How To Build an AI Search Strategy, Part 2: Top-of-Funnel Content

Michael 00:00 – 00:02

There's a test that we use internally, which is

Michael 00:02 – 00:31

You know, could you take this piece of content, take your logo off, slap your competitor's logo on this content, and could anybody notice that it doesn't fit their business? If the answer is yes, it could be their business or yours, there's no difference. I would say that that may be commodity content. And I would be pessimistic that there's an opportunity for that piece of content to achieve the ability to be cited

Michael 00:31 – 00:45

when users are asking informational questions, which is the whole goal.

Michael 00:45 – 01:11

Hey, welcome back to The Search Signal. I am Michael Transon. I'm the founder and CEO of a search marketing agency called Victorious. And here at The Search Signal, our goal is to help marketers understand and also take action on what is happening in the world of search marketing. So we have been running search campaigns for well over 13 years. We also work on hundreds of websites every year for some of the biggest and also some of the fastest growing brands out there.

Michael 01:11 – 01:37

And we bring what we learn from those experiences, plus a lot of our own first-party data in research to help you with your own search work. And today is part two of our four-part series on building a full funnel AI search strategy from scratch. In last week's episode, we built the infrastructure for an AI search strategy. And then this week, we are going to start at the top of the funnel, which is informational content.

Michael 01:37 – 01:54

And how we can also influence both humans and AI systems to cite our brand's content when users are trying to learn more about the operations and the industry that we work in. So before we get into today's episode, let's start by talking about what happened this week in the world of search marketing news.

Michael 01:54 – 02:22

Okay, let's start with Reddit and its influence in AI answers. So one of the AI visibility trackers this week, it's a company called Prompt Watch, put out some data that was showing that Reddit's share of voice and share of citations in ChatGPT search dropped significantly in the middle of August. Prior, it had been sitting at just around 4% of everything that ChatGPT had cited, which honestly is a really big share for just one website,

Michael 02:22 – 02:50

but inside of about four days, it had dropped to only about a half of a percent. And it looks like this is only happening on ChatGPT. And what I mean is for example, Google's AI surfaces is showing Reddit drifting down, but a lot more slowly. Nothing like what we're seeing in ChatGPT. I would also say this report comes from a single company and a single tracker's data. They've said themselves it shows

Michael 02:50 – 03:19

that this is happening, but not really why it's happening. And also, neither openAI or Reddit has really said anything about this. But I would say OpenAI might be changing how much weight right now it is giving to Reddit. And this matters a lot because a lot of AI visibility strategies over I'd say like the last year have leaned a lot on things like Reddit seeding, which is you know, going out and getting your your brand effectively artificially talked about

Michael 03:19 – 03:37

in Reddit threads and was doing so because ChatGPT loved citing Reddit. But if this is a durable change that we're seeing from the way ChatGPT decides to source Reddit, if you're doing those strategies, I think that you might seriously need to reconsider how you're approaching them.

Michael 03:37 – 04:05

The next update this week comes from Google and it's continued work to penalize scaled AI content abuse across the whole internet. So Google ran its third algorithm spam update of 2026. And this one was very fast. It started on a Tuesday and then it was done by Friday morning. So for context, these types of algorithm updates usually roll out in a couple of weeks. So all of that movement

Michael 04:05 – 04:29

just got compressed into about three days. the update itself was global. It was across all different languages, and it seems to be aimed at sites that are violating Google's spam policies, which these days prominently includes what they are calling scaled content abuse. That topic is going to come up again later in this episode, but scaled content abuse is very, very prevalent right now. And

Michael 04:29 – 04:58

it is because it is extremely easy and extremely cheap right now to produce content at a very massive scale. Google seems to be actively working to identify and also penalize the websites that they would say are abusing AI to produce inordinate amounts of content. So if you or your agency are producing significant content at scale, I would strongly recommend and caution you to just pause and

Michael 04:58 – 05:15

maybe reconsider this approach right now. And I would also, because of that, expect Google to just get more and more aggressive on penalizing companies that are using too much AI to produce too much commoditized content.

