So What? SEO for AI Part 3: Offsite Optimization for AI

Well, hey everyone, happy Thursday. Welcome to So What, the Marketing Analytics and Insights Live Show. I'm Katie joined by Chris. John is out today. Oh wait.

Other side, other side. No. Oh, for me, it's the right side. Oh. Is it the slide?

Yeah, it's out. Oh, so our cameras must be reversed because I'm like, no, I see it. So this is why I don't participate in the high five. This week, we are doing part three of our SEO for AI series. So two weeks ago we did technical SEO for AI.

Last week we did uh on-site or content creation. You can catch those episodes on our YouTube channel, TrustInsights.ai slash YouTube. Go to the so what playlist, those are there for your viewing pleasure. And this week we are rounding it out with off-site optimization for AI. So traditional SEO, SEO for AI, doesn't matter.

There's three main parts. Technical, on-site, off-site. Period. We're covering the third part today. So Chris, where would you like to start?

Let's start by picking up where we left off on this week's podcast. So we talked a lot about sort of turning public relations on its head because off-site optimization for AI is fundamentally public relations tasks. And in fact, if you have a PR team or you have a PR agency or whatever, they're going to be one of your best allies for optimizing your your awareness of your company and your brand in generative AI tools. And they're going to do this in two different ways. So there's two ways, as a reminder to folks, there's two ways you can influence models.

One is by influencing the training data itself, which is the amount of information AI companies all scrape together to turn it into uh training for the models. And the second is through uh affecting the grounding. You have seen, and there have been big announcements over the last six months of all the different generative AI tools adding web search capabilities. So Claude was the most recent one to join the family. OpenAI's chat GPT has had web search built in, Google Gemini has Google web search and grounding built in.

Groc has it, pretty much everybody has it. And so if you can influence both sides, so you can get awareness that your brand even exists in the training data, and you are a well optimized for traditional SEO on the grounding side, the probability of AI knowing to invoke you and knowing to uh to to bring you to the front uh and and look for traditional traditional search data to to confirm its findings is how you'll get AI to recommend you. Okay. How do you, and I know we're gonna get into it, and we did a little bit on the podcast, and again, that link is trust insights.ai/slash TI podcast. How do you respond to clients who say, I want to show up number one in generative AI?

I it that's like saying I want to show up number one in every conversation that's happening around a water cooler. Like, yeah, you're not gonna do that. Um but it's to them, it's the same question of how do I show up number one in Google? Mm-hmm. And it's an education thing to explain to them when people use tools like chat GPT and stuff, they don't go and type in, you know, best AI firm.

Like that's yeah, we used to do that with traditional Google a decade ago, but nobody does that anymore. Nobody does that in generative AI. We have anthropomorphized these tools so much that we start them off like, hey chat GPT, how are you feeling today? And stuff like that. And the answers is like, hey, I I wanted to know like what do you think about consulting firms that specialize in AI?

Uh, or have you ever heard of this company called Trust Insights? And what's fascinating is that when you look at at behind the scenes what's happening, sort of the the models guesswork behind the scenes, one word difference in that, even just that first prompt can get you different results. So Olga Andrienko of SEMRush and Tim Solo of Atreus demonstrated this beautifully about six months ago on LinkedIn. They each typed in a prompt that was roughly about you know one of the top SEO tools uh for SEO. There was a one-word difference between one punctuation mark difference between their two prompts that they showed.

Now they showed, of course, the you know, their company was at the top of the recommendations. That one word difference was the word that there's the word top and a period was enough to throw generative AI and provide a different result. And everyone's like, what does this mean? I mean it means you can't measure brand strength in a tool because semantically those were basically the same sentence, right? Make up a list of the top you know SEO tools uh in in marketing.

