AI Needs Facts It Can Trust

AI can mention your brand without understanding it. This episode shows why semantic triples and consistent entity facts across the web decide which brands AI actually recommends.

Transcript

SPEAKER_00 0:13

What if I told you that uh publishing another 50 blog posts for your biggest client this week won't just waste their budget, but it will act it'll actually actively destroy their visibility.

SPEAKER_01 0:27

Yeah. I mean, it sounds completely counterintuitive, right? But it's true.

SPEAKER_00 0:30

Right. Because for the last 20 years, getting your clients found online felt a bit like, I don't know, shouting in a crowded room. You bought the biggest megaphone, you shouted your keywords the most times, you published the most pages, and eventually, well, the search engines would point to you.

SPEAKER_01 0:47

Exactly. That was the entire agency model. But today, if you're a digital marketing agency owner or you're the one making the big strategic calls for your business, you really need to realize that we are standing in a completely different room now. Aaron Powell Yeah.

SPEAKER_00 1:00

A room where the lights are totally off. And instead of a crowd, you're talking directly to an incredibly smart but uh incredibly literal artificial intelligence.

SPEAKER_01 1:10

Aaron Powell And that shift, I mean, moving from traditional search engines to these AI-driven answer engines, it requires a total dismantling of the old agency playbook.

SPEAKER_00 1:18

Aaron Powell, which is exactly what our mission is for this deep dive today. We're going to break down this massive shift to agentic optimization and figure out how you can actually stop wasting money on tactics that are just dead.

SPEAKER_01 1:30

Aaron Powell We have a phenomenal stack of sources to guide us through this today, too. All fresh dispatches from late June 2026. We're pulling from uh reports in Search Engine Journal, insights from Aleda Solis' CFOMO weekly newsletter, this really brilliant essay from a publication called Search as a Channel, and deep strategic notes from Steve Toth's AI Notebook.

SPEAKER_00 1:52

Aaron Ross Powell Right. An AI Notebook is actually the world's first strategy-focused newsletter dedicated entirely to LLM optimization, right? Like specifically large language models.

SPEAKER_01 2:01

Aaron Powell Yeah, exactly. It's incredibly relevant right now because we are truly at a pivot point in the industry.

SPEAKER_00 2:05

Aaron Powell We really are. I mean, Search Engine Journal just recently reported the sad passing of Bruce Clay, who is, you know, literally one of the founding figures of the whole SEO industry. Rest in peace. It really feels like the baton is officially being passed from the pioneers of the early web to this entirely new paradigm.

SPEAKER_01 2:22

Aaron Powell It does. And at the exact same time, you have Google CEO Sundarpachai out there downplaying the whole Google Zero fear, you know, that panic that AI is going to steal all website traffic forever.

SPEAKER_00 2:34

Oh, yeah, everyone's terrified. No one will ever click a link again. But looking at these sources, it's not that the traffic is gone. It's that the rules of how you actually get that visibility have fundamentally changed.

SPEAKER_01 2:45

Right. It's not just about getting your brand mentioned anymore.

SPEAKER_00 2:48

Aaron Powell Exactly. That's the huge takeaway here. In 2026, just having an AI scrape your brand name isn't enough. If the AI doesn't actually truly understand what you do, you just won't get recommended to buyers. Trevor Burrus, Jr.

SPEAKER_01 3:00

Because agenc optimization means we aren't just trying to rank on a list of blue links anymore. We are trying to be the recommended solution when an AI agent is tasked with actually solving a user's problem.

SPEAKER_00 3:12

Okay, let's unpack this. Because before we figure out how to get AI to understand a brand, we have to talk about the old strategies that are actively backfiring right now.

SPEAKER_01 3:21

Yeah, the old more is better content strategy. According to some recent MIT research that was featured in SEJ, the sheer volume of output just it doesn't work the way it did even two years ago.

