Latest / The Search Signal / How To Build an AI Search Strategy, Part 4: Measurement and Reporting

Transcript

How To Build an AI Search Strategy, Part 4: Measurement and Reporting · The Search Signal

• michael: So the moment that an AI assistant recommends your company to somebody, it happens in a conversation you cannot see. And the journey from that, private AI conversation to getting to your website and converting is just not a simple or straightforward motion right now. So the amount you can know about how this channel is working for you is actually going down while the channel itself is growing like crazy. Meanwhile, our leadership wants more clarity. Hey, welcome back to The Search Signal. I'm Michael Transon. I'm the founder and CEO of a search 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 at Victorious have been running campaigns for over 13 years, also for some of the biggest websites, some of the fastest growing brands out there today. And we bring what we learn from those experiences as well as a lot of our own first-party data and research into The Search Signal to help you with your own work. And today is the fourth and final part of a series on building a full funnel AI search strategy from scratch. And last episode, we covered the bottom of the funnel and that whether an AI system chooses to mention your company or not in a response to a user is mostly decided by how often other websites write about your company and your brand in relevant content. Now, today is going to be about measurement and reporting. We're gonna talk about how to know whether our brand is improving or declining in AI search as a whole. And then also we will talk about how you can explain that to people like leadership or your board or anybody else at your company who has a vested interest in that performance. And before we get into today's episode, let's spend a little bit of time talking about what happened this week in the world of search marketing news. Okay, let's start with Google's Search Console and what Google said this week about its own AI reporting in GSC. So Search Console got a report over the summer. That report simply just shows how often your pages are appearing inside of AI overviews and AI mode. And this week, John Mueller from Google's search team had answered a question about how positions get counted for those appearances in the regular performance report. And his answer was that an AI overview gets tracked as one single block, which is the same way that Google seems to handle other search features. So every link that is inside of the AI overview is going to be assigned the same position. And is in his own words, the the old position you know of one through 10 is he said hard to map through AI overviews, or to make it useful for site owners. And then a couple days later, he also added that he expects this to evolve over time as AI overviews and AI mode themselves keep changing and evolving, and that there are also, in his words, some edge cases. Google has not yet worked out. So, point being, the one Google provided number that you have for AI visibility right now is a count that Google itself also expects to change. And when you put the the Search Console AI number, we'll talk about this later in front of leadership. I recommend you present it as a directional reading and tell them that we are going to expect this counting method inside of GSC to change. And then the next thing to bring up this week, Similarweb published a study this week on how much AI influenced traffic your analytics can fully see. So Similarweb, it's a traffic measurement company. this is vendor data. They use their own US desktop browsing panel to follow thousands of people who'd gotten a brand recommendation from ChatGPT in finance and travel and beauty over the second half of last year. So, what they found was that those people were about two and a half times more likely to visit the recommended brand than a direct competitor within the next week, and that of the visits from people who hadn't been to the brand site in the prior month, fewer than one in ten came through a channel that analytics tools would label as AI. And the rest just showed up as like search and direct or other channels. And the conclusion that they brought is that attribution models are handing credit to direct traffic and other channels for demand that AI was creating. And I think that's right. For your company, this is the number that you are gonna bring to somebody when someone asks why the AI channel in GA4 might look so small. Because the answer based off of this research, and we have it in the shownotes from today's episode, is the channel is actually seeing a fraction of what it's actually influencing and driving. And then the third news from this week is back to Google. They have started to pay some publishers for content that is shaping its AI answers. And that is showing up inside of Search Console. So Digiday reported this week that Google is running what it's calling an AI, I think they call it an AI contribution pilot, where they have a couple dozen publishers. I think they're mostly small, mid-sized, and they're getting an earnings panel inside of Search Console with a monthly dollar figure for content that, in the wording of the pilot's own documentation, contributes, quote, significantly to a Gemini AI overview or AI mode response. Now there's no detail on how that number is actually getting calculated. Google did also confirm to Digiday that it's an early stage, I think they called it a learning pilot for rewarding high quality content. And there was a source that's close to the pilot who told them that the early payouts were like pretty small next to traditional advertising revenue. So I don't think most of us listening are going to be in this pilot. Wouldn't certainly plan around it, but what I would take from it is that Google is now putting a dollar figure on AI visibility inside of the same tools that also hold your impressions, which means that Google has some internal way right now of scoring how much a page is contributing to an answer. And it's not really being shared with anybody outside of