate search. And it's actually creating a much better experience I'd argue for the people landing on your pages. And you know there's something I really want to talk to you about and that would be internal linking. As you know we started this podcast talking about internal linking. We're gonna keep talking about internal linking but I imagine with 30,000 pages like internal linking is probably impossible or extremely difficult. But I've noticed on your pages you've got tons of internal links and you tend to go very deep with those internal links to other like very topically related pages within each one of these plant species. So how the heck are you scaling internal linking
in such a way that you're doing it very thoughtfully across what is 30,000 pages?

‍

Alex Ross (Speaking)

The thing to keep in mind is that every one of these pages is completely crafted using LLMs. Chad GPT mostly, but it doesn't really matter. And it's that 10,000 line prompt framework. And so what we do is we go section by section and we give the LLM the text, and then we give it a list of other topics and pages. So we basically say, go into this text. And if you see any words or word pairs or phrases that are very relevant to one of these other topics and pages, then go ahead and add a link to it. And so that has been fascinating because again, this is a ChatGPT4 thing. So this is all only possible in the last year
because we try to do these things with older versions. And it, again, it's the reasoning, the complex reasoning that it's capable of that makes this possible. But I've been so amazed at how it will go like three levels of association away from a keyword. So a lot of internal linking is like, okay, we've got an article about toxicity. And so it looks for the word toxic or toxicity. Right. Whereas in this case, ChatGPT is very good at synonyms. It's like poisonous. It's even really good at understanding like symptoms. So if there's some discussion about stomach issues, it understands that that in this context is almost certainly caused by toxicity. And so it can link that symptom or that really quite removed, but still related topic. And it's able to associate it back to that page. It's really just as good as a human, in my opinion. It's just very difficult to like build, right? But the reasoning abilities are amazing. So one thing that I think is very impressive with your internal linking strategy is that you've used like varying types of anchor text that are sometimes very targeted, other times very descriptive of like the pages were actually internally linking to. And it sounds like ChatGPT or the engine that you're using is very helpful for Google too as they try to understand how all of your pages are interconnected and related to one another.

‍

Nate Matherson (Speaking)

And we recently had Cyrus Shepard on this podcast where in his research he stressed and found
that having varying anchor texts on your internal links is critically important. Like you don't want to exact match everything and that's definitely not what you're doing. And so I would say your internal linking strategy, at least with Anchor Text, is very in line to what Cyrus described to us a couple weeks back. And so my question to you,
like how important do you think Anchor Text is for Google in terms of understanding like how all of these pages are unique and relevant to one another? 

‍

Alex Ross (Speaking)

I personally have to believe it's important because I just, you know, as an engineer, I think that Google's job is extraordinarily difficult. And I think about it in terms of our domain. So plants, for example, I spend all of my time thinking about plants. I have database engineers. I have thoughts on this. We are just thinking all about plants. And it's hard for us to understand the associations between plants and topics and things like that. And so I think that Google kind of I'm going to guess that Google depends partially on these links to build what I would call like a knowledge graph. I've seen a little bit of research in this area, but they basically build knowledge graphs where you try to associate different topics to each other. And so I have to imagine that like what anchor text does is it gives Google a broader sense of the whole universe
of topics that this page is about or like how one page relates to another. And so if you're using only like the word toxic again and again and again and again, you're really only giving Google one data point, right? Whereas if you can then say, actually, it also has to do with like stomach pain or like, you know, throwing up and things like that,
then you give Google a much broader perspective about your page. And I think people may just overestimate like how much Google really knows as soon as you're past like the first layer of depth of information. We're talking about long tail plans are really complicated. And so I think we need, I view it as like,
we need to help Google understand this domain better. And so when we give it more rich, sensible information that adds up, I think it picks up on the signal and I have to believe that it helps. 