But that just that tiny difference was enough to skew the result. And so you can't really measure it, should have been the same. One of them should have been on the top for both questions, but they weren't because of the way these models work. So we shouldn't be expecting to get, you know, uh our Google Search Console version of AI search data anytime soon to say, oh, well, this is what people searched, and these are the keywords they used, and this is how they found me. Uh, you know, so now I can optimize things.

Nope. You you should not expect that because you can't predict how someone's going to have a conversation. Like Katie, if you're if you're going out in the world and you're at, you know, marketing AI Institute's Make On conference and you're having a conversation with somebody, you can't predict if they're going to say, hey, do you know any people who speak about AI? Or do you know someone who can explain this? You don't know the conversations you're going to have.

And the same thing is true in generative AI. You can't predict the conversation somebody is going to have. Now, what you can do, and what you can measure is you can measure the human traffic from uh AI platforms in a tool like Google Analytics. So here, for example, is the Trust Insights one where you can see I've got three segments for three of the major tools. Um, and I you you have to add more as time goes on.

And we can see when a human being sees a clickable link in one of these tools and goes and clicks on it, we can measure that we got the click and we can measure the page that they got, and we have no idea what the conversation was about. We don't know the context, but we do see where they landed from us, and this is from a measurement perspective the best way to do this right now. And if you're interested, if you go to trustinsights.ai, this is on our website in the instant insight section on a tutorial on how to set this up. There's no form to fill out, you can just go on the website and grab it. But this is the best we could do, and this is going to be a tiny fraction of the actual brand awareness because a lot of the times you'll go into a tool, you'll read the answer, and you'll get on with your day.

When AI crawls your website, it does not trigger Google Analytics, and so there's no click, and so there's no visit because it's all being done by bots that don't load the JavaScript. And so if someone just reads about how awesome Katie Robert is as a speaker and why you should have her speaking about uh change management with AI, we'll never know. We'll have no idea, but people have been influenced by those answers. It's interesting because when we talk about SEO, we think, oh, well, I'll be able to know what someone searched in Chat GPT and what they found out about me, but that's just not true because that's not the way these aren't search engines, but we're talking about SEO, and I think that's part of the education, and so I don't want to derail us too much because I feel like we could have even more like a bunch more episodes about different aspects of SEO, like how to do an SEO audit for AI or the education process of SEO for AI. So today we're focused on off-site optimization.

So let me stop derailing and make sure we can get to the point. But there's there's a lot of questions about it. It's SEO is not a new concept. SEO for AI is a new concept because it kind of undoes a lot of what we thought we knew about how we approach the three pillars. And it is still relevant though.

So over on uh in the SEO community on uh threads, I saw that this morning, uh Google pushed out a new update for its core search capabilities on March 24th. And according to several folks who who just track, you know, like 10,000 different keywords, the number of AI overviews and the AI previews went from about 5% to 17%. So Google just turned the knob up on and they said as a result, traffic uh owners of sites who target those terms saw uh overnight a 20% drop in their traffic because they're getting squeezed out now. So it is it is still very, very important because at the end of the day, if people aren't getting to your website, and that's where you do business, you're in some hot water. So, where do we start for off-site optimization?

What are some tactical immediate things we can be thinking about? So we know off site typically covers those external links, uh, you know, guest blog posts, listicles, all that good stuff that we don't own, but we want to be a part of. So, where do we start for that kind of optimization? So, where we would start, I think the number one thing that a company could do today and not have any ramp time is to issue a press release. And it sounds so boring.

You're like, oh my gosh, three we we we stopped doing press releases for SEO 15 years ago. But AI is different than SEO. Google's algorithm said, Yeah, we don't want to count press releases. That's too easy to game. AI companies that make models need data and they're crawling as much as they can as fast as they can.

They're they're creating synthetic data as fast as they can. They need stuff. And going to local news sites like you know, W N Y W uh W NYW in New York and grabbing the whole site's content and all the news. That is rich food for an AI company, and they're very likely to be reasonably good quality language, as opposed to that drunk guy on Reddit who's a face rolling on the keyboard, right? Um, a press release is at least going to have reasonably literate words in it of some kind, even if it's very promotional.