SPEAKER_00 3:33

Carolyn Shelby actually wrote a piece for SEJ that just drops an absolute bomb on the old retainer model. She reveals that publishing more content is actually making your SEO worse. Right. But wait, if an AI is, you know, hungry for information and training data, how does adding content hurt? It feels like we're telling people to stop throwing hay on the haystack when the AI is just looking for the needle.

SPEAKER_01 4:00

Aaron Powell Well, because AI retrieval systems don't index pages the way traditional crawlers do. They map concepts mathematically.

SPEAKER_00 4:06

Okay, so it's a completely different mechanism.

SPEAKER_01 4:08

Aaron Powell Exactly. So when an agency floods a client's blog with a hundred slightly different articles on the exact same topic, you know, maybe targeting 50 variations of a long tail keyword, they aren't casting a wider net anymore. They're actually creating overlapping, conflicting data points.

SPEAKER_00 4:23

Oh, wow. So you're diluting the AI's confidence in what that brand is actually about.

SPEAKER_01 4:28

Yes. You create noise in your own signal. It makes it so much harder for the AI to extract a definitive single answer about your client's core expertise.

SPEAKER_00 4:37

So we're essentially confusing the machine by overexplaining. And the internet is already drowning in noise as it is. I mean, SEJ reported this crazy statistic that TikTok is now showing three times more AI slop than YouTube.

SPEAKER_01 4:51

Three times more? It's staggering.

SPEAKER_00 4:53

It's gotten so bad that YouTube actually had to add automatic detection for undisclosed, photorealistic, AI-generated content just to protect their platform.

SPEAKER_01 5:02

And that proliferation of slop is exactly why Google is changing the rules so aggressively. Their late June STEM update is officially targeting and penalizing AI answers and AI manipulation.

SPEAKER_00 5:12

Because their entire business model depends on providing accurate answers, right?

SPEAKER_01 5:16

Exactly. If their AI overviews are easily manipulated by low-quality mass-produced agency content, users lose trust and leave. So the algorithm now actively filters out that volume-based spam and rewards what we call semantic precision.

SPEAKER_00 5:28

Okay, so pushback here. Are we supposed to tell clients to just fire their content writers and stop publishing altogether?

SPEAKER_01 5:34

Aaron Powell No, no. Google explicitly states that AI visibility still hinges on content people actually want to read. It's not about stopping content creation.

SPEAKER_00 5:42

Right. So instead of 50 blog posts defined what a CRM is, you need like one definitive, perfectly structured resource that proves why your client's CRM is the best for a very specific use case.

SPEAKER_01 5:56

Yes. It's about shifting from volume to precision. You're no longer creating content just to feed a crawler and hit a monthly keyword quota. You're creating it to build a factual, authoritative footprint that an AI can parse, verify, and actually trust.

SPEAKER_00 6:11

Okay, so if the old volume playbook is dead, what is the exact mechanism that agencies need to use to achieve this semantic precision? Like how do we actually teach the machine without writing a novel?

SPEAKER_01 6:22

This is where Steve Toth introduces a brilliantly simple concept in his AI notebook. It's called the semantic triple.

SPEAKER_00 6:28

The semantic triple. Okay, what is that?

SPEAKER_01 6:31

It is the absolute foundation of LLM optimization. A semantic triple is just subject plus predicate plus object. You are providing structured, clean facts.

SPEAKER_00 6:42

So no marketing fluff.

SPEAKER_01 6:44

Exactly. So instead of a flowery, rambling paragraph about your client's history on their about us page, you feed the machine a triple. Brand X is the subject, is a B2B analytics platform, is the predicate and object. Brand X integrates with Shopify.

SPEAKER_00 7:00

Oh, I love this because it makes so much sense. It's like organizing a completely messy junk drawer into a highly indexed filing cabinet. The AI doesn't have to like read the files and guess what they mean, it just scans the labels.

SPEAKER_01 7:12

That's a great analogy. And that essay from Search is a channel really hammers this home. They point out that a basic brand mention just getting your name dropped in a blog post only tells a machine that your name exists.

SPEAKER_00 7:22

Which doesn't help if it doesn't know what you actually sell. Right.