‍

Nate Matherson (Speaking)

Google has recently walked back some of its statements about SGE, at least from my perspective and in my opinion,
they've like reiterated that it's just an experiment. They're not totally sure what's gonna happen with it. And it sounds like they're not totally sure when, if at all, it's going to roll out everywhere. Google might disagree with me on this, and they're welcome to come on the podcast anytime. But from my view, it doesn't feel like Google knows what it wants to do with SGE. Do you think SGE is going to be released everywhere? Is it something that you're worried about as you're thinking about the long-term of your SEO channel?

‍

Alex Ross (Speaking)

44:46

Such a good question. We first had this idea to go deeper into SEO last summer. And I remember talking with an advisor, and the advisor was basically like, you shouldn't invest deeply in that being your channel, because it's existential for us and our company, right? And his point was, he thinks that search is going to change too dramatically in the next year.
I ended up thinking a lot about that, and my personal stance is that even if Google changes, I don't think users are going to. And this is my experience in consumer land, is that I think there's billions of people that are habituated to searching Google, and I think there's like no chance that's going to change in the next two years, you know? At the same time, I do think that my personal behavior in relationship to search has changed dramatically. I used to be super, I was so not into ChatGPT. Like I was actually the most like bearish person, and then at this point I use it multiple times per day. So the word has changed. In my opinion, it's kind of undoubtable that things are changing for sure. The way we access information is changing. I just think it's impossible to tell what direction. There's so many different... Yes, I'm using ChatGPT a lot. A lot of people are using Perplexity.
But I think it's really important to keep in mind that we are the early adopters. And so we're like 1% of the internet or less. And so I just think that I've been doing consumer for a while and getting a change to propagate across the entire internet is shockingly hard and it doesn't happen in two years. And so I would bet that Google is dealing with such an existential major product design change
and it's so hard for big companies to innovate their own product that dramatically. There's a whole book about it, The Innovator's Dilemma. It's just so difficult for a company to disrupt itself. So I think I personally would not be surprised. I actually would be surprised if we saw dramatic changes to Google search in 2024.

‍

Nate Matherson (Speaking)

And that's the bet that I ended up making is that I just don't think this is the type of ship that turns that quickly. When we pivoted to Positional, we pivoted to get here. We emailed some of our investors and we told them we were building tools for SEO. A few of them emailed back and said, SEO is dead. I've heard SEO is dead for 10 years now. I think SEO has changed a lot in the last 10 years. I'm an optimist in the sense that even if search fundamentally changes, there will still always be a place for helpful content. Maybe the retrieval engines or systems change, but with a lot
of the content you're creating, I imagine there would still be places that this content would be helpful and could be served, even if like the UX UI of search as we know it today changes. I do want to talk a little bit about startups. Being a startup founder, it's a hard job. This is your second company. And so what made you want to do it round two after round one?

‍

Alex Ross (Speaking)

That's a really good question. Startups are so hard. It's pretty brutal.

47:43

I think for what it's worth, really enjoy it. And my advice to any entrepreneur is a personal take, but I loved being at Tinder for four years because I didn't have to think about payroll. Other people were paying the paychecks and health insurance was covered. So after taking some rest and recovery, I just felt like I had the energy. And if I had to really simplify it down, I would say the first thing is that it feels like the most authentic way for me to spend my time. I care a lot more about like, do I enjoy what I'm doing? Do I enjoy who I'm working with? Is there a story and narrative behind the contribution I'm having to the world? And all of those things,
I just have a really hard time imagining doing as an employee at a big company, you know, cause you really, it's impossible to not be a cog in the machine. So for that reason, like entrepreneurship, even as difficult as it is, you don't make money for a long time. I did feel very compelled to start a new company. And when I landed on this idea and plants, like I really love plants. I really, really enjoy being around them. I love the feeling that other people get when they help a plant grow and bloom. It's very satisfying. That just was easy for me to make that decision. The last thing that I'll
mention is, like my advice to anybody who's young and thinking about like financial planning and everything, if you think you have the ability to build a company, it is the surest path to wealth in the current economy. So I think like if you really, if you do the math and you want to like build a life of freedom where you're not, you know, like money has diminishing returns on happiness, but if you want to earn a certain amount, like 10 million or more, it's very difficult to do that with a salary. You could be a top earning AI engineer earning 500k for 20 years and okay, that's 10 million. What about taxes? And so I think if you do that math, you can kind of reason like, oh, if I want that for myself and my family, then I should get started sooner than later because it does take a long time.
My first company took almost 10 years to exit. How crazy is that? That's a long, long time. I think most people start a company thinking, I'm going to sell it in four or five years. That never happens, almost never.