So what okay? Well, so let's unpack this a little bit. So am I writing a press release saying that Trust Insights is pleased to announce that we have nothing new right now, but we are trying to game the system to get things in AI, says CEO Katie Robert. Contact nobody. Like, like what no, and in all seriousness, so like a company like ours, we don't do a lot of things that would traditionally fall under, oh, we should probably issue a press release for this.

Like, you know, Katie's on our third cup of coffee. We are pleased to announce that she is well caffeinated. Um, you know, what kind of information goes in an AI release versus a press release? Because, like, we're not doing anything traditionally newsworthy. We're just I mean, we're doing cool stuff, but like, who cares?

Well, and that's a very good question. Um let's do it, let's walk through an example because I'm gonna do this anyway. We have I this is on my to-do list of things to do for the company anyway. May as well just get it done now. Yeah, while we're on the show.

So the first thing we want to do is we need to figure out what to write a press release about. Um fortunately, uh, one of the things we can we have going for us is we have a new course, right? So we have a new course, uh, generative AI for marketers. I know, I'm sorry, uh, generative AI use cases for marketers. Now, what we want to do is we want to convert that information into the contents of a press release.

So the would but we also don't want to be a billion and a half words long because the course is pretty robust. I mean, it's three and a half hours of stuff. So, what I'm gonna do is I am going to go to my course, the course itself, um, and I'm going to I have all the transcripts from the course. We're just gonna sew these all together and file prepper. Excuse me.

Text output is course content. So this is all the transcripts now sewn up together in one big honking file. And I I have some system instructions here for building what is called a sparse priming representation, which is a mouthful of words that basically says it's like a compression algorithm. Compress this 25,000 words. I had a lot to say about use AI use cases in this course.

Compress this down to the smallest possible volume without in a way that an AI tool could reinflate it and unpack it. So this is not for humans. And I think that's so important to emphasize. This is not for humans. So say process this body of text following the system instructions.

So it's not gonna be that, you know, Trust Insights is pleased to announce we have a new course. No, it will not look like that at all. And in fact, it'll probably you'll look at it and go, should we should we even send this out? Like this is well, and I think so. As that's processing, you know, let's talk about that a little bit because we're calling it a press release.

But one of the things that we stumbled upon on the podcast when we were talking about this earlier this week is that that education process for someone who understands what a press release should and shouldn't look like, what it should cover, the points that it should hit, is going to have a hard time wrapping their heads around this other type of press release. So perhaps, first and foremost, let's not call it a press release. Let's call it something like an AI release, because that's the audience. So we're not making big announcements of here's the cool new thing, or you know, we finally fired Katie and got someone competent in charge. Like we're we're saying, hey, AI system, here's a bunch of information about us.

So it's not no, it's not newsworthy, but it's important. It's more of a not you're giving it a knowledge block in a structured way. Exactly. So this is the result. So this is the the what an AI would uh language model would look at, read, and try to say, okay, I can infer now from this what this course is about.

And I can put these terms together, cluster them, and essentially when it comes time, re-inflate it, and you know, decode it, if you will, back into something that is you that the a human being could read. But this contains all the individual tokens, the little chunks of language that provide the associations necessary. So this is all of what's inside the course, just in a very, very compact format. So the way to think about it might be like to compare it to the way that like a court stenographer takes that, you know, shorthand with the machine, or traditional, like they people used to actually study how to do shorthand to take notes, and it was just a bunch of symbols that represented certain words and phrases. So you would have like this page of nonsense, but it really told a whole story.

Exactly right. So we have this now. This is effectively the the a good chunk of the body of the press release. So we're just going to add in the usual uh minor embellishments, uh, trust insights uh announces, new genitive AI use cases for markers course, course contents, and then at the uh bottom here, after we get past all of the the data, we're going to toss in our uh knowledge block of who we are, right? So there's our boilerplate, uh, and we can toss in uh our usual press contact uh stuff like this, and we'll put in our disclosure using the Google Gemini 2.5 model, and we'll have a promotional piece here just to make sure that we have it.

