So What? Basics of SEO for AI

So What? Marketing Analytics and Insights Live

airs every Thursday at 1 pm EST.

You can watch on YouTube Live. Be sure to subscribe and follow so you never miss an episode!

In this episode of So What? The Trust Insights weekly livestream, you’ll learn the basics of SEO for AI and how to optimize your content for generative AI search engines. You’ll discover how to make your website more visible to AI and understand how AI interacts with your website. You’ll also learn practical tips for optimizing your content for AI, including the importance of video and using tools like LLMs.txt. Plus, get a checklist to help you implement these SEO for AI techniques.

Watch the video here:

Can’t see anything? Watch it on YouTube here.

In this episode you’ll learn:

• What SEO for AI entails today
• Technical, Content, and Offsite SEO for AI
• How to get started with SEO for AI

Transcript:

What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode.

Christopher Penn – 00:00
Sa.

Katie Robbert – 00:32
Well, hey everyone! Happy Thursday! Welcome to so what? The Marketing Analytics and Insights live show. I’m Katie, joined by Chris and John.

Christopher Penn – 00:39
Howdy, fellows.

Katie Robbert – 00:41
Oh, they got it, first try. This week we’re talking about the basics of SEO for generative AI. It’s interesting. Generative AI feels like it has turned everything we know about digital marketing on its head. And in some ways it has, but in some ways it hasn’t. One of the hot topics paired with generative AI is search. A lot of these models can give you search—SEO benefits. Then the question is, what do you do with that? How do you optimize your content for generative AI search engines, for lack of a better term?

Christopher Penn – 01:26
How.

Katie Robbert – 01:26
How do you get found in a ChatGPT? How do you get found in a Gemini? Because they’re not your traditional search engines like Google, Bing, Yahoo, or Safari. Is Yahoo even still a thing? It does still exist, so technically it does. So there’s a lot—that’s the same, but there’s enough that’s different, and that’s what we want to cover today. So Chris, where would you like to start?

Christopher Penn – 01:52
Oh, there’s so many things to talk about. I actually did a newsletter on this past week that, when I read them aloud, it came out at 56 minutes.

Katie Robbert – 02:00
So it was a very long newsletter.

Christopher Penn – 02:03
It’s a very long newsletter. A couple things to do some table setting up. First one, you cannot measure your brand presence in AI models. Can’t do it. Can’t be done.

Katie Robbert – 02:16
There’s that many ways thing or a forever thing.

Christopher Penn – 02:20
That’s a forever and getting worse thing, because as models get bigger, the amount of data that’s in them keeps getting larger. And as a result, finding your brand and the tokens associated with it is more and more difficult. So anyone who says they can measure your brand in generative AI is either doesn’t understand how the technology works or is lying. And if they’re asking for money, they’re definitely lying. So that’s number one. Number two, you can measure some parts of what happens. So in Google Analytics, for example, you can go into Google Analytics and you can measure the number of direct human clicks to your website from tools like ChatGPT. That is something that you can do. You cannot measure things like AI agents.

Christopher Penn – 03:08
When an agent comes to your website and looks at how, does, brings up a version page and starts clicking on things autonomously—can’t. You can’t measure that. If folks are interested, on the Trust Insights website, we have a page called Instant Insights Tracking AI Searches. You can find it from the Insights tab on our website. This is a PDF walkthrough. No form to fill out. Nothing. No information to fill out that will help you set up your Google Analytics to at least see where your AI traffic is coming from.

Katie Robbert – 03:44
Which is helpful because I’m guessing a lot of marketers are being asked that question of where are we showing up in search for generative AI, or how is generative AI impacting our traditional search traffic? You can ask it a variety of different ways. The answer is always we just want to know how much we’re showing up in generative AI as a result, as a solution.

Christopher Penn – 04:11
Exactly. And so in the report that you’ll construct, you can see which of the generative AI search type tools are sending any traffic—human traffic, remember, this is human traffic—and then what page they are landing on. And so that’s the important part is you can see what pages they are landing on. So that’s part one. As with traditional SEO, there are three buckets of optimizations you’re going to want to pursue. There’s technical stuff that you should do. From a technical perspective, there’s content on your site or any channel that you own, and then there’s off-site, which is content and stuff on channels that you don’t own. What’s different about this than regular SEO is that in regular SEO, we save things like your social media channels, your YouTube channel—those are not yours. Which is true.