the pilot just yet. So I picked these three news reports from this week on purpose because every one of them is about the gap between what these systems know about your visibility and then what they let you see about your visibility. And that gap is what today's episode is going to be all about. So with that being said, let's dive in. I want to begin with talking about the direction that measurement is moving in right now. So for about the past 15 years or so, marketing measurement, in my opinion, has gotten more and more precise pretty much every single year. So we went from, you know, knowing roughly how many people visited a website to knowing which specific ad they clicked, to, you know, which page they landed on, how long they stayed on that specific page, and which one of those visits ultimately became a customer and how much they spent. right? Marketing teams have been building their habits around that. And a lot of the teams have gotten to the point where they have only done things because they can measure it. Because honestly, those are the things that they could defend most successfully in a budget meeting. And what's happening in AI search is it is moving in the complete opposite direction. So the moment that an AI assistant recommends your company to somebody, it happens in a conversation you cannot see. And the journey from that, private AI conversation to getting to your website and converting is just not a simple or straightforward motion right now. So the amount you can know about how this channel is working for you is actually going down while the channel itself is growing like crazy. And it's not because we're lazy or we don't, we're not smart enough to figure it out. It's just how the channel is actually working. Meanwhile, our leadership wants more clarity. There was a survey I read about earlier this year of about 500 CMOs at big, large US companies. And close to nine in 10 of those CMOs had had a board or a CEO mandate to build an AI search strategy. And only about one in five of them even had the operating practices in place to run one. More than half of them had said that their board is asking about AI visibility almost every single week. So leadership knows that this channel is critical to the growth of all of their companies, but we cannot treat it with the same level of granularity in reporting because none of the platforms are providing it to us. And one of the things that I hear a lot, and I've said it myself here. Right, is why don't these companies just give us reporting. Google gives us Search Console. Why can't OpenAI or Anthropic just show me what people asked when my company came up? Nobody at these companies, to my knowledge, has gone on the record and talked about why. Google's got a public position, which is that more metrics are going to be coming out over time. But you know, as of right now, I think that there are maybe three reasons why we aren't getting that level of attribution and reporting we'd need to match the same level of specificity that we are getting from other marketing channels. And the first one I would say is that a conversation with an assistant is private and private in a way that a search query on Google has never been before. So people tell these systems things that they would never ever type into a search box. Your company might come up. 30, 50, 100 messages into a conversation that might have started with somebody's, I don't know, private health condition, or they've got a legal problem they're working through. And the only way to tell us as a business what was asked that resulted in our name or our website being used in the response is to summarize that person's private conversation for us and the platforms for very obvious privacy reasons are not going to do that. That's the first reason. The second reason is that. Search Console and Google Analytics, those work only because they are limited on our own website. They show you what's on your site. They show you what people searched to get to that website. And then what they did once they actually arrived. There is no equivalent boundary for a conversation about your category where the assistant's talking about your company without actually using your own website. And then the third issue is that a search engine has a very real business reason to send you clicks. Because clicks are where the advertising lives. And most assistants do not have that reason yet. Now, this is also starting to change. As advertising is starting to show up inside of these products. We talked about it in today's news recap. And if it does, you know, the reporting may follow. But my concern is that it will be centric to paid advertising. And I'm not totally sure if we are going to get the same reporting visibility that we can get with non-paid activity. So, in my opinion right now, taking a step back, is this channel generally mostly resembles a version of word of mouth. these are conversations between a person and something that ends up talking back and they play out differently every single time and we cannot listen in. You can, as a brand, influence word of mouth. You can make sure the people that are doing the recommending know what your company is good at with good marketing, but you can't fully measure it. And also nobody really is expecting you to, right? Should your should your CEO be able to reasonably ask you to track word of mouth? If you got asked that in a meeting, the answer would most likely be no. How would you even do that? You have to like actually listen in on people's conversations. The assistants are in the conversation. And they don't share it. So we need to start thinking about using this channel and holding this channel to the word of mouth bar. You can influence it, you can see its effects, and you can report on it as a picture built from several different directional numbers. So measurement in this channel is just less precise. I don't know if I see that reversing. But now that we have established that, I think we can talk about what you can measure because we're not completely stuck in the dark. We do have, in fact, ways to measure more broadly brand performance and