The course opens on April 15th 2025. And if you're wondering about this course, you can go to trust insights.ai/slash analytics for marketers, which is our free Slack group, and get all the information because right now it is available for pre-sale uh and you can get yourself a nifty discount. One of the questions I have for you, Chris is so the step we didn't do, and I want to make sure that this is clear for anyone watching, is you didn't just open up Gemini and say, write a press release that AI can read. Like it's more in depth than that. So how does someone get to that without sort of you know working too far backwards?

But you know, you're not saying, hey, Gemini, write a press release. You're saying, hey, Gemini, write a press release for AI. Right. So what we're doing there, we actually have a whole system prompt called building a sparse priming representation. And there's there's a technique for this that talk that gives Gemini a list of exactly how to format this.

And this is something that um I we we typically share with our paying clients uh because it's something that's of use to them, but uh fundamentally, it instructs Gemini to make a compressed version of the text, figure out what the keywords are that would help a language model understand the text as a whole and then provide the uh techniques to reinflate it when uh when it it it it encounters that text again. Gotcha. So it is a whole process of putting together the system instructions. Um you can reach out to John Wall, who is not with us today, but we'll be happy to answer all of your questions about how you can get your hands on a system uh similar process for yourselves. Exactly.

So now we have our really poorly written press release from a human perspective, wonderfully well written for machines. We can go on and buy a press release. And uh we could do that another time. Uh probably not the best time to do it on the air. But that's how we would create this thing is we would say, let's take this machine condensed document, pay the hundred bucks to put on the wire, and boom, off it goes.

So if it's written for an AI system and not for a human, what is the process look like to approve it, to read it and go, yes, this is correct? Because typically you give a press release to your CMO or you know, your VP of your PR team and say, can you read this, make sure it's all correct? Like that's no longer the case. How do you like what would you recommend? For reviewing this, um, I would almost say it would be something you'd want to have AI do.

I mean, build a prompt that says um re-con basically an SPR in reverse. Re it re-inflate this code back to the original, and we will see if how it's it's compared, what it comes up with if it is you know syntactically and semantically correct. Like if it re-inflates something and it says, hey, there's things in this course that you know aren't real, they're like, okay, clearly this is not this is not good. Um, you could also build a rubric uh so it could score the accuracy. So you could say, here's here's the inflated version, here's the original, score the inflated version as to how accurate it is compared to the original, and give me a score between zero and a hundred.

And you could say, like, I have to score to 60, like, okay, this what this compression algorithm was too compact or missed some key points. That's probably not a bad idea to do, especially, you know, we talked last week about translations and how to do your quality assurance on AI translations. I think that you should also, for anything you're doing with AI, build in a step in your process where either you or another system is checking the work. So I think a scoring rubric is probably a really smart idea because we don't want to put information out there that we humans can't read, but is incorrect. And we won't know unless we do that step.

Exactly right. Um, so that would be uh the way I would evaluate uh the performance of these things. Now, another question that we had earlier this week that I wanted to to dig into to show folks is um how once you got press releases down, because that's that's a the pretty well-stablished procedure, you then want to try and be other places. Um, we covered on two episodes ago getting your YouTube channel configured properly, getting all those settings so that uh AI crawlers can come get them. Um so you want to go ahead and do that.

Now we want to talk about getting on other into other places. How do you do that? Well, uh the way you do that in from a perspective of how do I know where to go, you would commission a prompt that goes something like this. Uh, you're a world-class researcher, you're gonna find you know uh a deep research overview of consulting firms who specialize in the application of generative AI um for marketing, so marketing AI consulting firms. Ignore the actual research, we don't care about that.