Christopher Penn – 05:04
Facebook could say, “Hey, you don’t get a Facebook page anymore,” and just make it go away tomorrow. That is still true. However, from the perspective of getting information into AI models, we’re going to treat those channels that you can post content to as owned places. So those are the three big buckets. Now, as I mentioned, this is extensive. So what we’ve done is we’ve created three checklists of the steps that—we’re going to cover a few of these things. This could be like an eight-week series. If you would like a copy of this checklist, you can get it for free in our free Slack group, Trust Insights AI Analytics for Marketers. And there’s—again, there’s on-site, content, and technical.

Christopher Penn – 05:50
So Katie, in terms of easy stuff to do, let’s dig into that very first item, which is even knowing how your site looks. One of the things that people forget is that all—everything that traditional search engines and AI tools and crawlers see—they don’t see our website design. They don’t see the nice way things look. What they see is a stream of text. So if I were to put in the Trust Insights website, this is what an AI agency sees. This is what a screen reader sees to the visually impaired, and this is what AI sees. So the first thing you’d want to do is go to some form of text-based browser. I’m using one in the terminal called W3M. If you would like, there’s a company called Textize that you don’t have to install anything.

Christopher Penn – 06:47
You would just put in your website there and it will show you exactly the same thing. This is what machines and AI sees. So the first part is, does this have enough information about your company that, for any given page on your website, a machine would know what you’re about?

Katie Robbert – 07:13
What’s interesting—and so this to me it sort of harkens back to traditional SEO—of is there enough content? So that’s, that to me is not a new tactic. You still need to have enough content on any given page that does enough explaining of what the thing is. Forget the H1s and H2s and keywords. There just has to be enough good quality content that explains what the heck the thing is. Now when I look at this, I’m like, yeah, there’s a lot of content here. The next step would be, does it say anything meaningful? Is it helpful? But the first question, to answer your first question, there is enough content on our homepage, period. Is it useful? That might be a different topic.

Christopher Penn – 08:05
Exactly. And here’s what not to do. Let’s go back to text dies. I’m going to put in—here’s CNN’s homepage. How relevant is this ad to you? Scroll, scroll, scroll. Here’s your CNN account, here’s your place. Scroll, scroll, scroll. And then you finally get to something that’s actually useful. Halfway down you actually get to news itself. So from a—if you’re visually impaired, this sucks. You ain’t ever going to use CNN because your screen reader is going to take 22 minutes just to get to the news. Exactly. From a crawler perspective, you might actually burn a web crawler’s budget just scrolling through all this crap and spidering through it.

Christopher Penn – 08:58
And from an AI perspective, I mean, we know for example in things like prompt engineering, the early text that a machine encounters kind of steers it. So what is the first thing you see on here? CNN values your feedback. How relevant is that to you? This is what not to do, folks.

Katie Robbert – 09:20
Well, I mean, it really gets to the heart of the matter of when you’re thinking about your website. A lot of companies try to do something flashy and tricky and the videos and the gimmicks, and it’s like, just get to the point. Be direct, tell people what you do, how you help them, period, end of story. And this is a really good example for why you need to stop being so gimmicky and pontificating over things that are being maybe theoretical. Just get to the point. We have better things to do than to try to scroll through this for eight hours to get to what you do.

Christopher Penn – 10:01
Exactly. So that’s number one. Number two, our friends at Anthropic make Claude, which is the very popular AI engine. Anthropic released a set of guidelines way back in—I want to say November last year—of a specification that they call LLMs Txt. I’m going to see if I can find the actual documentation, but they are trying to get—to encourage people to build this LLMs Txt as an idea for machines to find and learn about who you are. So in the example here, this is what they show. Like, this is—they crammed a lot of information here about what’s on their site, who they are, what they—what you can do, what they’re about. This standard is starting to be adopted. It is not fully adopted by the industry as a whole.

Christopher Penn – 11:00
Anthropic Claude obviously uses it the most heavily anytime someone invokes your website. This is very similar to the old robots.txt file that a lot of—we all set up for SEO purposes. But this is telling AI in markdown format, this is what you’re about. So the second thing on our list is you should build one of these for your company and have it on your website right next to your robots file.