connect what are disparate dots to tell a really strong, clear narrative that should reasonably satisfy two different things. Number one is your own confidence in the work that you're doing to grow the channel and to understand what might be working and not. And then second, your leadership's confidence that this is a good use of money and budget and it is helping the business grow. So let's talk about what we can measure. Okay. As we dive in to what we can effectively count and measure, I want to start by saying the things that you measure in AI search channel reporting are really the same things that marketers have always measured. How visible you are, how much traffic you get, and how many of those visitors just ultimately become customers. What has changed though is what each of them look like, right? Visibility, for example. That used to be something that you could just count down to the individual keyword in Search Console. In AI search, you can really only sample it. And you do that by asking an LLM a generally worded question and then just see what it responds with. Traffic, on the other hand, is lower now than it has ever been for a given amount of visibility. Because these systems are answering the question in place. And most people do not click. They don't click through either because the system doesn't give a link for them to click or the conversation just, you know, naturally continues on. But in very good news for conversions. Conversions from the traffic that does arrive from AI search is much, much higher because somebody who visits your site after getting an assistant's recommendation has usually already made most of their decision. So now these numbers that we're going to track are going to come from two different places. And I think we should keep them in two separate groups because they behave differently. But also more importantly, they are going to answer different questions. The first group is everything that happens before anyone ever gets to your site. And those are what we are going to call our leading indicators, meaning that they're just the numbers that move first. When the work is just starting to take hold and impact results. The only place where you can get these leading indicators is something like a prompt tracking tool, which is usually software that asks all the AI assistants a set of questions and they do it on a schedule and then they record what comes back. And the set of questions it should be asking is, harkening back to episode one, the prompt strategy that we built. So if you follow that episode, you probably or should have somewhere between 10 and 30 informational prompts and then 10 to 30 commercial prompts for each of your commercial entities, which to remind you means each product or service that you are trying to sell. Those get loaded into the tool, and each of the prompts are tagged as either either informational or as commercial. Then you also add three to five competitors you named, which we did in part one. And you turn on the platforms and track the ones that your buyers most likely use, which for most companies are going to be ChatGPT, AI mode and AI overviews for Google. And then, Gemini, Perplexity, Claude. You can add those in if you think that your buyers use those platforms. So once that though is set up, the tool should be giving you four different specific numbers each month for each stage of the funnel. Which are your share of voice against your competitors, how often you are being named, how often your pages are being cited in answers, and then also how the assistants are describing you when they do name you. That's the first group. Those are the leading indicators. The second group is everything that happens on your website and in your CRM after somebody has already heard about you. And these are the lagging indicators, meaning the business results that are showing up later. So every one of these can be pulled from tools you likely already have. So I'll go through them real quick. In GA4, for example, the traffic acquisition report has a channel called AI Assistant. And if you filter it, you can pull down sessions, you can pull down conversions, conversion rates for any given month. I do you want to say Perplexity and Claude aren't included in that channel. So you also want to have a custom channel group that adds perplexity.ai and claude.ai as sources. Otherwise, you are going to be under counting. In that same report, though, you can pull organic search conversion rate as its own line. You can also pull direct traffic with the landing page dimension added. So you can actually see which pages these direct visitors have arrived on. Then also, out of GA4, you go over to Search Console. If you take the performance report and you filter the queries to ones that are containing your company name, that gives you something called branded impressions and clicks. And the new generative AI report actually gives you the number of times your pages appeared inside of AI overviews and AI mode. And then the last one comes from your CRM where you would count how many new leads or customers if you give them the option, wrote in chat GPT or AI or whatever assistant's name in a like how did you hear about us field. So that's about seven different points of data. You don't need to buy anything new to get any of them if you're already paying for those. But also as we're talking about this at a high level before we jump into the details of each of them, how often I would also look at this, I would recommend not looking at these every day. Looking at the tracking numbers, probably once a month. I would definitely resist the urge to check them on a much more frequent level. The reason because this is a change on a page that, you know, has already been cited can show up in the answers within a couple of weeks. But a change to your own content or how often you're being mentioned on the web that can take one or two months to really just settle into the numbers. And then on top of that, we've talked about this two episodes ago, citations can just come and citations can just go. You can be a source for an answer on a Monday, and then you're gone on