Michael 05:15 – 05:43

And the third update this week is regarding ChatGPT launching paid advertising infrastructure. So it's finally happening. OpenAI made two ad moves inside of this last week. First, ChatGPT ads expanded to 31 different European countries. They also expanded to Brazil and Mexico, and they also shipped a bunch of performance marketing infrastructure. We saw automated bidding and view-through conversion reporting. Those are the two big items. They're saying

Michael 05:43 – 06:12

tens of thousands of marketers have advertised since the US tests started back in February. And right now the ads only show to free tier users. And the reason I would say I care about this one is it connects to something that we're gonna talk about a little bit later in today's episode, which is that these platforms like ChatGPT are designed to answer your question without ever sending you to a different website. And the exit that they eventually will want people to take

Michael 06:12 – 06:41

is a paid exit. You can see that future starting to show up in ChatGPT ads release notes. So what I would probably do, don't do anything expensive yet, but it's probably time to start thinking about how paid placement inside of these assistants and your or organic AI visibility are going to divide the work in your funnel. And if you have the ability, maybe a small test budget, if you can go through the application process and start to test it before

Michael 06:41 – 07:10

the self-serve fully opens up and the advertising prices start to settle. So I picked these three news updates this week on purpose because they're all versions of the same situation, which is that these platforms are controlling fully the answer surface. And what they cite and also what they sell on it can really change in as little as a week. But the thing that you control is the content on your own site and the authority.

Michael 07:10 – 07:14

behind it. So that is what today's episode is about. So let's jump

Michael 07:14 – 07:18

into things.

Michael 07:18 – 07:45

Okay, so quick catch up if you missed last week. Last week we discussed how to build the infrastructure and also the strategy for a full funnel AI search strategy. And we went through taking an inventory of your entities, which to remind you is just like the products and the services and the capabilities that your company makes money from, and then tiering them, mapping keywords and prompts to each of them,

Michael 07:45 – 08:11

and naming an anchor page, a commercial page for every one of those entities, which is effectively the page that sells or advertises that thing. And then we also talked about the plumbing underneath all of that, making sure that your website can have AI crawlers reach it and read the site. We also talked about using schema to declare, you know, who you are. And then we also talked about getting corroborated off of the site. So if you missed that episode, I would

Michael 08:11 – 08:37

really recommend you go back and listen to it because a lot of what we're going to talk about today assumes that you've seen that and the foundation that you have is in place. there are also two things that I want to talk about from last week before we move on. We talked about picking prompts to track, and we also talked about making the site discoverable and also indexable. both of those topics got probably less time than they could totally fill. There's just not a lot of room to get into the details. Like prompt selection could be its own

Michael 08:37 – 09:04

own fully dedicated episode. And honestly, probably at some point on the show, it will be. And then discoverability. I mean, there's an entire discipline, decades-long discipline of technical SEO that lives behind that one word. And we just got to compress it down to some checklists. So, what we gave you is a really good starting point that you can definitely use and build on. But if you want more information on things like indexability and discoverability and crawlability to improve the site's performance,

Michael 09:04 – 09:33

we're going to be releasing more episodes in the near term that're going to dig into those things deeper. okay, so, but let's talk about today. You've got your entities and you've got your anchor pages. The next decision is going to be what content we build around them. And we're gonna be starting at the top of the funnel today with informational content, which is the blog posts and the guides that your buyers are going to be reading when they're just trying to understand a problem. So

Michael 09:33 – 10:02

Let's start. Here's the thing about informational content. For basically the entire history of search engine optimization, the deal that we were running with was pretty simple. You write a post on your site that answers a question that people are searching in the form of a keyword. It hopefully ranks. People click on it. They land on your site. And then, you know, some percentage of them eventually become customers and buyers. And traffic was how you measured all of that.

Michael 10:02 – 10:32

Because traffic was like what the whole model was producing. And what has been happening and today, and this has honestly been going on for close to a decade, it just didn't start with AI, is that people are just clicking less and less in search. Somebody searches, they get what they need and the answer right there on the results page. And you know, oftentimes they never actually go and visit anyone's website. Today, about two thirds of Google searches are ending without a click on anything.

Michael 10:32 – 11:01

And with AI answers, like ChatGPT and other systems, it's kind of the whole design. It's like these systems are built to resolve the question inside of the conversation you're having. And none of them really have any incentive to send that visitor to your site, right? The platforms want people staying on and engaging with the platform. At some point, like we talked about in the news today, they're gonna want users to leave, I think through ads, because that's where the money is. But

Michael 11:01 – 11:29

if the click is going away on some of these types of informational questions, our question becomes why publish informational content at all? The answer I think is that the informational content that's hosted on your website is still doing a very important job. It's just a very different job than the one that we have been using it for years before this. And here's how I would think about the role of informational content in building your

Michael 11:29 – 11:56

AI search strategy and getting customers from AI search tools. Your informational content now has two distinct readers. The first one is the one that has always been, which is a human. The fact is, organic search still drives the majority of traffic to most websites, way more than AI tools are referring today. So real people are still finding and reading this content.