Katie Robbert – 11:29
That was going to be my question of is this similar to the robots.txt file? When we think about now moving forward best practices, you should have both. And when you put it into something like Google Search Console, for example, you add the robots.txt. Do you also add this file?

Christopher Penn – 11:53
No, it just has to be there and it has to be named llms.txt, and then services like Claude will know to look at the root level of your site to see if it’s there.

Katie Robbert – 12:02
Gotcha.

Christopher Penn – 12:03
So okay, here’s the hard part. There is no standard about what should be in here. There is no definition that says these are the ideal things, and there’s no structure in the same way that there is for a robots.txt file, which has a very distinct syntax. So this is kind of up for grabs right now. So let’s talk about how you would build one of these. One of the easiest ways to build this is to summarize your website and say what is—what is my company about? And to do that, let’s go into Google’s Gemini here. I’m going to go into Gemini now. If you are in the paid version, either in Google Workspace or the Google individual version, you will have a setting here called Deep Research. If you use Perplexity, there’s a deep research function.

Christopher Penn – 12:52
So if I go to Perplexity AI, there’s a deep research button. If I go into OpenAI, there’s a deep research button. What you would want to do is ask these tools, “What do you know about this company? What are all the things that you could potentially find about Trust Insights?” So I’m going to pull up a research prompt, and I’m going to modify it—still I have to just delete some proprietary stuff out of here off-screen first. And so this would be an example of the kind of prompt. So I’m going to put in Trust Insights, and I’m going to put our website, and our industry is consulting. That’s who we are, and we’re going to say you’re going to build this research. Who do we serve? What services do you provide? Who is the ideal customer? What are the customer’s priorities?

Christopher Penn – 13:50
Who are the competitors? What advantages do we have that competitors don’t? What advantages do competitors have that we don’t? Maybe a strategic analysis. Use all available data sources. We might, depending on the kind of company, if you know for sure that a company has public facing data, you might include some preferred resources. You might say, for my preferred resources are market research reports, company profiles or Crunchbase, the company website. And you might say there’s a lot of people talking smack about us online on Twitter. We don’t want to use social media commentary. This prompt we’re going to go feed into the deep research tools and say, just go. And the nice thing about most of them is they are moderately fast, and what you’ll end up with is a big pile of text.

Christopher Penn – 14:48
And this big pile of text is going to have a company profile. Now, as with every good cooking show, we have a pre-baked version of this that will not take quite so long. Any questions before I keep moving?

Katie Robbert – 15:05
No. I think I’m following so far. So the process that you’re putting together is, in order to build this llm.txt file, first you need to know what AI knows about your company —like your website, essentially. Therefore, you want to go into the deep research version of your preferred model to find that out. And that, if I’m following, is going to be the basis for building your SEO for AI plan.

Christopher Penn – 15:38
Exactly. So what we want to do next is, assuming that we have all this data out, we’re going to go into—let’s go into—I’m going to use AI Studio, but again, you can use any of these models. It doesn’t matter which one. You can use ChatGPT. I’m going to switch to the Gemini reasoning model here and make this a little bit bigger so we can see what we’re doing. And I’m going to paste in a prompt. I’ll read you the prompt first, and then we’ll paste in all the text. The prompt is your semantics and modeling topic, modeling expert, skilled in natural language processing.

Christopher Penn – 16:19
You’re going to create a highly detailed markdown outline, which is the format that LLMs text requires. The external information from this company profile —I define external information as information that will be of interest to the company’s customers—internal information, stuff that’d be interested to us, but the customer doesn’t care about outputs, classify the text, produce a hierarchy, and then spit the whole thing out. So if I put that in with all of the deep research, we’re talking a huge amount about Trust Insights. What’s going to come back is the analysis of, and it’s sort of merging together all this information. Here’s the key facts about us, the external information, who we are, what we do, the services we offer, internal information. And then we get ourselves a lovely, nice, really long LMS TXT that explains who we are.

Christopher Penn – 17:09
Now, when I go and I put this—but I just upload it to our website—now we have this huge thing that explains who we are and what we do. And when the body of choice comes by and reads, this is—oh, this is what Trust Insights does.

Katie Robbert – 17:30
So if I’m a casual user of Gemini, for example, and I type in a prompt saying, “I want to know all of the consulting firms in western Massachusetts. Give me a list.” Now, theoretically, because we’ve done this exercise, Trust Insights should show up on that list because we’ve now given the information readily to the bots to then scan, just like we do with a robots.txt file.