SPEAKER_01 7:25

But a semantic triple tells the machine what your brand actually is. Steve Toth argues that mentions plus triples will beat mentions alone every single time.

SPEAKER_00 7:35

So the architecture of these models actively rewards this exact structure.

SPEAKER_01 7:39

Oh, absolutely. Aleta Solas actually highlighted a fascinating analysis of a Google patent in her CFOMO newsletter. The patent reveals that Google's strategic goal is literally to quote, teach AI who you are.

SPEAKER_00 7:51

Teach AI who you are. So they aren't looking for just a vague brand presence.

SPEAKER_01 7:56

No, they reward confidence. If the language model has a high confidence score that your client's software integrates with Shopify, it will recommend it to a user asking for Shopify integrations.

SPEAKER_00 8:07

But if that language is buried in five paragraphs of jargon, the confidence score drops and your client vanishes from the AI overview entirely.

SPEAKER_01 8:16

Exactly.

SPEAKER_00 8:16

So let me get this straight. If a customer review just says, you know, this agency is awesome, the AI essentially ignores it because it's subjective fluff. But if it says, this agency provides PPC for healthcare companies in Austin, the AI absorbs that as a hard fact. It's literally like feeding a calculator numbers instead of poetry.

SPEAKER_01 8:34

That is exactly it. When a large language model parses your site, it isn't reading for nuance or prose. It's extracting data points to plot on a massive semantic map. Clarity is the new competitive advantage.

SPEAKER_00 8:46

Because the AI model is a synthesizer.

SPEAKER_01 8:49

Right. If it can confidently connect hard facts about your business, you get the recommendation. If it has to parse through paragraphs of buzzwords just to figure out what you sell, it moves on to a competitor whose site is easier to process.

SPEAKER_00 9:03

Wow. So you get filtered out before a human buyer ever even sees your brand.

SPEAKER_01 9:07

You do. You're just invisible.

SPEAKER_00 9:08

Okay. So let's say I'm an agency owner. I listen to this. I immediately go to my client's website and I update the homepage with perfectly structured semantic triples. The filing cabinet is organized. Boom. We're numbers, not poetry. Am I done? Can I attend the invoice? Not even close. Ah, of course not. Why?

SPEAKER_01 9:25

Because the homepage is only a fraction of your client's digital identity. AI models build their picture of a brand from the entire web.

SPEAKER_00 9:33

Right. They scrape everything.

SPEAKER_01 9:34

Yeah. So your client's homepage might have perfect semantic triples, but consider their LinkedIn company page, their crunch base, or their profiles on G2 and Captera. Often those profiles were set up like five years ago by totally different departments using completely different messaging.

SPEAKER_00 9:51

Oh man. Which means you're forcing the AI to guess again.

SPEAKER_01 9:54

And contradictions are the absolute fastest way to kill AI trust. If the facts don't align perfectly across all those third-party platforms, the AI receives conflicting signals.

SPEAKER_00 10:05

So it's confidence core plummets and the brand is excluded from the final recommendation anyway.

SPEAKER_01 10:10

Exactly. And here's another trap. Even if your site is perfect, it might be literally invisible to the AI.

SPEAKER_00 10:16

Wait, really? How?

SPEAKER_01 10:18

SEJ had this wild report about the fintech industry specifically. They found that a third of fintech homepages are completely invisible to AI agents right now. One in three.

SPEAKER_00 10:28

One in three massive bank or financial websites. Why? Because they stip the rendering step. Aaron Powell Okay, explain the mechanics of that. Why is an AI just ignoring a massive bank's website?

SPEAKER_01 10:38

Aaron Powell This is a huge issue in technical SEO right now. A lot of modern sites, especially in fintech, rely heavily on client-side JavaScript to load their content. They want these dynamic app-like experiences.

SPEAKER_00 10:50

Right. They want it to look flashy.

SPEAKER_01 10:51

But executing JavaScript requires significant computing power and time. Many AI agents and scrapers navigating the web simply don't have the compute budget to render every script on every single page they visit.

