In this episode of Revenue Boost: A Marketing Podcast, titled From Traditional Search to AI-Driven Strategies: Future-Proofing Your SEO, host Kerry Curran is joined by Paul Shapiro, SEO expert from Uber’s web intelligence team, to discuss the evolving role of AI in search engine optimization. Paul shares his experience with AI’s transformative effects on SEO, from leveraging tools like GPT for content creation to navigating potential threats from AI-powered search engines. Together, they explore strategies for future-proofing your SEO efforts, optimizing AI-generated content, and ensuring long-term success in a rapidly changing digital landscape. This episode is packed with actionable insights for executives focused on digital marketing and revenue growth.

Podcast transcript

Kerry Curran, RBMA (00:01.462)

So welcome Paul, very excited to have you with us today. Please introduce yourself and tell us a bit about your extensive background and expertise.

Paul Shapiro (00:11.338)

Thanks for having me, Kerry. Yeah, my name's Paul Shapiro. Right now I'm working at Uber on what we call our web intelligence team. So the web intelligence team is focused on Uber's website, which is an interesting channel for Uber, which is an app-based company. But we're really focused on building out a lot of our data tooling for marketing channels like SEO, and PPC.

CRO, et cetera. And it's, it's, it's, it's been a really interesting ride so far. Prior to Uber, I was at Shopify and I ran the technical SEO team there. great experience. We did lots of very interesting experiments and all sorts of things. was, was, it was there for like two or three years. I couldn't say better things about that experience and then prior to that, I was on the publishing side. I was at Conde Nast running the technical SEO team for publications like Vogue Wired and Bon Appetit. which had its own separate set of challenges from Shopify. Then, you know, prior to that, I was on the agency side, working alongside very talented people like Kerry Curran.

Kerry Curran, RBMA (01:37.921)

Thanks, Paul. So yeah, so Paul and I go way back, back when he was leading the SEO for both Catalyst and parts of Group And we originated the tech SEO boost, which someday we'll bring back. But always great to speak with you, Paul, and definitely miss the days when we ...

We're brainstorming together and always learning a lot from you. So I know in all those years, it has come to life and impacted SEO so there have been a lot of changes back from when you and I were partnering together. Talk a bit of what you've seen with that evolution and how AI is impacting SEO today.

Paul Shapiro (02:32.588)

My God, can I first say what a change from when I started playing with AI? remember probably back at Catalyst, was something called, they called it more, it had a specific name because that's what it was. Then it was a very niche part of AI called natural language generation. The hot model was GPT-2, with no association with open AI, I believe.

Kerry Curran, RBMA (02:51.766)

Mm-hmm

Paul Shapiro (03:02.396)

And it spits out texts that are pretty incoherent. And I was completely caught off guard. I didn't think this was going to become anything in the near future. But here we are with chat, GPT, and all these other models that have changed pretty much every way of working, I think, for most people. What a time to be alive. Yeah.

There are sort of two ways I've been thinking about AI within the SEO and marketing world. One is, what is the threat model? How is AI gonna jeopardize SEO and what do we do as marketers? B is how can we, or number two, rather, how can we leverage AI to do better marketing?

Kerry Curran, RBMA (03:56.139)

Mm-mm.

Paul Shapiro (03:56.62)

two very, very different ways of thinking about how we need to conceive of AI in the workplace, both important.

Kerry Curran, RBMA (04:08.586)

And so, when you talk about the impact AI has on SEO, where do you kind of see that going as maybe a threat to the future of SEO?

Paul Shapiro (04:22.498)

Sure. So one, SEO is very much grounded in what I will call the traditional SEO model. People are optimizing for keywords. They want to appear higher in search so they can get more traffic and convert that traffic into customers. AI challenges that whole experience.

On the Google side, these AI overviews, which provide sort of a summary of what you would expect in the search results themselves to users so they never have to click on a website. On chat GPT, there's there's a they're introduced this was it search GPT. So it's sort of like a replacement search engine, where it, you know, combines much like Google does like the large language models with traditional search results.

Kerry Curran, RBMA (05:10.176)

Mm-mm.