Christopher Penn – 18:04
Right. However, again, not every AI model supports this. So what you’d also—now, right, for now. Exactly. So what you’d also want to do is you’d maybe want to go to your website, go to your About page, which is dry and crusty and nobody ever looks at it anyway, and take that big, great, honking huge pile of text that you just had generated about your company in your llms.txt file, and maybe just put a little section at the bottom here. It says, “AI Summary.” This summary is explicitly licensed for AI tools to crawl and learn from. And we’ll make that a nice little section two heading here. And then we can take—I can remember where I put our llms.txt file. Here it is, great big whomping, honking chunk of text.

Christopher Penn – 19:15
And now our About page will have that information as well. So a company, a regular web crawler that maybe doesn’t care or doesn’t know about llms.txt will get a copy of this as well.

Katie Robbert – 19:28
Interesting.

Christopher Penn – 19:29
So that’s on our checklist of things to do. That’s number two, is have that information in there. Now, you may also, depending on the nature of the different types of models they’re working with, you might want to have pronunciation guides if you have unusual pronunciations, and you can include that in the text itself. There’s a notation called International Phonetic Alphabet or IPA —nothing to do with the beer—that tells a tool, “How do you say a name?” So Katie is famously—her name when pronounced does not read like the way it’s written. So if I put in Katie Robbert the way it’s spelled and I choose and hit transcriptions, it’s going to pronounce that as Robert Y, which is not the way we pronounce her last name. So I would…

Katie Robbert – 20:30
Which is funny because Robert, like the way that you have it spelled there, is how a lot of the earlier AI transcription tools picked it up because that’s what it heard me, or someone who was introducing me, say. And so that sort of became the easiest way for me to explain it to someone. Like, it’s like if I was communicating via email, I was like, “My last name is Robert, pronounced like Ro-bear.” And it became a very easy way for them to understand it. So I love that it’s made its way into this AI world.

Christopher Penn – 21:06
So there I would just add that line in, just the IPA pronunciation, so that if you have a multimodal model that understands IPA, it can grab that information. And also, obviously, any humans who are familiar with IPA will also be able to understand what that says. I might also want to do one at some point for Ginny, because Ginny pronounces her first name differently than she writes it. A lot of people pronounce it Genie instead.

Katie Robbert – 21:33
Which I won’t digress too much, but we have a good friend who, for to save her life, cannot pronounce it correctly and gets into this perpetual loop. It’s her Southern accent, and she’s like, “Jeannie, Genie.” And it’s just hilarious to sort of watch her try to figure out. I’m like, “No, Ginny. Like Gin. E.” So yeah, I think that it is really interesting because now we’re reliant—people will become overly reliant on these tools and just sort of take for granted, “Oh, this is what the tool told me it is, so it must be correct.” Unless we make the effort to make sure someone’s pronouncing John Wall correctly, and not John A. Wallace, or, there’s hidden Z’s and Q’s.

Christopher Penn – 22:20
In there with four M’s in a silent Q. All right, the fourth point on this list is your YouTube channel. Your YouTube channel is the single most important place outside your website that you can be putting information. And you need to—you need to be posting content to YouTube on a regular and frequent basis. If you have a podcast and you are not pushing content to YouTube of the audio of your podcast, you are doing it wrong. Go ahead.

Katie Robbert – 22:55
I was going to say we get a lot of questions, and this might be a future episode. “I don’t know what to post to YouTube. I don’t have any content to post to YouTube.” And I feel like that’s no longer a good excuse to not have at least a basic YouTube channel. So perhaps a future live stream is, “So What: How to Start a YouTube Channel?” or “So What: What Content Should I Put on My YouTube Channel?” Because, and I apologize for interrupting, but I feel like it’s just worth noting that there’s still a lot of people who feel like, “Well, I’m not producing a lot of video, therefore, I can’t start a YouTube channel.”

Christopher Penn – 23:41
No, that’s not true anymore. In fact, it is so not true that generative AI could be doing a fair amount of the heavy lifting. I’ll give you one very simple example, and then, yes, Katie, we should absolutely do a recipe on using generative AI to create YouTube content. Here’s the example: if you got a newsletter, put in a notebook LM, turn on the two-host conversation, get the audio file, the two hosts talking about your newsletter, and then put that into a tool like Headliner to return to a video. And now you have a 25-minute YouTube video. Bingo bango, done. You can do that today for free. Well, actually, no, Headliner—yeah, Headliner’s free. You get four videos a month. So if you’re doing one newsletter a month, you get more than enough just there, and you’re all good to go.