‍

Nate Matherson (Speaking)

Well, I think Noah Kagan, one of our recent guests on this podcast would totally agree with you that building and starting companies is a great path to wealth and you can make life-changing amounts of money and you can also buy $6,000 toilets. Thanks for that suggestion, Noah.My next question is about Greg. Where do you see Greg in, let's say, 10 years from now?

‍

Alex Ross (Speaking)

Like I mentioned, we started, Greg, as the plant information company. And so the unique insight we have into plants is that they matter. So plants produce all of our oxygen and they also produce all of our food. The whole food chain starts, almost all of it starts with plants, right? And the climate is changing. And the problem is the climate is changing faster
than plants can adapt to. So we'll see if we end up being right, but we have a belief that plants need our help. And so my goal for Gregg over a 10 year time span, this is definitely a much longer than 10 year company. I started this company with like, I wanna work on this for, I don't know, however long it takes, you know,
but our goal is to build these layers of knowledge about plants and the ability to help people with them. So that we're starting with consumers in their homes and gardens in offices. And then we expand to agriculture. So people who are growing plants for food systems, I don't know how much, you know, but there are a lot of countries across the world that are in pretty serious food insecurity situations, and that's only going to grow. And the problem is that so many people are affected by that,
that it becomes everybody's problem, right? It is impossible to have hundreds of millions of humans who are food insecure, and for everybody around the world not to be affected by that. And so, you know, we'd like to do our part there. It's a kind of like grand mission, but I think that information and knowledge
around what plants need is an important part of that solution. And so if we can position ourselves to eventually be able to do something and provide a helping hand, I think that'd be really cool. But that's also like, right now I just want to make payroll. I just want to build a successful business, but we're making sure that the things we do, it builds towards that long-term vision.

‍

Nate Matherson (Speaking)

51:50

Payroll is good. I'm sure your employees listening to this podcast are looking forward to the next Friday. I love that you've walked us through that bigger vision and that nobler cause that you have for Greg. 

‍

Lightning Question Round:

Nate Matherson (Speaking)

And if it's okay with you, I would love to ask you a few lightning round questions. Does that sound good? 

‍

Alex Ross (Speaking)

Yeah, let's dive into it. 

‍

Nate Matherson (Speaking)

I do have a quick question about plant care. Do mums grow back or are those like a one-time, one-use plant? Because currently I have like eight dead mums sitting on my porch right now.

‍

Alex Ross (Speaking)

That's so tough. You know, there's a lot of seasonal plants that people like. Poinsettias and mums are very popular at the Christmas season. And yeah, they definitely do grow back. The only type of plant that I'm aware of that doesn't grow back are certain flowers that they have a life cycle where it just grows and then it dies. But no, mums, yes, you can grow them back. And same with poinsettias. You can actually plant them in your backyard. You can totally grow them. They might just look a little out of place because these big red holiday Christmas things in summer, like maybe not, but yeah.

‍

Nate Matherson (Speaking)

All right, well, to my wife, who's hopefully listening to this podcast, I told you so, they will grow back.
VC funding, if we're starting a new company, should we take it or should we bootstrap it?