SPEAKER_00 11:04

Oh. So to save resources, they just scrape the raw HTML.

SPEAKER_01 11:09

Yes. So if your beautiful semantic triples are locked behind a script that the agent doesn't render to the AI, your page is completely blank.

SPEAKER_00 11:17

Wow. That is an incredibly expensive mistake for a client to make. So you basically have to serve the facts on a silver platter in the raw HTML. The AI needs to see it immediately without having to think or process any extra code.

SPEAKER_01 11:29

Exactly.

SPEAKER_00 11:30

And speaking of where AI gets its facts, we have to talk about Reddit. Because Steve Huffman, the CEO of Reddit, actually called user content the modern oil for AI.

SPEAKER_01 11:40

He did. He flat out said LLMs wouldn't even exist without Reddit data.

SPEAKER_00 11:44

Which is wild.

SPEAKER_01 11:44

But it makes sense because language models desperately need conversational context, they need to understand how humans talk, and they need to source authentic opinions. Reddit provides this massive, constantly updating corpus of human interaction.

SPEAKER_00 11:56

It's so vital that SEJ actually reports buyers are literally buying Reddit right now to win AI citations.

SPEAKER_01 12:04

Yeah, because Reddit is essentially the new link farm of 2026.

SPEAKER_00 12:08

If the AI is looking there for the truth, you have to ensure the truth about your brand is represented there. So let me ask you this pointed question. If an AI pulls five third-party sources about my listener's brand today, is it going to find the exact same facts repeated, or is it going to find a total mess of half-finished, contradictory profiles?

SPEAKER_01 12:29

And that is exactly where most brands lose. Agencies often focus all their budget just on their clients' own properties. But AI recommendation quality only improves when your brand facts stop competing with each other across the entire internet.

SPEAKER_00 12:42

So if your client's story is fractured across the web, their AI visibility will be fractured.

SPEAKER_01 12:46

Exactly.

SPEAKER_00 12:47

Here's where it gets really interesting for our listeners, though. Understanding this problem is great. But how do decision makers actually operationalize this across their teams? Like what does the actual 2026 playbook look like for an agency?

SPEAKER_01 12:59

Well, Steve Toth provides a highly operational framework in his newsletter for this. He separates all off-site brand mentions into three distinct buckets.

SPEAKER_00 13:08

Okay, three buckets.

SPEAKER_01 13:09

Bucket one is company controlled and editable. Bucket two is community-driven and influenceable. And bucket three is editorial and completely out of your hands.

SPEAKER_00 13:18

Aaron Powell So you obviously start with bucket one, you audit the edible footprint first.

SPEAKER_01 13:22

Yes. The very first step is to stop treating directory listings, G2 profiles, and crunchbase as minor administrative tasks that you just hand off to an intern. Trevor Burrus, Jr.

SPEAKER_00 13:32

Right. They're primary AI visibility assets now?

SPEAKER_01 13:36

Exactly. Agencies need to go in and rewrite all of them for absolute factual consistency.

SPEAKER_00 13:41

Aaron Powell What does that actually look like in practice? Let's say I'm looking at a client's crunch base profile right now.

SPEAKER_01 13:46

Aaron Powell You strip out the marketing fluff. If the Crunchbase says we are a revolutionary synergy engine empowering seamless digital transformations.

SPEAKER_00 13:54

Ugh, the worst.

SPEAKER_01 13:56

Right. You delete it, you replace it with clean semantic triples. We are a sauce platform. We provide inventory management. We integrate with AWS.

SPEAKER_00 14:04

Keep it simple.

SPEAKER_01 14:04

And then you ensure those exact same triples are on LinkedIn, Facebook, industry forums, and the company website.

SPEAKER_00 14:12

Aaron Powell You really have to break down the silos in the agency to do this, though, because normally the PR team writes the crunch base, the social team runs LinkedIn, and the SEO team is the only one touching the website.