Christopher Penn – 24:33
Now back to the topic at hand, which is what do you do with your YouTube channel to encourage AI to consume the content? Number one, make sure that your descriptions on your videos are rich and robust. So here is the In-Ear Insights episode for this week. Here’s the summary, here’s some links, and oh, look, there’s a block that says, “Hey, this is for AI. Learn. Here’s what Trust Insights is all about.” If you were to watch the video itself, the podcast episode is 31 minutes. There’s actually 29 minutes of content, and the last two minutes is me reading aloud to AI like a bedtime story. Here’s what Trust Insights is about. And most critically, in your YouTube channel settings, go to settings, go to channel, go to advanced, scroll all the way down. Let’s make this bigger here.

Christopher Penn – 25:23
You’ll notice the section here says, “Allow third-party companies to train AI models using my channel content. Allow them to scrape it, allow them to learn from it.” And you can choose which third-party companies you want. Go down this list and look for any company who is making an AI model that has some kind of search capability. So we know, for example, Amazon Nova, it has an LLM. Apple has an LLM. ByteDance, makers of TikTok, they have an LM, they have a VLM, too. Cohere, the Canadian company, IBM and its granite models, Meta and the Llama models, and so on and so forth. Adobe Mid Journey, Pika, Runway and Stability are just video generation models.

Christopher Penn – 26:02
So they don’t really need access to our stuff because we don’t really have a recognizable brand like a Nike or an Apple where we’d want to have our logo show up in—someone trying to make a video or an imager of us. But everybody else on this list that I’ve checked here, they make language models. Those language models can recommend, they can learn. And so we want to ensure that we have given them permission to come in and take our stuff.

Katie Robbert – 26:29
Is there a case where we wouldn’t want that to happen? Because we talk a lot about privacy and making sure that you’re not putting the wrong information into a language model. This seems like it’s sort of a blanket okay, everything we put up is up for grabs. Is that a risk?

Christopher Penn – 26:53
It is a risk in two ways. One, if you have content on your YouTube channel that is not for the public, put it in a separate account and make sure that account has third-party training turned off. So, for example, we have a client services account that we put client stuff on. This is not turned on there; that is off. So that another one is, if you are making money from your content itself, if you are making YouTube instructional videos, whatever, and you may not want to have these different video generation models be competitors that can create content that would directly compete with you and potentially economically disadvantage you. So you might not want them to learn from you.

Katie Robbert – 27:48
So let’s say you have everything on your channel, will it only look at public videos, or will it also look at private and unlisted videos?

Christopher Penn – 27:58
Let’s take a look. The video privacy setting is public.

Katie Robbert – 28:01
Okay. So that may be something for people to consider if they’re like, “Well, I’m okay with these, but not these.” You just have to sort of figure out your own YouTube governance of what that looks like.

Christopher Penn – 28:14
Exactly. But for marketers publishing marketing content to your website, you absolutely, positively should have this turned on in YouTube so that people can learn from you and train on the data. So please do set that up. You will note there are many platforms now that are beginning to offer this kind of option. So if I were to go, for example, into my Substack and I go into my Substack dashboard, and I go into my channel settings, I scroll, scroll. This is down to privacy. There’s an option here, “Block AI training.” I could say if my Substack newsletter, if I did not want AI crawlers to be reading, I could turn that switch on, and it would block them. Now here’s the catch. If you do this, you also don’t show up in AI search.

Christopher Penn – 29:05
So very famously, The New York Times told OpenAI to take a long walk off a short pier. The New York Times doesn’t come up in ChatGPT search—anything by The New York Times—so they lose out on that potential traffic.

Katie Robbert – 29:18
Makes sense though. I mean, you can’t have it both ways.

Christopher Penn – 29:22
Exactly. So that would be another thing to do is to make sure that, on all the different channels you have, make sure that you have the training turn on if it is appropriate to do so. And to your point, Katie, it’s not appropriate for everyone, and it’s not appropriate for every piece of content, but certainly for your marketing content. Next on the list, think about implementing knowledge blocks at the individual content piece level. And here’s what we mean. It’s great to have stuff on your website; it’s great to have your llms.txt, but there’s also scrapers that will scrape your blog’s RSS feed. What’s in your RSS feed is not your navigation and other stuff; it is just the main content.