Michael 11:56 – 12:25

And you still need to be writing for them. But the second reader is the AI system itself. And there are two moments where those AI readers show up on your site. One is when these systems refresh the giant body of content that they are learning from, which is called the corpus. And that happens on some cadence, right? Depending on the platform, it could be like quarterly-ish. The other one is when somebody asks a question.

Michael 12:25 – 12:53

And then the system will fire off live web searches to assemble the answer. And your content either gets pulled into that pool for the citation and the answer or it does not. and the point I'm trying to make here is that in this stage of the AI search strategy, you're building brand awareness with the system, arguably more than with the buyer and the human. Every solid piece that you are publishing

Michael 12:53 – 13:22

on a topic or an entity teaches Google and the LLM that your company knows this subject, right? That is topical authority. And the system can treat you as a credible source because you've covered the subject very well and consistently across a very broad body of work on your site. And the reason why this matters is the actual hypothesis that this whole series is built on.

Michael 13:22 – 13:51

Which is that when a question or a conversation with AI turns commercial, when someone finally asks, like, okay, who should I reach out to? Which company about this, right? That system that already trusts you on the subject as a whole is much more likely to put your name as a mention, a brand mention in that commercial answer. And this is a working theory, but it's built on how these systems are described to work. And

Michael 13:51 – 14:05

what we can measure does seem to continually be pointing in that direction. And I do expect in the future there to be more research on this. But in my opinion, this is probably the most defensible bet that we can make in AI search right now.

Michael 14:05 – 14:35

Okay, now that we've established that informational content is a dual audience approach in modern AI search strategy, humans still click on and still read your content, while also AI systems are now ingesting that same content and it can influence those systems' opinions about your brand. Let's talk about where AI itself does and does not fit into actually creating this content. So

Michael 14:35 – 15:02

as you're well aware, probably AI has made the production of content essentially free. For like 20 bucks a month, you can get a subscription and you can produce more content in a single day than you shipped for the entirety of your career. Which means that for everybody, volume really has stopped being an advantage because virtually anybody has access to an AI assistant that can do this. And it's

Michael 15:02 – 15:27

already very much happening on the internet. And we have a lot of data to prove this. About half of the new articles being published on the web right now are AI generated. And roughly three-quarters of new pages have at least some AI content and written text inside of them. So the temptation in this, and I I get pitched versions of this constantly, is to just

Michael 15:27 – 15:56

point the machine, the AI assistant, at your list of content ideas and then just start to publish everything. Because if content builds authority and content's now free effectively, why not just do it all as quickly and as cost efficiently as possible? Well, here's what's been happening to companies that do that. Let me take an example. There is a website out there called National Today that had done this. They built out a section of their website with over 800,000.

Michael 15:56 – 16:24

AI generated articles on it. It was essentially a bunch of news, local news written by nobody. And for a while it worked very, very well for them. The site was starting to rank really effectively. It was getting picked up in things like Google News. These were all very positive indicators that this scaled content was a really good idea. Then Google rolled out an algorithm update. Google hit it with a manual penalty for scaled content abuse.

Michael 16:24 – 16:52

And the entire section got totally de-indexed out of Google. And then their citations in ChatGPT just totally disappeared too. It tells us, and we know this, like some of the LLMs lean on some search infrastructure to find their sources. So when your website is going to get penalized for AI abuse, other systems theoretically and often do inherit that same change. And that's a obviously crazy example, but this isn't also a just

Michael 16:52 – 17:21

spam site problem either. We all are familiar with Business Insider. They spent last fall pulling down dozens of pieces of content from their website after they had figured out that one of their freelancers didn't actually exist. It was AI written work that was getting filed on their website by a made-up byline. And then there's also Deloitte who ended up having to refund part of an Australian government contract because a report

Michael 17:21 – 17:46

published content on their website with citations to studies that were never written because they used AI and they had no human in the loop. And every one of these cases is just an example. It's an example where it's production for content was easy and fast and cheap. And they thought they could just eliminate the human in the loop, and each of those sites lost

Michael 17:46 – 18:14

major credibility and literal search performance in traffic from their websites. And they ended up being way worse off than when they had started. And these are not unique stories. These happen all of the time. Here's a fact that I think is very important to acknowledge and also internalize when we consider whether AI-generated content can be performant over human-written content. As of right now, across all of Google search.