SPEAKER_01 14:22

Aaron Powell Which is a recipe for disaster now. SEJ actually had an article pointing out that search and agents are now essentially one product.

SPEAKER_00 14:30

Aaron Powell You can't have these teams operating in different realities. You need one single playbook.

SPEAKER_01 14:34

Aaron Powell The SEJ report calls for an integrated search brief. This is a foundational document that forces the alignment of SEO, PPC, and content teams around a single factual identity.

SPEAKER_00 14:46

Aaron Powell Because the whole ecosystem is becoming hyper-targeted and agent-driven, right?

SPEAKER_01 14:50

Trevor Burrus, yeah. So every department must be feeding the AI the exact same data points.

SPEAKER_00 14:54

Aaron Powell And the data totally supports this shift to extreme targeting. I mean SEJ reported that desktop click-through rates are actually climbing while mobile dips. Trevor Burrus, Jr.

SPEAKER_01 15:03

Which suggests people are doing deep, complex research sessions with AI rather than just quick scrolling.

SPEAKER_00 15:08

Aaron Powell Exactly. And Google is testing strongest match labels on search ads while simultaneously launching tools like Ask Ad Manager, which is their first AI agent specifically for publishers. Everything is being handed over to these incredibly specific, prompt-driven agents.

SPEAKER_01 15:25

The deliverables for agencies have to change to reflect this reality.

SPEAKER_00 15:29

But this is a massive opportunity to upsell clients. It's no longer just, you know, selling an SEO retainer for 10 blog posts a month or a standard digital PR package. You have to sell a cross-functional entity strategy. Aaron Powell Right.

SPEAKER_01 15:42

You're selling them a unified, machine readable identity across the entire Internet.

SPEAKER_00 15:46

Aaron Powell, which is way more valuable.

SPEAKER_01 15:48

It is, but the SEJ sources offer a really strong word of caution here. Selling AI simply as a replacement tool, like telling clients, hey, we use AI so we can fire our writers and charge you less, it might win some short-term attention, but it totally kills long-term trust.

SPEAKER_00 16:03

Oh, absolutely.

SPEAKER_01 16:03

The modern agency won't survive by using AI to just do the old jobs cheaper. They will survive by using AI to build a live data stack, aligning their clients' teams around a shared factual identity, and actually teaching the machines who their clients are.

SPEAKER_00 16:19

It really is a whole new ballgame. So stop chasing hollow brand mentions. Just seeing your client's name in a block of text isn't the win anymore. Start building a web of undeniable, consistent facts.

SPEAKER_01 16:31

And there is a massive silver lining here for the agencies that get this right. SEJ shared this incredibly motivating statistic. Daily AI overview users actually click on sources 3.5 times more than occasional users.

SPEAKER_00 16:45

Wait, really? Three and a half times more?

SPEAKER_01 16:47

Yes. The consumers who rely on AI the most aren't avoiding links. They are actively looking for trusted sources to validate the AI's answers.

SPEAKER_00 16:56

That makes total sense.

SPEAKER_01 16:56

So if you provide that factual clarity and become the source the AI cites, the traffic isn't gone, it's there, it's highly qualified, and it's literally waiting for you.

SPEAKER_00 17:06

So the pie isn't shrinking, it's just moving to a completely different table.

SPEAKER_01 17:09

Exactly.

SPEAKER_00 17:09

So what does this all mean for you, the agency leader listening to this? Stop throwing hay, start organizing that filing cabinet, and get your teams working from a single integrated brief. But before you head into your next strategy meeting, I want to leave you with one final thought to just mull over. We've talked a lot about teaching the AI who you are today. But consider how these large language models are actually built. They are trained on massive data sets that eventually get cut off or frozen to build the next iteration of the model.

SPEAKER_01 17:38

Right, the training cutoff day.

SPEAKER_00 17:40

Exactly. So if you wait until 2027 to clean up your client's factual footprint, what happens if the AI has already frozen its training data based on today's messy internet? Are you risking being permanently locked out of the AI's foundational memory? Think about that. Get to work on those semantic triples.