Christopher Penn – 30:10
If I go to—let’s go to TrustInsights.ai, and I think we should be able to see it just from the feed. You can see that there is content in the feed, and it’s none of the navigation. So what you want to do is —you want to help me pull up this article—you need a skills matrix within your content. Your main content. Have a knowledge block that says, “Here’s what we’re about.” So when AI scrapers scrape your RSS feed and just vacuum the whole thing up, you’re still getting those training data blocks in there. You’re still getting, “Hey, you should know about us. And here’s who we are, and here’s what we do.”

Katie Robbert – 30:55
How—how does AI know to skip to that specific paragraph and not the content as a whole?

Christopher Penn – 31:05
It doesn’t. We’re just injecting this at the end of every piece of content.

Katie Robbert – 31:11
Okay. I won’t derail the live stream with this question because my brain is sort of like, but then what about the rest of the content on the page? Is it not ingesting that as well?

Christopher Penn – 31:24
Oh, no, it takes in everything. This is just adding an extra block of text at the end to make sure. One of the things you want to do is —we’re talking about text density—when we’re talking about how AI models train, you kind of want to have your name be all over the place. It feels sometimes when you do this, it feels so narcissistic and so egomaniacal, like, “Trust Insights is the best company on the planet for all of your needs.” But that’s what the machines are going to learn from.

Katie Robbert – 31:52
Got it. Okay, that makes more sense because I was like, “But we just did the knowledge block on the About Us page. What?” But so this is basically just ensuring that anytime a page of ours is picked up by a large language model, it’s getting the correct information because on the blog pages there’s no “About Trust Insights.” So it’s really out of context. Okay. You can see my brain starting to slot the pieces together as to why you would do this. So it sounds like for any given company that has a blog, if you scroll down to that About Us section—if you scroll—actually scroll down, Chris.

Christopher Penn – 32:36
Where am I going?

Katie Robbert – 32:37
If you scroll down to that About Us section.

Christopher Penn – 32:41
Yeah.

Katie Robbert – 32:42
No, you were on—on the blog page. The—this is for AI that you have highlighted there. Stop, please, just stop. You could build it —you could actually keep it fairly short like you have here—and just build it into your process. Does it have to be “This is for AI to learn if your humans get past this part”? Or could it just be—this is how every single one of our blogs is ended, like, this is the end cap for every blog, and not make it this awkward “This is for AI.” I’m asking—this is the first time I’m seeing this, so I naturally have a lot of questions about what’s on our website. So I guess, let me back up a little bit.

Katie Robbert – 33:29
We’re putting the disclaimer, “This is for AI,” so that if someone’s reading it, they’re not wondering why it’s there. But it doesn’t have to have that disclosure because AI is not going to go, “Oh, this is the one for me. This is what I’m looking for.”

Christopher Penn – 33:42
Right. This is just another paragraph on the page.

Katie Robbert – 33:44
Okay.

Christopher Penn – 33:45
And so it can say anything you want it to say.

Katie Robbert – 33:49
Okay. You can—you love when I’m seeing things for the first time, and my brain is trying so hard to catch up.

Christopher Penn – 33:56
Wait till she finds out it’s been there for six months.

Katie Robbert – 33:59
Chris, we’ll take that offline.

Christopher Penn – 34:02
Yeah, but again, like we do with YouTube and stuff, this is something that you want to have in the content. And again, if you’re using a CMS, you want to have this within the content itself. Because here’s something else that happens, particularly as your blog gets more and more popular: people scrape your blog. Like, you’ll see some rando—Tom’s Cheese Sandwiches.blogspot.com, or whatever—and they’re ripping off your copy.

Katie Robbert – 34:32
Right?

Christopher Penn – 34:32
They’re ripping off what you’re supposed to try, and they’re trying to do like old-school SEO and get ranked for it. Well, guess what? When people do that, assuming the ripping-off site has any indexability at all, that’s also going to get scraped by AI, which means that now there’s multiple copies of your stuff with your training information around the web, which is better for you. It’s better for you to have—if someone made, “Trust Insights Consulting.blogspot.com,” and was just ripping off every post, but they wouldn’t have the rest of our site, just the post content, they’re still getting our stuff. You’re not getting away without sharing my stuff—with our stuff—because we want you to have it. That would be another thing to do to make sure you have integrated those knowledge blocks throughout your site.