Michael 18:14 – 18:43

Almost nine out of 10 websites that currently rank on the first page of Google, those pages that are ranking are still human written, even with half of the new stuff on the internet right now being AI generated. Before we talk about how we should be utilizing AI to build our informational content strategy so that we have a strong influence over AI systems perception of our brand in our space.

Michael 18:43 – 19:08

As an authoritative and as an ideal candidate for, brand mentions when we start to see users ask commercial recommendations. We have to understand how these AI systems are designed to retrieve content to help support the informational questions your buyers might be asking them. Only once we have that, we can really start to create content that matches the way that these systems work and effectively influence them.

Michael 19:08 – 19:37

I want to start by talking about what good means to these systems because it is much more specific than just writing great content. And I want to work and talk through how the machines, these AI systems, read your page first. Because once we understand how they read our content, we can start to set our own standard of excellence that we need to achieve before we can actually start to draft the content itself.

Michael 19:37 – 20:05

So when an AI system reads and evaluates your content, it breaks the pages into passages. Typically it's a cunk of like a couple hundred words, a few hundred words, usually at most. And each of these chunks gets judged on their own merits, and then in head-to-head comparisons against chunks from other sites that are also covering the topic from people like your competitors. You may have probably heard this called chunked retrieval. So

Michael 20:05 – 20:33

point being is each individual section of your page is out there competing on its own against everyone else's sections. And a few things follow from that insight and they really change how we write our content. And the first one is that when we're producing informational content, every section has to be able to stand on its own. And what I mean by that is if a paragraph only makes sense because the paragraph above it

Michael 20:33 – 21:00

opens with something like, as we mentioned earlier in this, you know, piece of content. Then if that specific content gets pulled out, alone as a standalone chunk, it doesn't actually mean anything because the chunk doesn't include the context from the prior paragraph. And the system as a as a function will deprioritize it because that's how these systems work. We have to aim within reasonableness to write each section of our informational content

Michael 21:00 – 21:27

as if it might be the only part of the page being read by these AI systems. And then the second thing is that density is a huge variable and not just getting retrieved, but by being an actual source for the citation. So there's actually academic research on this that shows that adding real and verified first-party statistics to a page, and these are essentially just you know verifiable numbers, instead of

Michael 21:27 – 21:56

adjectives or generic or broad claims. Including that improves that piece of content's odds of showing up in AI answers by 40%. And then citing credible sources and then quoting recognized human experts were also very, very strong variables. Basically, the machines here are they're comparing our content to somebody else's content, and they're asking, you know, which one gives it more

Michael 21:56 – 22:22

verifiable confidence to work with and published statistics and verifiable experts, human experts both do this very well. and then there's also a third one, I would say probably, in my opinion, the most important. and it's also something that your AI assistant cannot do on its own. Your content has to offer something that the competing pages do not have.

Michael 22:22 – 22:51

These, what I would say are unique insights that you've spent the time researching and establishing on your own. First party data, case study data, quotes, point of views from your executives, your employees. Because if your page says what every other page in this category says that AI systems are looking at, but in like maybe slightly different words, the system really has no reason, right, to pick yours over theirs. There's a test that we use internally, which is

Michael 22:51 – 23:20

You know, could you take this piece of content, take your logo off, slap your competitor's logo on this content, and could anybody notice that it doesn't fit their business? Right. If the answer is yes, it could be their business or yours, there's no difference. I would say that that may be commodity content. And I would be pessimistic that there's an opportunity for that piece of content to achieve the ability to be cited

Michael 23:20 – 23:49

when users are asking informational questions, which is the whole goal. Now let's also talk about the quality bar itself of the content. This is very, very important. you have probably, if you've spent time in search, heard of EEAT, which is Google's framework of experience, expertise, authoritativeness, and trust. Experience is does the writer of the content have

Michael 23:49 – 24:16

firsthand or personal or practical involvement with the topic, like they've actually tested a product or they've lived through an event they're talking about. Then you have expertise, Does the creator have the formal knowledge or the skills or credentials that are required to discuss this subject very well? Then you have authoritativeness, right? Is the person or the website itself,

Michael 24:16 – 24:44

recognized by other people in the same industry as a primary and credible leader or source of information. And then trustworthiness. Is this page honest? Is it safe? Is it accurate? Is it transparent? Is it secure? That framework has stayed consistent with AI systems, but now it has actually extended further. The bar that we now at Victorious hold content to is an extended version of that called neat N-E.

