Latest / Found in AI: AI Search Visibility, SEO, & GEO / Schema.org's New Dataset + Claude Fable 5: What You Need to Know

Transcript

Schema.org's New Dataset + Claude Fable 5: What You Need to Know · Found in AI: AI Search Visibility, SEO, & GEO

• 0:41
  SPEAKER_00: Hey, welcome back to Found in AI.
• 0:43
  SPEAKER_00: I'm Cassie Clark, a fractional content strategist, an AI search optimization expert, and the host of the show where we talk about AI search, GEO, AEO, and what all of this means so we don't get lost in this new wave of user search behavior.
• 0:58
  SPEAKER_00: Today is Thursday, June 11th, and we've got two updates this week.
• 1:02
  SPEAKER_00: The first is from schema.org.
• 1:04
  SPEAKER_00: Last week they announced a public usage statistics data set in collaboration with Google that for the first time shows you how structured data is actually being used across the web.
• 1:15
  SPEAKER_00: The second is from Anthropic, who launched Cloud Fable 5 earlier in the week.
• 1:20
  SPEAKER_00: It's their most capable, generally available model yet.
• 1:24
  SPEAKER_00: And the first model in their new tier that they're calling Beethos Class.
• 1:28
  SPEAKER_00: One story is about how the web describes itself to machines, the other is about how smart those machines are getting.
• 1:34
  SPEAKER_00: So let's dig in.org because this is the one that most directly affects digital marketing.
• 1:51
  SPEAKER_00: If you've been listening for a while, you know that I talk about structure constantly.
• 1:55
  SPEAKER_00: The S in the FSA framework, Fresh of Structure Authority, is machine legibility.
• 2:01
  SPEAKER_00: It's how clearly an engine can read, extract, and reuse what you've published.
• 2:06
  SPEAKER_00: Schema markup is a big part of that.
• 2:08
  SPEAKER_00: It's the structured data that tells a machine, hey, this is a product, this is a price, this is an author, this is an FAQ.
• 2:17
  SPEAKER_00: Pretty much just kind of labels everything.
• 2:19
  SPEAKER_00: Here's what's changed.
• 2:21
  SPEAKER_00: Until now, when you were deciding which schema types to implement, you were mostly guessing.
• 2:27
  SPEAKER_00: You could look at documentation, you could copy what a competitor did, you could follow whatever your STL plugin defaulted to.
• 2:35
  SPEAKER_00: There was really no big picture view of what the rest of the web was actually using.
• 2:40
  SPEAKER_00: But now there is.
• 2:41
  SPEAKER_00: Schema.org working with Google, they publish this data set that samples a huge slice of the public web and then shows how those different schema terms like person or product properties like price or telephone, how often those are used across millions of unique domains.
• 3:00
  SPEAKER_00: It's updates, it updates monthly.
• 3:02
  SPEAKER_00: It's free on GitHub and CSV or JSON, and they're putting the stats directly on the schema term pages so that you can see adoption right where you'd go to look up the term anyway.
• 3:14
  SPEAKER_00: There are a couple details that we need to call out because it kind of tells you how to read this thing.
• 3:19
  SPEAKER_00: First, it's counted by domain, not by page.
• 3:22
  SPEAKER_00: So if you use a term on one page or a thousand pages on your website, that only counts as one.
• 3:29
  SPEAKER_00: So this is measuring how many sites adopt a term, not how heavily one site leads on it.
• 3:35
  SPEAKER_00: This is really an adoption signal, not so much a volume signal.
• 3:39
  SPEAKER_00: Now, second, they don't give you exact numbers either.
• 3:42
  SPEAKER_00: They group everything into ranges.
• 3:44
  SPEAKER_00: Buckets like 100,000 to 1 million domains.
• 3:48
  SPEAKER_00: They did this on purpose for two reasons.
• 3:51
  SPEAKER_00: To filter out the daily noise that you get from imperfect web crawling.
• 3:55
  SPEAKER_00: That's a big one.
• 3:56
  SPEAKER_00: And then also to protect privacy so nobody can reverse engineer how a specific site is being crawled.
• 4:03
  SPEAKER_00: So really don't go looking for precision here.
• 4:05
  SPEAKER_00: We're just looking at the shape of the thing.
• 4:08
  SPEAKER_00: Now, what do you actually do with this?
• 4:10
  SPEAKER_00: I'm so glad you asked.
• 4:11
  SPEAKER_00: There are three things.
• 4:13
  SPEAKER_00: One, prioritize.
• 4:15
  SPEAKER_00: If you're an early stage team with limited dev time, you don't need to implement every schematype under the sun.