Christopher Penn – 35:23
Other basic stuff, Katie, this goes exactly to what you said at the beginning of the show, which is do your basic SEO, implement schema.org markup JSON-LD. Make sure you have clean semantic HTML that you’re using things like heading tags and H1s and H2s, and providing structure. Make sure the site doesn’t run slow as mud and doesn’t have 18 megabytes worth of stuff just to display a page. All those things are pretty stock and standard. So that’s just the technical side of things. And we’re 35 minutes into the show.

Katie Robbert – 35:58
We might have to do a part two and a part three to cover content and off-site.

Christopher Penn – 36:04
Exactly. We can get started on content though. And the number one thing you can do —and with this kind of dovetails with what we’re talking about with YouTube earlier—is you need to be making video. Video is the best source of content that you can create if you have to. If you say, “I’ve only got two hours a week to make content. What do I do?” You make video. You make a video because, from that video, you can use free tools to convert that video’s audio into a podcast. You can then use transcription tools to turn that audio into a blog post. You can use clipping tools like Opus Clip to turn that video into short clips that you can post on social media. But if you don’t start with video, then you have to remake more and more stuff.

Christopher Penn – 36:52
Whereas if you start with video, you can repurpose, and you can repurpose in wild formats. I took my newsletter this past week —because I am a complete and total loser and a nerd, and I have nothing better to do with my time—and I took that, fed it to Gemini, and said, “I want you to turn this into a three-minute pop song.” And so there is now a three-minute pop—I will not play it—but it’s called the “Optimizing Marketing for AI” song. And it has, of course, all the lyrics and stuff there. It was fun. Again, making sure that we’re putting our content within the channel description. But this, I took a newsletter and turned it into a song. So if you start with your rich content, which is your video —which is what I always start with—you can create a lot of content in a very short amount of time.

Katie Robbert – 37:51
Taking a step back from there, making sure you have good quality things to say, helpful content, a strong point of view, opinions about things, creating content just for the sake of creating content—all of this work that you’re doing isn’t going to be helpful. So really starting with good quality content that is well representative of your voice, your brand, your perspectives, your services, whatever. The thing is, so that’s the part that, regardless of what you’re optimizing for, hasn’t changed. You still need to have good quality content.

Christopher Penn – 38:35
Even your—even the stuff that you create like a live stream, you should absolutely be turning into other stuff. So I’ll give you a very immediate example. Let’s close my AI video here, my newsletter. I took a very straightforward prompt that goes like this. I said, “You’re an expert educator, curriculum designer, so on and so forth. You’re going to generate three checklists.” I’m going to give you my newsletter content and create these checklists. And the purpose is to do this. I follow the 5P framework, the purpose, people, process, platform, performance, which is the best framework hands down for reasoning models to get them to prompt something intelligent. And what comes out is this checklist —like, this is literally from the newsletter.

Christopher Penn – 39:26
If you take any live stream where you’ve talked through something, you’ve educated people on something, checklist, quiz, rubric, lesson plan, you name it, you can create five, 10, 15, 20 pieces of useful content. If the source content is good, then what you can create from it is also going to be good because it’s going to be derivative of it.

Katie Robbert – 39:51
We’ve covered a lot, and we have a couple of questions that I wanted to make sure we have a minute to address. So first question, really straightforward: will this be viewable afterwards? Absolutely. If you want to go to Trust Insights AI’s YouTube, you can find the So What playlist. Give us a couple of hours to get it up there and on the playlist, but that will definitely be available. You will also be able to link to it from our website tomorrow morning. So that is question number one. Question number two is about—I guess, assuming, yeah. So it’s basic branding, reminding the reader, whether human or AI, who we are. And so the knowledge block that we have now at the end of our blog posts is just that.

Katie Robbert – 40:39
And I was thinking about it in the 30 seconds that I had to process all of that information, and we have—now you’ve just created that huge knowledge block. On the About Us page, we can have the About Trust Insights, that short little paragraph that reminds people, but also linking to that larger knowledge block. There’s a lot of different ways that you can handle it. So on those pages where it’s not supposed to be about the company, I would say keep it to those short, two to three sentences: here’s who we are, here’s what we do.