Michael 24:44 – 25:14

E E A T T, which adds two letters, notability on the front, and then transparency on the back. Because AI systems need both of those before the middle four like really even get evaluated. Let me talk through it with you in just some like plain terms because this needs to be a checklist for every informational piece that you publish. We talked about the EEAT. Let's talk about notability and transparency. So notability

Michael 25:14 – 25:44

is whether you are recognized across other people's websites inside of your specific niche. And it is definitely also relative, which is good news if you're not a huge brand, right? A focused company can literally be more notable inside of one topic than a giant company whose presence is spread across a lot of different topics. And then transparency is gonna be whether a machine can tell who wrote this and also who owns it.

Michael 25:44 – 26:14

This would be things like authorship and ownership and intent and also machine readable. So this is where I would strongly recommend you consider establishing Wikidata profiles for both of your company and then also the authors on your informational content. We'll link in the show notes to a timestamp section of an earlier episode where we talk about this in detail about Wikipedia and essentially how to set up those profiles. But now none of a none of what I just said.

Michael 26:14 – 26:43

Around NEAT and around how we need to be producing content that matches what AI systems are looking for, and we shouldn't be doing scaled content. None of that stuff that I talked about means that you shouldn't use AI to help you make this content. I would say please use it, but it's how you use it. Please use it to plan, use it to structure or to get a first draft of the content

Michael 26:43 – 27:11

on a page. But the thing I'm saying is that the machine can't supply the unique and human-led portions of content that actually, in my opinion, earn the citation. Right. It cannot magically supply you with some unique data and it can't give you your own expertise or your experience because it doesn't have any of that and it hasn't done any of the actual work. It also cannot take responsibility for what gets published under you.

Michael 27:11 – 27:37

Or your company's name. I got into this a couple episodes back here on The Search Signal when we talked about what AI is doing to search budgets and how we resource search campaigns. I'm not gonna relitigate that all here, but the short version is these tools were never going to be a differentiator for us. They're just a new tool to use. And everyone now has it. And

Michael 27:37 – 27:54

a tool doesn't create value on its own. It needs a human to use it and to leverage it. But what you are adding as a human being, it's the expertise. And the expertise is what these systems are ultimately trying to detect.

Michael 27:54 – 28:20

Okay, now that we've established that informational content has a a new standard of N-E-E-A-T-T NEEATT and that our content approach should be measured and should be thoughtful in both the volume of the publishing and also the utilization of AI assistance in, I would say, appropriate manners. Let's talk about how to diagnose your existing content versus creating brand new content

Michael 28:20 – 28:50

for influencing these AI systems so that you are showing up as a reliable citation source for informational queries from users and also influence those systems as a reliable downstream commercial mention and also recommendation. So, first things first, you have to have your company's entity strategy done and in place. This includes establishing which commercial entities you want to be associated with in the AI systems,

Michael 28:50 – 29:18

having an active and also unique commercial page on your website associated with each one of those entities, a completed keyword research strategy for traditional search that identifies which searches have strong commercial volume and also reasonable competitiveness. And then also a prompt strategy built on top of that keyword strategy that reflects the type of high-level questions that users might ask AI assistants that

Michael 29:18 – 29:45

are going to ultimately correlate with those keywords. Those prompts should be both informational and commercial. So split into those two categories so that each entity has a category of informational LLM queries and commercial LLM queries. I'd probably recommend you start with 10 to 30 of each per entity. Now, once that is done, we have to run a content audit

Michael 29:45 – 30:14

against each of these entities and also its associated keywords and its associated prompts. We have to do this because we need to understand what existing content is already published and is designed to answer the prompts that we have built in our list. Some of the content on your site will not map to anything. We can set that stuff aside for now, we can come back to it. But for everything that does map, once we've established it, we need to pull how those pages are performing today on both sides. Are they ranking