• 4:21
  SPEAKER_00: You can look at what's widely adopted, well understood, and clearly relevant to your business and then start there.
• 4:27
  SPEAKER_00: Widely used terms are the ones the engines are most practiced at parsing, and the structure of working with the grade instead of against it is gonna do you some good.
• 4:37
  SPEAKER_00: Two, you can now justify the work.
• 4:39
  SPEAKER_00: Cannot tell you how many times structured data gets deprioritized because someone on the team says, Well, does this even matter?
• 4:47
  SPEAKER_00: Now you have an official public data set that you can point to.
• 4:51
  SPEAKER_00: That's not a small thing when you're trying to get engineering hours for something that's not as flashy as, say, a new feature for your product.
• 4:59
  SPEAKER_00: Now three, don't dismiss the small buckets.
• 5:03
  SPEAKER_00: There's a bucket for terms used on fewer than a thousand domains.
• 5:07
  SPEAKER_00: And schema.org is explicit that this includes both brand new terms and highly specialized ones.
• 5:14
  SPEAKER_00: Think medical, legal, government.
• 5:17
  SPEAKER_00: A niche term staying in a low bucket doesn't mean that engines ignore it.
• 5:21
  SPEAKER_00: It often means you're in a specialized territory where the total addressable web is just well, it's just smaller.
• 5:27
  SPEAKER_00: And using those terms builds deep authority for your specific niche.
• 5:32
  SPEAKER_00: That maps straight to the A and FSA.
• 5:35
  SPEAKER_00: If you're in a vertical specificity as an asset, not a liability, so do keep that in mind.
• 5:41
  SPEAKER_00: Now there's one more thing I want to say about this.
• 5:43
  SPEAKER_00: Schema.org is openly inviting other crawlers and indexers to publish their own statistics in the same open format.
• 5:51
  SPEAKER_00: That's the same pattern that I keep pointing at on the show.
• 5:55
  SPEAKER_00: Remember when Bing shipped their AI performance dashboard back in February, and I said that the big story here wasn't Bing's market share, but it was that a major player was building infrastructure around AI visibility.
• 6:08
  SPEAKER_00: This is another brick in that wall.
• 6:10
  SPEAKER_00: The semantic web is getting measured out in the open, and when things get measured, it gets managed.
• 6:17
  SPEAKER_00: The guesswork era is ending.
• 6:19
  SPEAKER_00: Okay, let's talk about Claude Fable 5.
• 6:22
  SPEAKER_00: This week Anthropic launched what they're calling a Mythos class model, a new tier that sits above their previous model.
• 6:29
  SPEAKER_00: Fable 5 is a version they've made safe for general release.
• 6:32
  SPEAKER_00: If you remember me talking about Mythos when it launched a few weeks ago, they pretty much said it's too powerful for general use.
• 6:39
  SPEAKER_00: So Fable 5 is the solution to that, and they're calling it a state-of-the-art technology on nearly every benchmark they tested, which was software engineering, knowledge work, vision, and scientific research.
• 6:54
  SPEAKER_00: And the longer and more complex the task, the bigger is lead.
• 6:59
  SPEAKER_00: Now, I can already hear some of you saying, hey Cassie, this is a content show.
• 7:03
  SPEAKER_00: We're talking about AS Search visibility.
• 7:05
  SPEAKER_00: Why are we talking about a coding model?
• 7:07
  SPEAKER_00: Well, here's why.
• 7:08
  SPEAKER_00: The engines that we optimize for, like Chat, GPT, Gemini, Perplexity, Claude, any of them, they are getting dramatically more capable of the exact things that decide whether you get cited.
• 7:20
  SPEAKER_00: They're getting better at retrieval, they're getting better at synthesis, they're getting better at reading, they're getting better at reasoning across a lot of sources at once.
• 7:29
  SPEAKER_00: When that underlying model gets smarter, their bar for what earns the citation moves with it.
• 7:34
  SPEAKER_00: Now, they had a big update here, so let me pull out the pieces that actually touch our work with this particular model.
• 7:40
  SPEAKER_00: First up is vision.
• 7:42
  SPEAKER_00: Fable 5 is by their account the new state of the art for vision task.
• 7:47
  SPEAKER_00: This means it can pull precise numbers out of detailed scientific figures.
• 7:51
  SPEAKER_00: It can rebuild a web app from nothing but screenshots.
• 7:54
  SPEAKER_00: Think about what that means for your content.
• 7:57
  SPEAKER_00: Engines are getting better at reading the visual parts of your pages, like your charts, comparison tables, diagrams, anything like that.
• 8:06
  SPEAKER_00: Not just the text on your page.
• 8:08
  SPEAKER_00: For a long time, if your key insight lived inside of an image, it was effectively invisible to machines.