Paul Shapiro (05:21.962)

You have search engines like perplexity, that are doing just that, like on their own, you know, being that is powered by chat GPT doing something very similar to Google. and then you have people that are just, they're, skipping search engines altogether. They're going to chat with GPT to find their answer. not the smartest thing to do, but, you know, people are, people are doing that. So … 

Kerry Curran, RBMA (05:44.193)

Hm-mm.

Paul Shapiro (05:51.03)

… you know, anything that's going to take away market share from those traditional search sources is going to impact the bottom line of your SEO efforts.

Kerry Curran, RBMA (05:56.94)

And so do you think that it's more going to kind of take over the role of your traditional SEO manager brand side or do the SEO managers and brands need to kind of think about it differently?

Paul Shapiro (06:18.722)

It's usually the latter. I think SEO at this point has a rich history of being a very adaptable marketing channel. it first started, as some people say it was even before search engines and it was more about link building. So like you had these portals and people found websites through these portals and then they built links from those websites and people got to your website. And like that's how the OG like SEO works.

Kerry Curran, RBMA (06:36.982)

Mm-hmm.

Paul Shapiro (06:48.39)

… and then like SEO came along with search engines like Google and people were trying to like, make sure you appeared at the top of Google. and, and then, you know, there were penalties for, you know, exploiting some of the things that these search engines worked out of. then SEO became a little bit more, closer to traditional marketing or rather a combination thereof.

Kerry Curran, RBMA (07:15.03)

Yeah.

Paul Shapiro (07:16.174)

… and you know, now we have AI at play and I, you know, I think it's just like another thing to adapt to at least, at least in the near term.

Kerry Curran, RBMA (07:27.018)

Yeah. Well, and you know, we talked about too, is that the AI tools create a lot of efficiencies and they can, you know, help you as a starting point to your point, can go directly to chat GPT and kind of type in what you're looking for. And I, I'll say like the format it comes out with is a lot easier to kind of read through versus clicking into multiple websites. Maybe that wasn't what you were looking for.

But however, to your point, I get it, it's not yet as robust as a traditional search engine.

Paul Shapiro (08:07.906)

I mean, I think it's pretty robust, especially when you like to add in follow-up questions, but what it's lacking is reliability. Like, it's the you know, give like hallucinations with these large language models. And you can't 100 % trust what they say. There's there's, you know, at the end of the day, these large language models are like advanced, really, really, really advanced, like … 

Kerry Curran, RBMA (08:11.956)

Mm-mm, yeah. The accuracy.

Paul Shapiro (08:36.062)

… autocompletes. They try to, with a certain probability, try to predict what the next word will be. So it's pulling from this large corpus of text that they were trained on and guessing what the next word would be. And oftentimes that makes a lot of sense. Like it's coherent. It has a semblance of logic because the English language or whatever language you're speaking has a semblance of logic. But it makes stuff up.

Kerry Curran, RBMA (09:03.968)

Yeah, it makes it up or draws conclusions or connections that aren't necessarily accurate. That being said, from an efficiency perspective, you can use it in a lot of ways for content creation. But when it comes to creating blog content or landing page content that you want to rank well, what's your recommendation for how you're going to leverage a chat model?

Paul Shapiro (09:33.41)

I mean, there have certainly been people who have generated full-out blog posts and liked entire websites using chat GPT and large language models. A lot of those have also been hit hard by, you know, Google algorithm updates and they've lost traffic. I'm not going to just 100 % dissuade people from experimenting with that, but I think it does pose a certain riskiness to your marketing efforts.

Kerry Curran, RBMA (09:40.928)

Mm-mm.

Paul Shapiro (10:03.37)

So yeah, I think it's much better to sort of employ it in places like, you know, copywriting and on your landing page, like writing persuasive copy, filling in or even like filling in parts of blog content, just sort of like augmenting your writing. I often just sometimes struggle writing a conclusion and I'm like, okay, here's what I wrote. Can you help me like, you know, compose a conclusion?

Kerry Curran, RBMA (10:19.969)

Mm-mm.