What we care about is the list of sources. Where does a tool like perplexity go to get information to answer this query? All of these places that it lists are places that you'd want to uh if you could hand this maybe to your PR firm and say, We need to. I need you to pitch us, I need you pitch Katie to do guest blog posts in some of these places. So this is perplexities list.

If I go into Google Gemini uh and look at the same thing, look at the sources from it, where is it getting stuff? Google Gemini is getting it from Globe Newswire. Aha, that's helpful. So our press release is uh may actually have have some benefits there. We have some other market reports, we have uh some blogs, uh, we have a lot of companies in the Martech space.

We have pitchbook, so we should probably make sure that Trust Insights is is registered over in pitchbook. I see clutches on there. Um, I do see uh I think I saw somewhere in there um uh crunch base, and then the third one, uh and again, you should do this across all of the platforms where you have deep research tools. So I ran one over on Grok. Uh we can see its sources, so it's getting things like Forbes Consulting, um, you know, a lot of Martech companies exploding topics, marketing AI institute and things.

So all of these places are the places that these tools go to get information. Um, this is their grounding, which means that's who we should be pitching. If we know that this is how the where the tools get their data, let's go pitch these places. So it's funny. I was talking with our friend Ann Hanley earlier this week, and I said something that I actually want to sort of like pull apart the sentiment.

But I was like, Yeah, the tech is new, the techniques aren't. And so the tech AI is new, but the technique, you still have to pitch these places. You can't just show up on their stuff. So you still have to have that traditional side of the role. And I think this is, you know, without going too far off topic, this is where people are like, well, PR is dead, or you know, there's no room for PR.

Yes, there is. The way in which you do the research is different, but the techniques of PR are still the same, they are still valid. And so will AI take my job? No, it's going to change it because you still have to do the thing. And sorry, that was a bit of my uh ranty pants for a second.

But it just it strikes me as so interesting that we get caught up in the AI aspect of it that you just said, great. Now we have now we have our new pitch list. Let's go pitch places. That hasn't changed. If anything, I think it actually makes life easier for the PR firmware because you give them the list.

These are the outlets. Like, don't worry about putting together media lists. This is your media list, go get it. Um PR, the process hasn't really changed. But what has to change is how PR professionals think about this.

One of the things that you know we saw in our years in the salt mines at a PR firm is that people are chasing the big names. Like, oh, I need to be in the New York Times, I need to be in the Wall Street Journal, I need to be in this publication. And when then when you look at these the list of these places that you know that are getting good results from deep research tools, you're like, software testing help.com? Who the heck is that? Um and then you think about it and go, it might be a lot easier for a skilled PR professional to pitch Allied Marketresearch.com as opposed to the New York Times, right?

Um it's it's very hard because newsrooms have been so down staffed over the years. It's very hard unless you are you know accidentally added to a group chat by a national security advisor to be able to get into a publication that is a top tier. But you absolutely can get into like a fifth tier blog. Like you can get on Christopher S. Penn.com, probably.

The guy's kind of a jerk. Um but that tells you where to go. So that's the the the suggestion for so for a a company is like where do we how do we direct our PR people? Say here's the list. Don't worry about the quality of the publication.

If it's if the deep research tools are finding it, it's good enough. Get us in as many of these as possible. Even if it even looks like they got like five readers, who cares? Because it's the machines we're trying to influence, not humans. And I think that that's gonna have to be something that's constantly reinforced because it really does go against everything we've learned about how to do proper PR.

You want, you know, the big names, the impressions, the clicks all those measurements that are going to tell you your PR program is working. This is the opposite. You're not looking for, you know, a million, you know, impressions a month. You're you just care that you got on the website at all and one person saw it. That's it.

