In this episode of the Marketing School podcast, Eric Siu and Neil Patel discuss their 2025 predictions for AI in marketing and share a suite of practical AI‑driven SEO and social media hacks. They forecast a massive surge in AI‑generated content and avatars, emphasizing the need to adopt early to capture email leads via ManyChat. Neil predicts AI agents will soon automate low‑level marketing tasks such as keyword research and content‑gap analysis. The hosts then dive into specific tools and workflows: a proprietary API from Chirag that surfaces ranking gaps on LLM platforms; Opus Clip’s ability to extract mid‑form clips from long‑form podcasts for new YouTube channels; prompt‑engineered hook generation using ChatGPT to boost short‑form video performance; Google’s NotebookLM for summarizing videos, PDFs, and audio and generating scripts and multi‑platform captions; combining ElevenLabs voice cloning with HeyGen avatars for higher‑quality AI presenters; Google Dev Studio’s real‑time, zero‑latency AI co‑writing for scripts, CRO copy, and SEO content; and the use of Gemini for video summarization. They also discuss branding repurposed clips with watermarks, the importance of still optimizing for traditional Google SEO while also targeting Bing/ChatGPT, and a broader shift toward social‑signal‑driven ranking as search engines crawl more social content. Finally, they predict that SEOs will increasingly use AI to condense long‑form articles, removing fluff and delivering concise answers, which will improve user experience and traffic in the evolving search landscape.
Chapters
Prediction for 2025 AI. I'm going to go first. I think we're going to see a huge proliferation in AI-generated content in general and AI avatars. We've talked about on this podcast HeyGen or using ElevenLabs to, you know, port your voice over. We saw with Rowan Chang, and I saw with a couple other people with their AI avatars, they're getting hundreds of thousands, if not millions per video. I think now is the time to attack that. We're recording this in January. You have to hit it now because everyone's going to start doing this. And by the time everyone starts doing this, you're not going to have as much of an advantage, in my opinion. I think if you start doing it now, you start collecting a bunch of emails using ManyChat, that's going to be a very nice combo. So I think that that's one prediction.
Prediction in 2025 for me is you're going to see a ton of AI agents doing marketing grunt work and people will not have to do grunt work anymore. I'm talking about for low-level stuff. I'm not talking about really complex stuff that requires creativity, trying to optimize for conversions and generate revenue. I'm talking about just low-level stuff like keyword research or content gaps. Things like that that can be easily done through AI. All right, let's do our favorite AI SEO hack.
You go first.
My favorite AI SEO hack right now is we have this API from our mutual friend Chirag. Do you know Chirag? You remember Chirag from Meddly?
I do remember Chirag, yeah.
And it ends up breaking down all of the mentions in your rankings for LLMs and the keywords. You know, it'll talk about random queries and stuff that you're ranking really well for. You take that data and you take all the gaps, in the essence, the stuff that you're not ranking really well from. There's a high correlation from Google rankings to, you know, being mentioned on these platforms like ChatGPT. You take the gaps and then you go and optimize for them and you can have AI help you with it, whether it's, you know, creating content, refreshing content, building better backlinks, optimizing your title tags, meta descriptions. Taking the pages that you have on those subjects and becoming more specific and answering specific questions. So not being as broad increases the likelihood of you getting mentioned as well. So just doing little things like that, we see that as a cool future in which people should be optimizing more and more for these AI platforms.
So what he's built, is it kind of like Try Profound where they show how you're ranking on certain LLMs?
Correct.
Okay, got it, got it, got it.
Cool.
And it works well?
It works good enough.
Okay, it works good. Hey, that's all you guys need to know. It works good enough. It's not public, I'm assuming?
Not that I know of.
Okay. All right. So my favorite AI SEO hack that's coming, actually, I just spoke with these guys today, um, you know, Opus, Opus Clip, right?
Yeah.
Um, so there is now the ability, um, well, I can't reveal too much. Let's just put it this way. There's going to be— Neil creates, you create a lot of content. Would you say your content library historically is really big?
It's massive.
Of all the content you've made, right?
It's massive.
And mine's pretty big too, right? That's what she said. But anyway, so my point is, sorry guys, I had to. But my point is, there's a lot of things that we say within like a 90-minute conversation or 30-minute conversation that are hard to find. Like, it's hard for a human to have to go through all those. Now, you have long-form clips and you have short-form clips. But what about the mid-form clips, right? So mid-form clips, Opus is going to help you find the Midform clips. And I think we're going to see a proliferation of even more content, right? Because I know there's, there's, there's, there's some cool stuff within the stuff we've already published. And that's exciting to me.