‍

Alex Ross (Speaking)

I think the most important thing is just to ask that question. Because I think that most people, especially people in tech, kind of assume that VC is like the success metric and it's the only way to build a company. Actually, I think most companies should not. That would be my answer. I think that it's important to view money as a tool. It's a very specific part. Financing a business is a very specific part of your strategy. And I think that a lot of people let it grow because of their egos and because of the news articles that come and stuff like that. They let it become bigger and bigger. And the problem is that VC capital, and honestly, any type of capital can do this, it's not just VC, but anytime you're taking somebody else's money,
you are reducing your options in the future, right? And so I think it's important to think about, for example, if you raise VC capital at a certain valuation, that means that you can effectively never sell your company for less than that amount, right? So if you raise capital at a $50 million valuation, it'll be very difficult to ever sell your company for less than $50 million.
And if you have a company that is worth $30 million and you own 100% of it, I guarantee you, you're gonna wish you could sell your company, you wouldn't own 100%, but basically the point is that accepting that capital reduces your options. And some of those options you may wish you had in the future. And so I think it's just important to get good advice and to think through a little bit what your strategy for the company is before basically walking through that one-way door. And that's actually the most important advice I could give to entrepreneurs is there's a very few one-way doors when starting your company. Your cap table is one of them.
You cannot get people off of your cap table, almost ever. And so if you put someone on it, you need to just kind of think through what are the implications of that.

‍

Nate Matherson (Speaking)

Who is that one advisor that if you had a really difficult question, you'd go to and ask?

‍

Alex Ross (Speaking)

For me, you know, one of the former CEOs of Tinder, Elliot Seidman is such a good entrepreneur. He's just such a critical thinker. My style of entrepreneur and leader that I look to kind of follow is somebody who is able to consistently deliver good results. So I've in consumer land, there's a lot of entrepreneurs who have a one hit wonder and they're successful, but can they do it again? And oftentimes the answer is no. So that's somebody who I look to and I see like, oh, he's consistently very good, which means that he has a playbook or like some decision-making framework that reflects reality. Cause that's really what it is, is you can look at the world and you can say the ball is going to go here for these reasons.
I admire that a lot. And so that's a person I look to get advice from. 

‍

Nate Matherson (Speaking)

Backlinks. Is that something you're thinking about?

‍

Alex Ross (Speaking)

Definitely. Yeah. I think my observation of Google could be wrong, but I think that Google's relying on brands right now. Because I think that with ChatGPT, there is a lot more spam. And so I think Google needs to rely on signals that this is a trusted publisher who has a reputation that they don't want to harm like Greg. And so I think that backlinks are one of the signals that demonstrate brand equity to Google.

‍

Nate Matherson (Speaking)

Well, Alex, this has been so much fun. Thank you for putting up with my questions. I know we veered off script here and there, but this was such a great episode. Thank you so much
for coming on the Optimize podcast. Is there anything else you'd like to say to our listeners?

‍

Alex Ross (Speaking)

No, I loved being on. I would say keep listening to the pod. It's such a great source of information. I looked at other podcasts and I don't listen to them anymore. So I like yours. I think you're doing a great job. Please, please keep posting.

‍

Nate Matherson (Speaking)

Yeah. Well, leave us a review, Alex, if you haven't already. Somebody recently left us a one-star podcast review, which is totally fine. Like, you know, maybe it wasn't one of our best episodes, but to all the listeners, if you did enjoy this podcast with Alex, leave us a five-star review and hit that subscribe button.

‍

Ad Spot:

56:58

This week's episode of the Optimize Podcast is brought to you by Positional. You probably know by now that my name is Nate and I'm one of the co-founders of Positional. And we've built what I think is a pretty awesome tool set for content marketing and SEO teams. We've got a few features you'd expect like tools for keyword research and keyword tracking, but we've also got a few tools that you've maybe never seen before. For example, internals for internal linking and content analytics, which is kind of like a heat mapping tool, but for a content team. It helps give you insight into where in your pages you might want to come back and improve. We've got about eight tools and we'd love for you to check them all out at www.positional.com.

‍