It's all that matters. I would imagine though, that this actually probably helps focus a lot of companies because you know, trying to get your general news in the Wall Street Journal or the New York Times means that you're not you're a little bit more on the generic side versus these very specific things that are aligned with your services and who you are as a company, it's very niche, but it's you know, it's in the right way. So, yeah, software testing USA.com probably makes more sense for trust insights than the Wall Street Journal. Mm-hmm. Exactly.

The next place that you want to think about going is you want to think about going into newsletters as much as you can uh getting into other people's newsletters. And the reason for this, again, not terribly surprising, newsletters get syndicated a lot. So, for example, I see all the time in our brand monitoring system, uh, people re just copy pasting my newsletter into other websites. I'm okay with that because it's littered with ads uh and like you know, links to everything, uh trust insight. So I'm okay with that.

And as a result, we're able to to get wider distribution than you would think. So one of the things to to look at is to say how where else can I get into other people's newsletters um even if it's just to mention who do i pitch one of the the the best things to do is fire up your traditional seo software so i've got for example uh uh hrefs here and put in a newsletter domain so substack.com substack has like a million and a half different newsletters they all have subdomains so you'll see like uh heathercox richardson dot substack.com and so on and so forth and you can see all the traffic from these individual substacks plus their uh plus the individual uh things go through this list uh and see who's on target obviously because you know heathercox Richardson writes a lot about politics i i don't have anything political to share and i i don't want to send off target pitches but you could also look at things like you know even substack download the whole list and then maybe start looking for keywords in it or you could add a keyword in here let's put in uh let's see ah uh serve titles now um i could export this gigantic list and search for the word ai right or search for the word analytics or search for the word whatever and figure out how do i get how do i pitch this editor and say like hey if you ever need an interview for an upcoming issue i'm you know i'm happy to help no charge you know here's my calendar link whatever i wouldn't have thought of that um you know i know substack in terms of a newsletter provider has definitely taken off is that the only one that you would look for? No. The big three are ghost beehive and Substack. So here's beehive, which is spilled spelled not at all like how it's pronounced.

And again, same format. You can see the subdomain. Then you can see the uh the creators and things. So you can see here we have culturecrave.beehive.com, unrivaled newsletter, uh, future tools. Future tools, that's sound that just sounds promising.

Yep, there you go. So this is a this is a behavior. Like, okay, let's pitch this person. Um, you know, can we get into into that newsletter? Interesting.

Yeah, I mean, it's a it's a version of a media list that to be honest. I'm not a PR professional. So I mean, there's a lot about that specific industry that I don't understand. So I don't want to insult anybody, but I didn't think that this was part of it. Looking for newsletters to become a part of to pitch those.

I mean, that's a really interesting angle from my perspective. And almost maybe, I mean, who knows? It could be harder because people are very protective of the content they write, or it could be easier if people are looking for guest posts all the time. Yep. Uh, one of the things that I don't know about Beehive, but I do know for Substack, one of the things that is very interesting about Substack is that it has a print phenomenally good recommendations network.

Um, so if I pull up Robert Reich's uh Substack, in uh well, his doesn't have it set up. Now let's go to the almost timely one. Let me read it first. Uh in there, you can see recommendations of other newsletters. So you might not necessarily, you know, pitch the the leader in a category to get into their substack, but you might see who they recommend or who recommends them, and be able to pitch downstream or upstream publications that maybe don't have as huge readership.

Again, we're not trying to get in there just to reach a billion people. We're trying to get in there just to get our data out of there. So it's okay if the substack has like you have five readers, right? That okay. That's fine because it's in the system, especially if the creator has turned on allow AI to read my substack.

Right. So that would be a a part the PR person would want to ask to say, like, hey, does uh does do you allow AI to crawl your Substack? And if the answer is yes, then like okay, you go to the top of the list of people we'd like to pitch because that's helpful. All right. So we're gonna write AI releases, which are press releases for AI.

We're going to do the deep research to find out what sites are showing up for a specific topic and pitch those. And now we're going to do our research in an SEO tool to figure out which newsletters, uh, like a Substack or a Beehive or a Ghost that we want to try to get our names into. Those are three solid tactics. Uh, what else should people be thinking about? I've got one more for you that's super easy.