What are you doing with all that Midform content?
Publish it to a new channel on YouTube. So this is YouTube AI SEO hack.
This isn't really an AI SEO hack, but, or even an AI hack. My team ran out of content the last few days. So, you know what they're doing on social media? Like on Instagram, they just go take my old popular videos.
Yeah, it works.
It works.
By the way, so my team does that, but they put, they put a watermark of Single Grain on it and it still gets a lot of impressions. Let's just get Single Grain impressions. They should put an NP Digital impressions. Yeah. Yeah. So, and look, there's so much, there's so much content out there. I think, by the way, I haven't touched Opus in years. And today I talked to them and they're like, you know, they're like, you talk about Overlap a lot. I was like, I sure do. And they're like, so we're trying to maybe work something out there. But they've really upgraded their product. And by the way, they've raised a lot of money too.
Really?
Okay, they raised a good chunk. They raised $30 million.
That's a lot of money.
Yeah, yeah, yeah, yeah. So I'm excited about that. Mine's more on the YouTube side. I think Neil's is more on the optimizing for LLM sides. And I think we can both share this part. Look, when you search for the best marketing podcast on Claude or Perplexity or Gemini or, you know, whatever else is out there, what usually pops up number one? Marketing School. Well, except for ChatGPT because they're on Bing. So the point here is that make sure in 2025, we're not done with traditional SEO yet. Keep optimizing for Google because a lot of the results pull from Google. Keep optimizing for Bing because that's where ChatGPT pulls from.
Yeah.
And ironically, because we don't optimize for Bing, that's why we're not even in the top results for Bing. But everything else, we're like number 1 for the most part.
Now for our favorite AI social media hack, you want to go first or you want me to? All right, so my favorite hack is using prompt engineering to come up with better hooks. So when you're creating a video, you have a specific topic, you may even have a script. We input a lot of our notes into ChatGPT. Then from there, we actually have them help us come up with hooks for each section. So when you're creating a long-form video, let's call it 15, 20 minutes, you probably can get at least 4 or 5 really amazing short-form clips from it. But the key with the short-form clips is you need to have amazing hooks for that section or else that short-form clip isn't going to do as well. So A, you can have AI help you with the hooks for these subsections so that when you're recording your long-form videos, you can also get more short-form clips. And B, what we found that it also does is it keeps people more retained watching the long-form video because you continually are using amazing hooks throughout the content. So it just keeps them there longer. Great.
Mine is I use a combination of things. I use NotebookLM, that's from Google, and I take our YouTube videos, okay, so like a video like this one, and I'll plug it in because I don't want to watch the whole thing again, right? So let's say my team wants to come up with, you know, like captions for a video that we're doing. So you dump it into NotebookLM, you take the YouTube video, put it in there, and it gives you a summary of what you're talking about. And then you take that summary, it's like, okay, create a script from that. To your point, create a script, come up with a really good hook, right? And so this can be rewritten for LinkedIn. It can be rewritten for a tweet, for example. And it ends up being pretty punchy. NotebookLM ends up saving you a ton of time because you don't need to read the whole entire article. I'll take a 42-page PDF on agents from Google and I'll just dump it in there. So I think NotebookLM is very heavily underrated. The other thing that we should maybe try for this podcast is NotebookLM makes a podcast of the summary, right?
Yeah.
We should take a lot of our content and just make the— dump like the little AI summary at the end.
What do you mean it makes a podcast of the summary?
Oh, we haven't done this.
Are you talking about a summary of the podcast episode or actually creates a new podcast?
It does a summary of whatever content you're uploading to it, whether it's a PDF, whether it's a video, or whether it's like an audio file.
Oh, so you're using that to create a new summary?
Yeah, well, I'm saying we could try that here.
To create a shorter episode.
Yeah, so NotebookLM's a big one. And I love using Gemini as well 'cause Gemini, Wait, wait, wait.
Let's go back to Notebook LLM. So let's say we upload—
LM, just LM.
LM.
Okay.
So let's say we go back to Notebook LM and we upload a script of our podcast, right? We're transcribing it. Boom. We pop it in there. It'll create a summary of it. Will it do it in audio format with our voices?
Not with our voices. It's, it's a guy and a girl talking right now, but I, in the future, I'm sure you can hook in with like ElevenLabs API and do it that way.
Or you can just use HeyGen and do it there.