Paul Shapiro (10:31.584)

Like things like that, it's good. And obviously, AI models are much better when they're paired with some sort of existing knowledge. So they talk about a ragged process. So I forget what the acronym is, like retrieval augmented generation or something like that. I'm going to have to Google this now. But basically, it gives it a foundation of … 

Kerry Curran, RBMA (10:42.86)

Mm-mm.

Paul Shapiro (10:57.75)

… of a knowledge base. you could use something like a rag or a knowledge graph and the accuracy and hallucination rate is going to be much, much, much better. Okay. I was off by a rag. It's a retrieval, no retrieval augmented generation. was right. Okay. Nevermind.

Kerry Curran, RBMA (11:18.445)

And so is that like, you're like, as you were saying before, like for it's great for e-commerce product details. So would that be if I'm, I'm uploading, you know, the product, all of the details I have from the manufacturing data or, you know, description, then say, write me a three sentence description of this product. That kind of, am I, is that right?

Paul Shapiro (11:25.71)

Yeah, I mean, a huge, huge problem for websites like e-commerce websites is that they're, you know, they have, you know, this whole catalog of products. and you know, most of their SKUs, they're just filling in something from a manufacturer. That, that's nothing unique. There's nothing special about your, you know, your product landing page, compared to the other guys. So like, yeah, I think AI fits very well into that sort of thing.

Kerry Curran, RBMA (11:48.629)

Yeah. All… Mm

Paul Shapiro (12:06.54)

Where you can sort of customize that experience for your brand and, you know, add some pizzazz and uniqueness to who you are on those landing pages.

Kerry Curran, RBMA (12:18.314)

Yeah, I mean, I definitely use it similar to what you were saying to kind of get a description or get some description or summary of something I already have written, especially, you know, with the podcast content or if I promote writing something on LinkedIn, I don't, it's not a straight copy and paste as much as a, you know, what are the key points here or check.

Paul Shapiro (12:42.508)

Right.

Kerry Curran, RBMA (12:43.404)
Write this more efficiently or check my grammar.

Paul Shapiro (12:49.002)

Yeah, I think using it to augment your existing work is the smart way of using it more or less now. And also knowing that providing it context is really important using something like a knowledge graph or rag and knowing how to prompt it properly. I've seen so many people be like, I tried to chat GBT and the answer was like, okay.
 

And I was like, well, have you, have you done this, this and that, to like, just make sure it gets like a better reply. Because you know, it's, it's, I guess it's, you know, it is sort of magical, but maybe it's not a hundred percent there yet. You can't guess what's inside your head. You asked her a question. Yep. You have to form, if you want a certain answer, you need to tell it what kind of answer you're going to want. And it doesn't have the knowledge of everything like out there. You need to provide it with that knowledge beforehand. If you want it to answer a certain way.

Kerry Curran, RBMA (13:23.082)
Mm-mm.

Yeah, no. You're right, you have to keep kind of testing it until you get what's accurate and what's helpful. But I know for the future of SEO and the role of organic traffic, there's challenges ahead. So what are your recommendations for companies to think about kind of future-proofing their positioning when it comes to especially brands that depend on SEO heavily.

Paul Shapiro (14:20.288)

Yeah, I mean, so I think, first of all, these AI search engines, they're probably going to take some market share from search engines, which means you're going to get less traffic and you're going to convert less of that traffic. I think certain industries and businesses are going to be more affected by that than others. Like I would worry about media companies, like publishers, like those would be the first I would … 

Kerry Curran, RBMA (14:29.121)

Yep. Yeah.

Paul Shapiro (14:48.95)

I would worry about it, which is why a lot of them are blocking chat GPT and other search engines from sort of training their models on that content because they pose a risk to their businesses. And then what sort of market share are these gonna take? I think that's probably growing steadily. You have the ...

Kerry Curran, RBMA (15:00.787)

Mm-mm.

Paul Shapiro (15:15.778)

There's a Gartner study. don't know if you've seen this one where they say, I think it was 20. There are several years. I think the first phase is like by 2026 or something. I have to look this up later. It's going to be like 25 % of search traffic. And then by like 2028, it's like 50 % of search traffic. I think these numbers are too high. But it sort of does show you that there's a risk there.