a Wednesday, and then you're back on Friday, and there's no real clear reason for any of it. And anybody who is reading this on a daily basis is gonna spend a lot of time chasing ghosts and reacting to movement that's just not very real, in my opinion. So those are the two main groups of numbers. Okay. And we can pull them on a monthly basis, but let's go deeper on each of these. Let's start with the first group. So the main number an AI prompt tracking tool is going to give you is something called share of voice, which just simply means how often your company is showing up in answers compared to your competitors. Last episode, we got into how some of these tools compute share of voice differently because they do it in different ways. So for example, Ahrefs builds its share of voice number from a very, very large set of prompts that it generates from its own internal search data. Then you also have prompt tracking tools like Scrunch or Profound or Peec. They are building the share of voice metric from the prompts that you load in. And these are not comparable. Today is how we read this number. The most important takeaway from understanding share of voice is reading it as a poll. Let's use an example. Like a political poll. Political polls. They ask a couple thousand people a question. And let's just say they report back that a candidate for whatever office has, I don't know, 49% of the vote or something. Nobody thinks that all 350 million Americans were asked that question and will vote exactly at 49% right? It's a sample. It's got a margin of error. Usually it's right, sometimes it's wrong. And like everybody understands that. Share of voice is the same kind of number. And I would also argue it could only ever really be that kind of number. And the reason is because the prompts people type are effectively infinite. Every prompt has got a ton of variations. Every variation runs on a model that it might get updated. The in the same prompt on the same model could also give a completely different answer on a different day you might ask it. So for example, a 40% share of voice means you were in there 40% of the prompts the tool ran. It is a sample, the same way that a poll is a sample. So, like I said, there are four numbers we need to track. Share of voice, which is against the competitor set that you defined in part one. Then there's mention rate, which is how often your name gets named in the answer, your company name gets named in the answer. And then you have citation rate, which is how often one of your own pages gets used as a source. we covered last episode why mentioned, and cited are different things that do different jobs. And so we need to keep them as separate metrics. And then there's also the fourth one: sentiment, which is how often an assistant describes you and how it describes you when it names you. It could recommend you, it could just simply list you, or it could name you and attach a warning. Unlike a search result in Google, an AI answer can formulate an opinion in it. And, for example, being described as like a super expensive option that's maybe difficult to work with, in my opinion, is worse than not being named. And then we need to look at it by the funnel stage. So we talked about this before, but in Victorious's Q2 AI search report we separated the prompts that we were tracking by where the buyers were in their process. The assistants tend to cite sources, which are like other websites for formulating an answer at pretty much every stage, but they only start really naming companies once a user is asking late funnel stage questions or the conversation is moved into a commercial context. A single share of voice number across all of your prompts is going to average all of those stages together. Your top of the funnel prompts are going to pull the number down for reasons often that have nothing to do with how well you're actually performing because top of the funnel prompts barely use company names. And then your bottom of the funnel prompts where being named is really the whole point, is gonna get diluted. So we need to track them in separate buckets so we don't get improperly influenced by the wrong data. We pay a lot more attention to bottom of the funnel share of voice because that's the stage where buyers are choosing. And in my point of view, the top of the funnel does have value because it's influencing the bottom. A company that the systems keep seeing in informational answers and using their websites to form responses is a company we believe they're more likely to name in the commercial questions and conversations. So top of funnel share of voice, but also more importantly, top of funnel citation rate is a leading indicator for bottom-of-the-funnel share of voice. But all of this depends on the prompt strategy and the competitor set from part one being done carefully because your share of voice is only going to be as good as the prompts that it's computed on. A tool that is sampling a prompt that doesn't make sense is not a prompt that you really care about or want to be tracking. But zooming out here, from you know, tracking all the individual prompts and all the stages and stuff. We also need to consider how we are tracking across individual platforms and we need to avoid the temptation to lump all of these together. And the reason why I say this is because the platforms themselves are all very, very different sizes. ChatGPT, it's about half of all of the web traffic going to and from AI assistants. Gemini is about a quarter versus Perplexity, it's like 1%. right? It's barely anything. However, Google's AI overviews, those reach more people than any of them because they are sitting at the top of Google Search, which gets the most traffic. So if our reporting runs our prompts on every platform and then averages the results equally. You could have a strong showing on a small platform like Perplexity, inflating your number in a place where your buyers aren't even looking and using. Some tools do weight for platform usage and some do not weight for platform usage. I would say we need to look at each platform separately either way. We spent quite a bit