Well, hey, I think ElevenLabs is better from a voice standpoint. So I don't know if you do this with your— okay, this is a social media hack, I guess. So I use HeyGen and then I plug— I record my voice in ElevenLabs and there's an API that hooks them together and it sounds way better than just the HeyGen voice.
Oh yeah, I haven't tested out, but you know, with HeyGen, the big mistake that people make is they don't upload enough information. Like HeyGen will give you stuff that you need, or my team at least, they send it to me and they say, hey, we need you to record all this stuff for HeyGen. And then it allows them to create better videos using the product.
By the way, because HeyGen, the lips are not quite there yet, it's maybe like 90%. So what that guy Rowan Chang does, Chang does, um, is he makes these videos talking about AI news, but his avatar will only show up for like 2 seconds, 2 seconds at a time. You can't see it for that long. So you, you, you intuitively know it's AI, but it— because it doesn't show it messing up, like you trust it enough because there's a lot of B-roll in a video.
Yeah, so we do something similar. So when we're using HeyGen, we typically mainly do it for non-English content. You could do it in theory for English, but we usually mainly, we mainly use it for non-English. We then go create the content and then let's say if I'm talking about an SEO tactic or strategy, we'll take the screen and showcase within that video that SEO tactic for whatever language we're in. So then that way you're not seeing me talking as much. Similar, pretty much just B-roll.
Yeah.
But we're just editing the video and adding in all the other stuff that we naturally would've. If it was in English.
That's good. By the way, I've noticed that, okay, me talking about NotebookLM, that's from Google. Me talking about Gemini, you can insert a YouTube video and it'll summarize for you as well. It does a good job of that. But also when you use Google Dev Studio and you do a share screen with it, I actually, I was so lazy to make a script because my team wanted me to make a script for one of our pages. And I was like, oh God, I don't want to think right now. I literally opened the share screen with Google's AI and I started talking with it back and forth, right? And I was like, okay, here's a single grain website. Give me a script based on this site. Give me a good hook and then tell me what I should be talking about. And the cool thing with Google Dev Studio is that when you talk to it, there's like no latency at all. So you're literally having a back and forth conversation like this, right? And we workshopped the entire script and I just used that script and made a video and that was it. So it takes, I don't have to think anymore. Great. And I can workshop things. And you can use that for CROing your website. You can do it for SEOing your website, or you can use it for a lot of different things. But to be able to show your webcam or to be able to share your screen and talk with an AI, I think what's going to happen later this year, this is our prediction for the end of 2025 maybe, is that we have these agents just writing for us in our style on different social media platforms and doing, I think we'll probably get to that by end of this year. So. All right, so I wanted to take a moment to tell you about my podcast co-host, Neil's agency called NP Digital, and they work with a whole host of global companies or a global organization. Also, Neil has SEO tools such as Ubersuggest and AnswerThePublic. All you have to do is go to npdigital.com to learn more, and we'll see you on the other side.
Yeah, and if I had to give one prediction for the end of 2025, I bet you're going to see a lot of SEOs have AI summarize their long-form content and make it more compact. Because there's too many articles on the web that used to rank really well with the model of let's create 3,000 words, 5,000 words, et cetera. And searchers want, they're getting used to getting answers quicker and they don't want to read 5,000-word posts when they could have got the data in 500 words, right? I'm not talking about just fluff. There's just a lot of irrelevant information that are in a lot of the older articles that may not need to be there because you would, before you'd want to cover topic super thoroughly to rank for it. Now instead, you just want to cover the topic and get straight to the point. And if you need to, you know, be more thorough on a specific topic, like let's say if it's SEO, instead of having an SEO article that talks about creating content, building links, and then go, you know, continually go from there, it's just easier to have an article on link building, an article on content, and just being super thorough to the point. And you can use AI to get rid of a lot of the fluff and the crap in a lot of your older content, restructure your overall blog, and you should see your traffic go up overall in the long run.
By the way, if you think about it, if you're Google right now, knowing that maybe website traffic is dying, and knowing that social is becoming stronger and stronger, you probably have to do a much better job of crawling everything on social because that's where things are going. And then what you would do is try to cross-reference, okay, Neil Patel on LinkedIn, on, on Twitter, on YouTube, whatever. Let's just look at all the things he's talking about and maybe rank him accordingly that way. But I think the way Google's going to choose to rank things in the near future is going to be completely different.
Yep.
Yep. Goodbye, guys. That's it for today. Go to marketingschool.io/agency if you want to grow your agency faster, and we'll see you tomorrow.