Kerry Curran, RBMA (15:37.139)

Mm-mm.

Paul Shapiro (15:46.134)

I think they're too high because it's not taking into account AI overviews in Google, which also takes away traffic. If people aren't clicking on your search results, that's getting cannibalized. Eventually, you'll have to, if you're like an e-commerce business, they'll have to transact with you. Maybe if Google doesn't take that away as well. You have businesses like...

Kerry Curran, RBMA (15:52.673)
Right.

Paul Shapiro (16:16.012)

Amazon aims to cannibalize small e-commerce businesses and independent e-commerce businesses and make it an all-in-one store. So I think you're constantly at risk, and AI is only making that worse. So I guess to get back to your point, what are you doing about that? I think you have to keep doing really, really … 

Kerry Curran, RBMA (16:38.376)

Mm-mm.

Paul Shapiro (16:45.74)

Good marketing outside of SEO, start doing things from a product-led growth approach. So make sure that you're creating a product and product features that are going to benefit the users the most so that you become the go-to source for everything. Like if I, if I'm selling, you know, e-commerce products, like you want, you want to be Amazon. You want people to like to know to go to Amazon first.

And that's, you know, because they created a great e-commerce experience with everything you could want at a decent price that gets to you the next day on your doorstep. And that's, that's, that's true of, of all businesses. Like if you're just creating run-of-mill content and hoping that that will rank in Google, and you know, then people will go through the sales funnel and eventually like to convert on that. Like, yeah, that might work, for a little bit, like what happens when like, like Amazon.com pops up in your industry. You need to, you need to sort of design around that thinking for the future. And that's, that's truer more than ever now with these AI models.

Kerry Curran, RBMA (18:02.057)

Yeah, and I definitely, you know what you were saying about publishers, I've been, you know, speaking with a lot of publishers and through the Google updates that happened over the past year. Just had Max Willens from eMarketer last week launch a paper that showed the decrease in revenue that the publishers have, the impact.

And I think your point about marketing, just brings it back to why that's also important. My recommendation to the publishers on a roundtable last week was you need to start thinking like a brand and start doing other marketing efforts and in promoting yourself through other channels as well and not just.

I think they've learned not to depend on the SEO bucket, but like you said, whether it's your differentiated product or just you need to remember your brand first and you need to be kind of attracting and drawing your customers. But no, this is great. So are there any top tips for how people should be approaching AI when it comes to, what we talked a bit about from a content creation, but any other key tips you want to share?

Paul Shapiro (19:21.03)

Yeah, yeah, I mean, I think you need to pay attention to these AI overviews right now. I mean, Google is still the top search engine worldwide. And they rolled out AI overviews to other markets and for logged-out users. And that's your near term threat at the moment. The good thing is, I think when they were beta-testing it, a lot of the search results were coming in from … 

Kerry Curran, RBMA (19:25.962)

Yes.

Paul Shapiro (19:50.178)

… outside of the top 10. It was all over the place within the top 100. And since they've launched, it's closer to the top 10 search results now. So if you're doing SEO and ranking within the top 10 results, you're setting yourself up for success there. And I've also, I've run tests in the past where you could do slight manipulations of those test results by altering the language slightly.

Kerry Curran, RBMA (20:01.388)
So yeah.

Paul Shapiro (20:20.05)

I know Cyrus Shepherd, also on his blog, has run some tests that were slightly different than my own, that's, that's, that's publicly shared on the internet. So if you're, if you're looking for resources, you know, check out his post. but yeah, I think you're, you're good just sort of like focusing on traditional SEO for the moment there. so do that in the near term and then for the long term, start making that, that product led, … 

Kerry Curran, RBMA (20:29.664)

Mm-mm.

Paul Shapiro (20:49.804)

… growth strategy for the longer term.

Kerry Curran, RBMA (20:54.806)

Great, well thank you, Paul. So great having you on today. As always, good to catch up with you. So thank you so much for joining.

Paul Shapiro (21:03.116)

Yeah, absolutely. Thanks, Kerry.