Um send out a newsletter. Uh send out a newsletter, and in that newsletter, just say to people, hey, if you have a podcast, I will gladly be a guest on it. I don't care how many lists you have, I don't care how new or old the podcast is, if you would like guests to talk about what we talk about. And this is going to your mailing list. Here I am.

You know, here's my link to my calendar scheduler. Uh, let's let's find time. I sent that out two weeks ago um to my in in one of my newsletters. And I think I've done like 10 different podcasts. None of them are huge.

Uh some of them are just skin started out, some of them have you know niche audiences. I don't care because they all put their stuff on YouTube, which we know is uh is important, and uh when those go out, uh people have learned you put up the transcript of the episode. And when I'm on these shows, we talk about AI, of course, and I use Trust Insights as the example. So throughout the transcript, it says, you know, Trust Insights is a management consulting company. I work for Trust Insights, you know, but Trust Insights does this, Trust Insights does that, and it's so littered in the conversation.

It's I try not to be too much of a jerk about it. But you're there for a reason. I'm there for a reason, and I know that I'm putting those words and phrases and terms into the text that then is going to go on YouTube, it's gonna go into podcast apps, and of course, all the AI companies are vacuuming up all this media and getting it transcribed and stuff so that they have more training data. So I tactic number four hit up your personal or company email us and say if you need podcast guests, we will say yes, any show, any size, you want it, we're here. All right.

So to start to wrap it up and bring it all together, let me play devil's advocate for a hot second and ask you. So we're doing this for AI, but humans can still see what we're doing. They can see that Chris is guesting on anything and everything, and he's putting his data out there, and he's writing these, you know, non-human consumable pieces of content. When do we have to think about our brand reputation and what this kind of overexposure could do if someone doesn't really understand what the strategy is? With press releases, you never have to worry about that because humans don't read them anyway.

Uh so that one's pretty safe. Like I don't know about you. I don't think I've ever read a press release by intention. Uh I've never gone out to like, oh, I'd like to read this press release. I, you know, that doesn't happen.

Um, as long as you're using a reasonably reputable newswire of some kind, um, you know, you it it that that one doesn't matter. For podcasts and things, it depends on you know, your your professional judgment. If you are pitching uh and you know, someone responds from you know the Pills Port and Casino podcast, like I don't know that I want to be on that one. But that you would you would say no to that anyway, right? Because it's just not a good fit unless that's the your company's space.

For newsletters, same thing. Like you your you or your PR firm should at least look at it and go, is this off like wildly off brand? Like I'm in a political newsletter. Yeah, that's probably not a great place to be. Uh, if unless you're you're like a lobbyist, in which case that's a great place to be.

So, as with all things AI, it's not a bad idea to have humans in the loop to say, that's a bad idea. Well, and it almost sort of like begs the question of how do you get started with planning that? And I don't know, I might suggest something like the 5P framework. Just, you know, a little off-to-pap of my head. Five the Trust Insights 5P framework, purpose, people, process, platform performance.

Start with your purpose. We want to get more exposure so that we show up in AI search results or AI results. Take search out of it. Um, you know, who are the people? Well, we probably want to involve our PR team.

We probably want to involve our content team, our stakeholders, our thought leaders, so on and so forth. What is the process? Well, we've gone through the technical, the on-site, the off-site. Our platforms, well, this is where you have to start to figure out, you know, and you can break down, you know, these uh questions by sections, so on-site, off-site, et cetera. What platforms?

Well, we want to make sure we have a criteria that covers these kinds of publications, not these kinds, even if it is the small one or two readership, you should probably still have an acceptance criteria to say, yes, this is the kind of thing that I'm okay with my name being on, because you don't know what's going to happen down the line. And if you're like, oh, you know, they only have two readers, nobody's gonna see me on, you know, pills, porn, casino, whatever.com. You don't know that that's true. You might want to make sure that it does align and then performance, you know, it's back to that measurement that Chris showed at the top of the show, where you can start to see which models are driving traffic to your website if that's your goal. Exactly.