of time in a dedicated episode earlier in The Search Signal, where I had made my argument for why I think Google surfaces are the priority for most businesses, with ChatGPT coming in second. I'm not gonna talk in more detail about that here or rehash that argument. So if you do want to know more about why I think that, we will link that episode in the show notes so you can read that or or listen to that later. But before we do move past that. I do want to say that on Google specifically, the way I've come to see their platform evolve. And also what you should plan for is that an AI overview, which is at the very top of search results and a very important component of an AI search strategy, in my opinion, the most important. That is also, on top of that, the first message of an AI mode conversation. They still call it an AI overview at the top of Google search result pages, but in form and function, it's the opening turn of AI mode. But a mention in AI overviews on a commercial intent search is, in my opinion, the single most valuable place that your company can show up right now. And more valuable still when you also rank in the regular results below, because you know, the person might read the AI overview question and then scroll down and they might find you a second time. Most of the AI tools and most of the AI prompt tracking tools in the market also find your brand mentions by matching a string of text. And if your company's name is also an ordinary word, they can sometimes count all of the uses of that word. And I just want to really quickly call this out because this has actually happened to us. One of our tracking systems had reported that a bunch of our leading indicators had just like shot up out of nowhere and we had no idea why. And when we dug into it, we realized the tool was counting every page on the internet where somebody else wrote about an ambiguous entity, "victorious" which, if you don't know, is not just our search agency, but is also a normal word, of course, but is also an old TV show on Nickelodeon. So if your company's name is a common word or if it shares the name with another company, we need to check how our tool matches and we need to make sure that it requires your name to appear alongside your category or your industry. So a mention only is going to count when the page is about you and also what you do. So that pairing with your name and your category is one of the is the idea we talked about from part one called co-occurrence. So take a quick step back. The leading metrics we've been talking about. And leading metrics are about how well you are performing an AI search. And they are sampling prompts for brand mentions, how often you're being named as a recommendation, citations, which is how often your website is being used as a source for answering questions for users, and then also share of voice, which is how often your brand shows up versus specific competitors. Those are the leading indicators. Now that's part one. Let's talk about the lagging layer. This is part two, which is what your own website analytics can actually show you about how this channel is performing. Because just simply looking at something like your AI referral traffic is going to be an incomplete and also inaccurate way to measure how these these systems are sending visitors to your website. Okay, for lagging indicators, most companies that we've worked with, the analytics that they use are going to be Google Analytics and Search Console. So I'm gonna go through those specifically. I'm gonna talk about what each of them shows about AI as of the recording of this episode, this month in September of 2026, because things have been changing a lot over the summer and will probably change a lot after this. So in Google Analytics since May of 2026, there's a default channel called AI Assistant and it groups visits from ChatGPT, Gemini, Copilot, and a couple others into one channel. I mentioned this earlier. It doesn't include Perplexity. and Claude would still show up under referral traffic. And it doesn't include Google's own AI overviews in AI mode. As of right now, Google still defines those clicks as organic search. That traffic will arrive looking exactly like a click on a regular search result, and there's really no settings right now that separates them. So your organic search channel now contains AI-influenced visits. And if those visits convert better, which I would personally expect, and we'll talk about this later, your organic conversion rate is actually gonna go up. What that means is a rising organic conversion rate with either a flat or even a declining organic traffic, is now a signal about how well you are performing in Google's own AI systems. And in Search Console, there's a new report called the I think they call it the Generative AI Performance Report, which went out to everybody, if I remember, at the end of August. So last month, it shows impressions, meaning the number of times one of your pages was included in an AI overview or an AI mode answer. It does not show clicks or queries or positions. And also AI overviews and AI mode are combined into one line. And those impressions are also counted inside of your normal search totals. So this is a view into visibility that was already in your numbers. It just tells you that you appeared. And Google has also, like I said earlier, said that more is coming. Separately, if you are in a company that sells products, Google Merchant Center also added an AI performance report this spring. It shows you your share of voice across similar brands, across AI mode, AI overviews, and the Gemini app. They also do split it by funnel stage, which is nice. Also along with some of the popular terms that people are using. So it also covers only English language queries in only a few countries, and also it shows aggregated terms rather than individual prompts, but it is the closest thing to a Google-provided share of voice number that exists. And if you sell products, I would definitely recommend that you look at it. So those are some lagging indicators. The next important one to talk about is going to be direct traffic. So direct