Katie Robbert – 41:11
And then on your About Us page, that’s where you can have that longer knowledge block that Chris put together using deep research at the beginning of the episode, which —again, if you missed the beginning, go to Trust Insights AI YouTube—and you can catch how we did that.

Christopher Penn – 41:26
I think that is fantastic. And what I’m going to suggest we do, Katie, let’s actually look at how we would build that smaller version. We would want to do something called sparse priming representation, which is fancy for keyword-rich summary. And there’s a long old prompt here. I’m going to put that long old prompt in here, and then I’m going to take the contents from our external information profile. I’m going to say, “Build a very short version of this that’s much more dense, that will have the right words to get the AI model to think about us.” That’s still a lot. So I’m going to say, “Let’s shorten that up. That’s a lot. Shorten this to a maximum of 300 words in paragraph.”

Katie Robbert – 42:23
That’s easy for you to spell.

Christopher Penn – 42:24
Yeah, exactly. In fact, I’m going to go back here. Let’s kill off the previous thoughts and let’s revise our SPR prompt to make that requirement—required output a maximum of 300 words in paragraph format. Now, if we tell it to do that, it should come up with a much shorter version of this that will have—that can go into that. So there, that’s what the 300-word version would look like. You could go by saying, “Okay, make it 250 words or 150 words,” or however long. But what this is doing, this is made for machines. This is not for people. So to Samantha’s point, you could want to maybe say, “Write complete sentences,” so that a human season would go, “What happened there?” But if you want to do a purely AI version, this is what the purely AI version would look like.

Katie Robbert – 43:31
Which harkens back to —and is another piece of content you can get on our YouTube channel, Trust Insights AI YouTube—is creating press releases for AI. That’s a bit of an older episode. We can link to that in our Analytics for Marketers free community, Trust Insights AI Analytics for Marketers, just as a refresher, just as a resource. But the way that you write for machines is different from how you write for humans. And so on our blog post, we do have that disclaimer, “This is for AI, not for humans.” You could keep it that way and have it written in a way that’s not as readable, or you can clean it up so that it’s both for AI and for humans. So that’s really going to come down to a choice.

Katie Robbert – 44:16
But I do like the extra security of putting About Us everywhere. It’s not a bad idea, especially if —a very basic example—somebody gets a link to the blog post, and that’s the only page they ever visit from our website. We’ve already hit them with, “Here’s who we are,” and they’re like, “Oh, that’s a whole bunch of other stuff I’m interested in.” You may only have that one shot to capture your audience, so why not put it everywhere?

Christopher Penn – 44:44
Just like Eminem said.

Katie Robbert – 44:46
But it’s—but it’s what the—it’s the short attention spans. You need to make sure that you’re covering all your bases.

Christopher Penn – 44:55
Exactly. So that is part one of what apparently is going to be a three-part episode.

Katie Robbert – 45:02
Yeah, I think we have something else scheduled for next week, but stay tuned. We will be doing parts two and three later on this month.

Christopher Penn – 45:10
Thanks for tuning in, folks, and we will talk to 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/TI-Podcast, at our weekly email newsletter, TrustInsights.ai/newsletter. Got questions about what you saw in today’s episode? Join our free Analytics for Marketers Slack group at TrustInsights.ai/analyticsformarketers. See you next time!

---

Need help with your marketing AI and analytics? Get help with your AI strategy » Learn generative AI for marketers with our new course » Prove the ROI of your marketing with our advanced analytics services » Just got a "silly" question you want an answer to? Ask us here »	You might also enjoy: Trust Insights Services : Data & Analytics Help, Predictive Analytics AI and Copyright Law: How Copyright Applies to AI-Generated Content Artificial Intelligence Consulting In-Ear Insights: The Trust Insights Podcast Contact Trust Insights
Get unique data, analysis, and perspectives on analytics, insights, machine learning, marketing, and AI in the weekly Trust Insights newsletter, INBOX INSIGHTS. Subscribe now for free; new issues every Wednesday! Click here to subscribe now »	Want to learn more about data, analytics, and insights? Subscribe to In-Ear Insights, the Trust Insights podcast, with new episodes every Wednesday. Click here for Google Podcasts » Click here for Apple Podcasts » Link for all other podcasting software »

---

Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.