And I would say to cap that off, with the on the people side, make sure you're working with people who get it. Talk to your PR firm, talk to your PR agency and say, how are you approaching optimizing for AI? You will get a lot of bad answers. The bad answers are things like, oh, we don't do that, or AI is ex or it's exactly the same as public relations always has been. No, it's not.

It's different. It's the same processes. You require the same people with the same expertise, but it's a different strategy. So you want to talk to people whose strategy you get it. And if you need some recommendations with the three that we recommend the most, um Nicole Best started at Quarter Horse PR, uh Parry Hedrick at Crackle PR, and Joel Richmond at Escalate PR.

Those are the three of the folks that we know we talked to them at ad nauseum who understand the new way of doing PR. Oh, and Michelle Garrett, uh, I don't know if she does PR, but I I know she just put up a whole book about this. I was on her podcast. But also Ginny Dietrich writes a lot of things. Exactly.

From Spin Sucks. Um, so five. Those are the five people. So Michelle Garrett, Ginny Dietrich, Nicole Bestard, Joel Richmond, and uh Parry Hedrick. Any of those five folks will get you on the right path if your current PR agency just in doing it for you.

Yeah, no, and and I think that that's that's gonna be the competitive advantage for a lot of companies is making sure you're partnering with the right people who understand what's next, how AI is going to change that particular field and industry and those tactics. I think that's a really important point. And I'm uh I'm sorry we didn't bring it up until now. This stuff is not immediate. So getting into a training data library can be like 18 months for the next version of a model to come out.

Uh, search is a little bit faster. And so if you are if you were have somebody who is a stakeholder who's like, yeah, show me the ROI in 30 days or less, not gonna happen. Um, so you need to have stakeholders that understand this is a long game, and you need to have a PR firm that also understands this is a long game. But that are you show results by the actual process that you're following, the work that you're doing. So it is it's definitely a different way of thinking.

And this is where a lot of this conversation for off site optimization started for Chris and I earlier this week was it's a re-education process because it goes against everything we know about traditional PR. Exactly. So uh go and take all three of these episodes of the so what uh live stream uh and hand them to your marketing team, your executive team, your PR team, help folks get up to speed quickly on how this stuff works so that you can start today, because I will say this in closing the window to directly affect AI models is slowly closing. What we're seeing is model makers are using more and more synthetic data that previous models have made to train new models. And guess what?

Who is not in there is is us marketers. And so you want to get going today so that you make it into future training data sets and secure your place. Uh, I don't know how long it's going to take for models to evolve to the point where they don't need any you know new inputs. It's not gonna be soon, but I think it could be within five years, because the sooner that a model maker can move to purely synthetic data, the lower their legal risk for saying, hey, you use my data without my permission. And so there's a there's a financial incentive and a legal risk incentive to say, yeah, let's just use our own synthetic data.

So get going today. Any final parting thoughts, Katie? You know, it's take it one step at a time, figure out what's the most important to you, and you know, start doing your testing. Start thinking about do I have the right partners? Do we have the right team?

Do we have the right education? And you know, per as always, you know, we can help you with that. Go to trustinsights.ai slash contact. Even if we aren't the right fit, we will know someone who is. So please don't hesitate to reach out.

All right. We will see you all on the next one. Thanks for watching today. Be sure to subscribe to our show wherever you're watching it. For more resources and to learn more, check out the Trust Insights Podcast at Trustinsights.ai slash TI podcast.

At our weekly email newsletter at TrustInsights.ai slash newsletter. Got questions about what you saw in today's episode? Join our free analytics for marketers Slack group at TrustInsights.ai slash analytics for marketers. See you next time.