Michael 30:14 – 30:38

for the keywords we want them to rank for? And are they getting cited in AI answers for the prompts that we want to be cited for? Once that's done, we need to then take the pieces that are getting cited and we need to study them. What are those pages showing up for in citations? And what do those pages have that your uncited pages do not have? And I'd say this is also a place where AI

Michael 30:38 – 31:03

Could genuinely be useful as an analyst. So you can absolutely hand it content that isn't getting cited, alongside the content that is getting cited, and then ask it to tell you, you know, what is different as it relates to the N E E A T T framework, the NEAT framework. That gap analysis is important because that becomes our reoptimization list. And we always want to start by re optimizing

Michael 31:03 – 31:31

what is already on our website because it is 100% the low-hanging fruit of the bunch before we get into any production of new content. Now, also, I would say the audit is likely going to surface some additional uncomfortable facts, which, for example, there might be content that should probably be pruned, or there could be two pages that are cannibalizing each other for the same type of question, prompt, or keyword. You might also have a bunch of thin posts

Michael 31:31 – 32:01

that would be one strong page if you consolidated them. And sometimes it's in the reverse. Where there's one big sprawling piece of content on your site that covers, you know, either a bunch of subjects or a whole subject at a level that's very general, so general that it might not really be answering anything. And it should be split into dedicated pages that each are answering something specific and going after specific prompts and keywords inside of our strategy. But

Michael 32:01 – 32:24

only after we go through the process of content auditing and re-optimization of the existing content, only after that do you ask whether or not you need anything new. And some of us, we may not need anything new. Optimizing what you already have is a much smaller lift, in my opinion, than net new production. And it's often where the performance

Michael 32:24 – 32:50

opportunities are at. And I get that new content is more exciting than a re-optimization sprint or whatever, but the systems are not scoring you on whether your content feels exciting to you, right? You create new content when your existing content library is genuinely thin, when the reoptimization list runs out, or when the audit shows your content was built around topics

Michael 32:50 – 33:19

that were never entities you wanted to work with to begin with. In those cases, that's where you prune those and build what should have been there instead. So let's say you are there. Audit's done, the reoptimization list has been run, and there are like some genuine gaps. Questions that are in your entity's territory and prompts and keywords that nothing on your site answers. This is how I would approach that net new work. Okay. every new piece begins as one core question

Michael 33:19 – 33:48

pulled from that territory map. It's a question that your buyers are asking somewhere in their informational journey that you currently have no answer for. But before anybody writes a word of content, we need to first go and look at who is getting cited and who is ranking for that question, that prompt, or that keyword right now. You can either read those pages yourself, or I would recommend sharing them with whatever AI assistant

Michael 33:48 – 34:16

you use and follow the same system and patterns approach we talked about earlier. What we need to do is figure out what the systems currently consider as a good answer and then where the opening is. Maybe that could be something like a sub-question that other pieces of content are glossing over, or more likely, you might have some unique data or some unique experience or insight.

Michael 34:16 – 34:44

That matches the NEEATT principles that others do not have that can create a better version of that content. If you've got that and you're ready to move forward, I would then build a content brief. And I know it sounds like very processy and it's not process for its own sake. This is really, really important. The brief for a content piece like this needs to name the entity that it's serving and which of that entity's attributes it's going to cover. Because

Michael 34:44 – 35:12

you know, attribute coverage is what is going to give these systems a lot of factual ground to stand on, whether they're considering to cite you or not. The brief needs to name the core question. It could be the prompt that you're focusing on, plus the related questions that are also going to belong on the same page. And I would say, you know, you could address multiple prompts in your strategy with the same piece of content. You can approach and optimize for

Michael 35:12 – 35:37

multiple keywords on the same piece of content, but they have to be semantically grouped, and we need to make sure that we identify ones that belong together as much as ones that don't belong together. So we identify which keywords that piece of content is associated with and where and how often they should be included in the content. It also sets the brief sets the structure and the schema for the page.