traffic is and has always been what analytics calls a visit from when it can't tell where the person came from. I want to be very clear: direct going up doesn't necessarily mean that there's just a bunch of people out there that are typing manually your web address, you know, your domain into a browser. A lot of this is coming from AI search. So for example, when somebody clicks a link inside of their ChatGPT mobile app, or they just copy and they paste it, the visit often arrives with no referring information and gets filed as direct. The other very important thing is cookie and consent blocking. Those also strip the source from a lot of visits. And then on top of that, bots that are fetching your pages can also show up there too. So it's another lagging indicator. And then after direct traffic, you also have branded search, which means that somebody is actually searching your company's name. So think about it this way: somebody who hears about your company in a chat assistant and wants to learn more about you, what are they most likely to do? Well, they're probably going to open up a new tab and search your company name in Google. So what I do, pull branded impressions and pull branded clicks out of Search Console as their own line in whatever reporting that you're using, because it is definitely a primary lagging indicator. If the assistants are naming you, here's the big takeaway on this one. If assistants are naming you, more people are going to search your name. And then the last, I would say, probably most important lagging indicator is conversion rate. We have talked about this ad nauseum in this podcast, but traffic from the AI assistant channel and traffic from AI converts significantly higher than search traffic or any other traffic source for that matter. So if Google AI overviews and AI mode influence people the way that other chat and AI assistants do, and their clicks are going to be inside of your organic channel, then your organic conversion rate is going to rise too. And you're also your overall site conversion rate is going to rise with it. So the lagging layer shows your arrivals to your site, your branded search demand, your direct traffic, and your conversion rate. Those are the lagging layers and metrics that we need to look at for AI search performance on top of the leading indicators . So now that we've got both of these layers, let's talk about reporting itself. And we can talk about, you know, who is this going to, your CEO, your CFO, your board, your CMO. Regardless, what I would recommend here is going to be two things. I would recommend getting your reporting on a monthly basis and then also getting your reporting and your narrative on a quarterly basis that explains what's happening in the numbers. So let's start about the talk about the monthly to start. I would recommend a monthly consolidated one pager report that is basically the sheet from segment two cleaned up from somebody who isn't going to read a spreadsheet. So the top half comes from our prompt tracker and it starts with your bottom of the funnel share of voice against your name competitors. And I would recommend you do one row per platform. So you got ChatGPT and then AI mode and AI overviews and then Gemini and then Claude, each getting their own line. Underneath that goes your top of funnel share of voice, labeled, in my opinion, in a good way, would be the leading indicator for the row that's above it. And then your mention rate and your citation rate. And I would say also flag if any sentiments on any of the prompts are changing or turning more positive or more negative than the last month. And then the bottom half of the report needs to come from your analytics and your CRM. And it's all the stuff that we talked about. It's got the AI assistant channel sessions, its conversions, its conversion rate, also with the organic conversion rate next to it for comparison. Then I would report on branded impressions and clicks. I would also report on direct sessions to the pages that the assistants are citing. Also your overall citation rate and the count of people who wrote an assistant's name in a field if you are tracking that when you're doing lead gen. That last one being pretty optional. Every row should show, in my opinion, also the last six months or so. So whoever's reading this is looking at a trend. I don't recommend looking at single numbers. And I would say you label the top half leading and the bottom half lagging. And the reader knows which numbers should be moving first because we are showing what's the leading and what's the lagging. And then also, I would also recommend you consider doing some sort of quarterly narrative reporting where the numbers are getting connected to each other. And I would recommend you could write this in maybe three parts. This is very basic, but first part is what did the team do this quarter? What's the work that was being done? We talked about how to build an AI search strategy in parts two and parts three of this series. So we're gonna talk about, you know, for example, how many pages we ultimately got published or rewritten. Also, we'll talk about how many placements were secured and what sites we are securing placements and brand mentions on. The second part is what happened to the visibility, which I would say, you know, the top half of the one pager that we saw across the last three months. And the third part is what happened in the business, which is the bottom half of that report. It's the lagging indicators. When it's written in that order, a reader can very easily follow a chain from activity, to the visibility, to the outcome, even though that last step, like we talked about, is directional rather than exact, because these platforms don't give us the exact numbers like we have gotten historically from GA4 and GSC for other channels like organic search. so also on top of that, I would also recommend creating some narrative, some paragraphs that you write out that says what the report can and what the report can't show. And say something like, we can show the sessions that came from assistants that passed a refer. We can show branded demand. We