• 8:14
  SPEAKER_00: Those were there for your human readers, but now that's changing.
• 8:18
  SPEAKER_00: Which means that the structure discipline that we apply to text also now extends to the visuals.
• 8:24
  SPEAKER_00: Go in and label your charts, caption your figures, make your tables clean and real, not just screenshots of tables.
• 8:32
  SPEAKER_00: The machine can actually see them now.
• 8:35
  SPEAKER_00: Then we have memory and long context.
• 8:37
  SPEAKER_00: These models can hold focus across millions of tokens and improve their own work using notes they take along the way.
• 8:44
  SPEAKER_00: This means an engine can hold more of your content and more of your competitors' content in a view at once, then weigh it all together.
• 8:52
  SPEAKER_00: I'm gonna say it until I blew a face in the face.
• 8:55
  SPEAKER_00: Shallow, thin coverage gets exposed faster when the model can actually see the whole landscape.
• 9:02
  SPEAKER_00: Depth is going to win in AI search visibility.
• 9:05
  SPEAKER_00: That's the authority again in play here.
• 9:08
  SPEAKER_00: Now, as these models get smarter, the cheap tricks get less effective.
• 9:13
  SPEAKER_00: Stepping conversational keywords, spinning up those thin pages, gaming the system with volume or prompt injecting, those kind of things are not gonna work.
• 9:24
  SPEAKER_00: A more capable, more discerning model is harder to fool, not easier.
• 9:28
  SPEAKER_00: It rewards the boring stuff, the durable stuff.
• 9:31
  SPEAKER_00: Content that's clear, contents that's fresh, content that is backed by a brand that exists across the web.
• 9:38
  SPEAKER_00: Yeah, freshness, structure, and authority.
• 9:41
  SPEAKER_00: The smarter the engine, the more those fundamentals matter.
• 9:44
  SPEAKER_00: Now, I'm gonna add a couple caveats here, because I always try to.
• 9:48
  SPEAKER_00: Fable 5 ships with safeguards that are deliberately tuned to be cautious.
• 9:52
  SPEAKER_00: Anthropic says some harmless requests will get caught up and rerouted to a different model while they tune it.
• 9:59
  SPEAKER_00: There are a few things that trigger this, like chatting with it about certain topics like cybersecurity.
• 10:03
  SPEAKER_00: I actually experienced this while chatting with Claude about Fable 5 itself.
• 10:08
  SPEAKER_00: It wouldn't continue on Fable 5 and then it just defaulted to Opus 4.8 for, and they quote safety, which was odd because we're just talking about the model.
• 10:18
  SPEAKER_00: Another thing on subscription plans, full access is really on stages rather than all at once.
• 10:23
  SPEAKER_00: So if you go test it and it feels a little uneven in spots, that's expected for a launch this big.
• 10:29
  SPEAKER_00: Don't read that as early friction, read the capability as the signal.
• 10:34
  SPEAKER_00: So let's put all of that together.
• 10:36
  SPEAKER_00: Schema.org is hardening the infrastructure, it's making structured data transparent and measurable.
• 10:41
  SPEAKER_00: Anthropic is pushing the intelligence, it's making the engines that read all of the things sharper.
• 10:48
  SPEAKER_00: Think of this as the plumbing and the brain kind of moving at the same time.
• 10:53
  SPEAKER_00: And both of them point back to the exact same playbook: freshness, structure, and authority.
• 10:58
  SPEAKER_00: Make your content legible to machines, make it deep enough to survive a model that can actually see the entire playing field and keep it current.
• 11:07
  SPEAKER_00: Build authority that travels beyond your domain.
• 11:10
  SPEAKER_00: None of that is new advice, it's just getting harder to ignore and easier to measure.
• 11:15
  SPEAKER_00: If you've already been building on the FSA principles, this week is more evidence, you're pointing it in the right direction.
• 11:20
  SPEAKER_00: If you're starting to take AI visibility seriously, this is a good moment to lean in.
• 11:25
  SPEAKER_00: While the tools to measure it, they're still new, and while most of your competitors are still kind of thinking, ah, that's a Sunday thing.
• 11:35
  SPEAKER_00: No, it's a today thing.
• 11:37
  SPEAKER_00: Hey, I'm Cassie Clark.
• 11:38
  SPEAKER_00: If you want help auditing how AI engines actually read your content before you start changing things, you can find more in the show notes or at CassieClark Marketing.com.
• 11:47
  SPEAKER_00: That's it for this week.
• 11:48
  SPEAKER_00: I will see you on Tuesday.
• 11:49
  SPEAKER_00: Until then, stay visible.