Michael 35:37 – 36:05

And also where the content should internally link to other associated pieces of content that are sitting within that broader entity. And then it also needs to name the unique things that only your company can add. Like we said, whether that's like a data set or whether that's first hand experience or client case study information. I would say if your content brief cannot answer that unique need

Michael 36:05 – 36:35

in the content, I would argue it may be a commodity piece. And we do not want to publish commoditized content. So if that's the case, we need to develop and get first-party data and expertise as a core component of our content before we start writing it. And then we can draft. And I would say this is the part where the AI can genuinely help us. And I would say like the structure of the content and the first pass of the content,

Michael 36:35 – 36:57

but we need to shape it the way we talked about earlier. The direct answer to the core question needs to go at the very top of the page. I'd recommend 40 to 60 words. and like we said earlier, written so that it could be pulled out, chunked out, whole and still make sense on its own. I would also recommend headings are written as the questions that people are asking

Michael 36:57 – 37:23

in the words that they are asking them. Each section needs to stand alone and it needs to cover one concept. And I would strongly recommend specifics over adjectives wherever possible. Because again, these machines are comparing our paragraph against somebody else's paragraph, and they're going to pick the one that has more substance to it. Once we've done that, before we publish,

Michael 37:23 – 37:50

we need to run it against the NEEATT principles we walked earlier. And we need to treat NEEATT as a gate. So does the piece show experience? Did somebody who's done the thing that it's talking about put something firsthand into the content? Is there depth in that content that a shallow competitor cannot match? Is there a named author that the machine can identify? Does that person have credentials that it can verify? And are the page

Michael 37:50 – 38:15

claims consistent with what the rest of your website claims? If that all passes, then we can publish it. We'll get schema on it. The brief will help us figure out where to link to other anchor pages or entity pages and then link those pages back to this page so that this new piece lands inside of the whole entity structure instead of floating in the website on an island on its own. And then I also do want to say

Michael 38:15 – 38:43

it's really important that I just talk a little bit about the pace of publishing, especially with our conversation today around AI-scaled content abuse and penalties that we've been seeing and the new algo update that happened this week. I would much rather see your company publish two pieces a month that are passing with flying colors that NEEATT gate than 20 pages that do not or 40 pages that don't. Because

Michael 38:43 – 39:06

we've already covered what happens to volume that does not have a strategic backbone underneath it. Essentially we don't really want to overwhelm our domain with a ton of commoditized generic content because we know for sure at minimum it won't get cited. But worst case scenario, we can absolutely run the risk of a scaled content abuse penalty. And then the last thing I would say,

Michael 39:06 – 39:34

taking a step back is on expectation setting for performance. So when a piece of content is working, I would generally say you'll see it in Search Console first. And I would say you would likely see it in long tail keywords and variants of your primary keywords. And they'll start to show impressions first, usually inside 60, maybe as far as 90 days, sometimes earlier, sometimes later.

Michael 39:34 – 40:03

And then after that, you should start to see it show up in things like AI systems as a cited page or cited source and answers. you can spot check this in some AI visibility tools that do broad citation looks across the whole internet, not correlated to specific prompts that you're tracking. Two tools that do this are SEMrush and Ahrefs. Again, those tools sample.

Michael 40:03 – 40:30

Because there is no tool in existence that can cover the literal infinite space of questions that people are asking, but they are definitely enough to see the direction. We are going to go deep on measurement in the final episode of this series. we're going to talk about leading indicators, what is trackable, what is not trackable, and then also how to build the case for your company's leadership so you can actually get the budget and the resources you need. So I'll save more details around

Michael 40:30 – 40:57

measurement outside of what I just said for that specific episode. But let's pull this all together. Let's take a step back. Like we've been talking about this whole time, the top of the funnel, informational content exists to influence the bottom of the funnel. And there are two different levers to move them. Up here at the top of the funnel, it's your ability to demonstrate subject matter expertise on your entities.

Michael 40:57 – 41:22

And your website is the one channel that you completely control for telling these AI systems who you are and what you do, what you know. And then down at the bottom, it's going to be how the rest of the internet talks about you. These are the mentions and the co occurrence in other people's content. And that is next week's episode. So today's awareness is tomorrow's pipeline.

Michael 41:22 – 41:51

And the companies that are treating informational content as authority building and they're doing it now, those, in my opinion, are the ones that are going to be in the commercial answer later when the question has literal money and pipeline attached to it. So thanks for listening this week to The Search Signal. If this was useful to you, please go ahead and subscribe. So the rest of this series and the rest of our episodes find you. Part three is going to be next week, and it is all about

Michael 41:51 – 41:53

the bottom of the funnel. I'll see you then.