can show what customers told us about, where they found us. And we can't show the full path, you know, from a recommendation inside of somebody's private conversation, like we talked about, to an actual signed contract. That paragraph is helping people to understand where the limitations exist. So a report that is built that way answers two questions leadership from my experience tends to ask. When your CEO comes to you and says, you know, I asked ChatGPT what's the best company, and we weren't in the answer. You can point to the top half, the leading indicators of your reports and say, you know, here's our share of voice on ChatGPT for all of the commercial prompts. Here's the trend. Here's what we're doing about it. And, explain that one prompt on one day is one response in a greater poll. And then on the other side, when your CFO might ask you, hey, how much business did I bring in this quarter? You can point to the lagging indicators, the bottom half of the report and what the report can and what the report cannot show. Which also brings up an objection that you're probably eventually going to hear. There are some executives, a lot of executives that hold a view of marketing spend that goes a little something like this. I put a dollar in, I want to be able to watch the whole journey. I get a dollar fifty out. And if I can't follow or watch that whole journey, I'm not going to spend that dollar. And I do want to be fair that because it is a legitimate preference, and there are companies that exist where that's the right way to think about it. But it also means that your company is not going to use this channel. And this is the fastest growing channel in marketing. And the way people are making decisions and buying things is moving quickly into it. So you might call that view that I describe some professionals or executives having as dogmatic. Maybe I'd call it that, but it is something that needs to evolve because you know, companies already spend their money without full attribution all of the time. Your company might have, for example, rebranded sometime in the recent past or it's going too soon and nobody can really trace the new logo to revenue, right? But you approved it because you expected it to grow the business. You also might do things like direct mail, which goes out and nobody can trace or nobody can see every recipient and how they might use that to make a transaction, even if you give them a special phone number. You never know the full return. Or you might invest into TV. Somebody sees a TV ad, goes to your website, drives to the store, follows you on Instagram or whatever. You can't see any of it. But all that spend gets approved on the basis that it builds the brand and the brand produces revenue. And AI search is now in that category. Search used to be filed under performance marketing, where every click that we had was very measurable and it's turned into brand building. Honestly, and brand building has never had click level attribution. And I'd also add that the certainty everyone got to with other channels, everyone that got used to feeling that certainty was never also as solid as it felt because, you know, analytics has misattributed traffic for as long as analytics has really existed. And you know, every so often somebody discovers that the channel that they thought was performing and was getting credit for was from another channel's work. The gap between what we could measure before and what we can measure now is actually, in my opinion, smaller than it feels. So the question becomes: what do you do all with all that? Right? You can't change who your executives are, and your plan can't be that leadership just needs to get more comfortable with being uncertain. So you have to work inside of the constraint. So what I've been recommending is four thing. You label the two halves of the report as leading and lagging indicators every single month. You write the can and you write the can't show in the reporting paragraph. And you want to be able to communicate that consistently to your team so they understand what is within our control to measure and what is not within our control to measure. And you also show a trend line. We need to be showing, four, six, twelve months of history on each of these metrics so that we can mark. changes instead of just restarting our baseline constantly. And then I would also connect the top half to the specific placements and the pages that your team's working on, your leading indicators. Your mention rate, your citation rate, your share of voice, map them to the placements that you've gained and the pages that your team has worked on so that the chain from activity to visibility to result is going to sit there for somebody to be able to understand in a report. That's what a defensible AI search report looks like in a channel that cannot be fully measured. And it gets more convincing the longer that you successfully are maintaining it. Okay, so let's pull the series together. Part one, we talked about building the foundation, the prompt strategy, your competitor set, and the pages and the entity work that is going to make your company legible and understood to these systems. And then we talked about in part two, the top of the funnel, and why you can produce informational content now to earn the trust of these systems and also the people that are reading them. And then part three. We covered the bottom of the funnel and why being named when the question has commercial intent attached to it is one mostly on getting your brand mentioned on other people's websites. And then today covered how you measure all of that in two layers, and then also how you report on it to people who want more certainty than the channel can natively give. These four parts are just one big program, and each one of them feeds on the next. And the measurement is what tells you which of the first three are going to need to get more attention or focus. So that's it for me. Thanks for listening to this week's Search Signal. If this series was useful, I would highly recommend you go ahead and subscribe so our next episode finds you. I will see you next week.