One Sunday afternoon in 2009, Eric Franzon’s wife took their daughter out of the house and he found himself with four unexpected hours.
He spent 45 minutes of it adding structured data to his wife’s website. She and her business partner ran a nail and waxing salon in West Hollywood for mothers of young children and pregnant women who didn’t want to sit in the fumes. Very niche. He was the webmaster.
Within an hour of tweeting about it, a friend at MIT sent him a screenshot: a satellite view of the strip mall, a map pin, and attached to that pin the address, the lat/long, the phone number, the hours, the services, and the prices. All of it read straight off the back end of the site.
Within a month, the salon dropped its $1,500 Google AdWords bill. Organic foot traffic had replaced it.
Eric has been working with structured data since 2004, which is seven years before schema.org existed. He was an Advisory Committee rep at the W3C. Schema.org was announced on stage at a conference he organized. His client list runs from RAND, Intel, and the official Simpsons website to a labrador breeder in Montana.
He came on the show for one reason: schema is not that complicated, nobody explains it clearly, and he wants to fix that.
Key Takeaways
- Schema is a vocabulary. A formal set of terms, like a dictionary, that describes things, their characteristics, and the relationships between them.
- Schema is a delivery system for ranking factors rather than a ranking factor itself. Reviews, services, locations, and people are the ranking signals. Schema is how you hand them over cleanly.
- A plugin’s default schema is a machine translating human words for a machine. It gets you the basics. It cannot tell Google you’re a veteran-owned plumber who works three neighborhoods.
- Wrong schema is worse than no schema. You are amplifying an incorrect signal, loudly and on purpose. Google does not reward sites that assert things that aren’t true.
- AI raises the stakes rather than lowering them. LLMs don’t have their own search engines, hallucinations are getting harder to spot, and asserted facts are how you keep a machine from making things up about your business.
- Test before you trust. Pick your three most important pages, run them through validator.schema.org and Google’s Rich Results Test, and read what you actually told the search engine.
Episode at a Glance
Eric Franzon runs Smart Data Websites in Missoula, Montana, and has been encoding structured data since before it had a name marketers would recognize. In this conversation he walks through what schema is, what it isn’t, why the plugin defaults on most WordPress sites are only a starting point, and why the AI era needs more structured data rather than less.
Action step: Pick the three pages that matter most to your business. Drop each URL into validator.schema.org, then into Google’s Rich Results Test. Read what your site is actually claiming. If your services page says “article,” you found your first fix.
Main idea: Schema is a delivery system for ranking factors. It lets you assert facts about your business directly instead of hoping a crawler guesses right from your text.
Why it matters: Bot traffic now outpaces human traffic on the web, and LLMs are assembling answers about your business from whatever signals they can find. Clear, accurate, human-curated structured data is how you stay in control of what they say.
Quick tool: validator.schema.org and Google’s Rich Results Test. Two free checks, three pages, fifteen minutes.
Listen & Subscribe
Podcast page: https://artillerymedia.com/podcast/
Apple Podcasts: https://podcasts.apple.com/us/podcast/slow-steady-a-podcast-by-artillery/id1840249614
Spotify: https://open.spotify.com/show/37ZzVcI8N7xnEDqkap2vCW?si=70b138e65b8c42ca
Follow along:
Instagram: https://www.instagram.com/artillery_media/?hl=en
Jake on X: https://x.com/jakekramer15?lang=en
Facebook: https://www.facebook.com/artillerymedia
Contact us: https://artillerymedia.com/contact/
About Eric Franzon
Eric runs Smart Data Websites in Missoula, Montana, where he builds sites, manages them, and does the schema work most agencies skip. He has 25+ years in web development and has been implementing structured data since 2004.
- Advisory Committee representative to the World Wide Web Consortium (W3C) for several years
- Former VP of Dataversity, where he launched the first non-academic conference on the commercialization of semantic technology
- Schema.org was announced from the stage of that conference in 2011, with Google, Microsoft Bing, Yahoo, and Yandex on it together
- Clients have included RAND, Intel, UPS Stores, Wondery Podcasts, the official website of The Simpsons, and a labrador breeder in Montana
- Has also taught improv, blues harmonica, early childhood education, and gender studies
His definition of slow and steady: “I have built my business on relationships, earning a reputation as ‘The web guy you wish you’d hired in the first place.’ Relationships take time. A truly successful website that supports business goals and business growth takes time. SEO takes time.”
Smart Data Websites: https://smartdatawebsites.com/
Free 30-minute consult for Slow & Steady listeners: https://smartdataseo.as.me/slowandsteady
Email him directly: eric@smartdatawebsites.com
Highlights & Timestamps
00:00–04:10: The Web Guy You Wish You’d Hired
How Smart Data SEO became Smart Data Websites, and the client calls that produced the tagline. “My web developer left in the middle of the night and I don’t know where they are.”
04:10–09:44: A Nail Salon, 45 Minutes, and a Dead $1,500 Ad Bill
2004 at Dataversity, Yahoo’s SearchMonkey, Best Buy’s local results, and the Sunday afternoon Eric proved this works for businesses without a technical bench.
09:44–14:42: What Schema Actually Is
Facebook’s like button, the Open Graph protocol, and the 2011 announcement on his own stage. Classes, properties, and the guitar hanging on Jake’s wall.
14:42–20:26: The Kansas City Plumber Test
Woman-owned, veteran-owned, service radius, differentiators. Why the long tail is where schema earns its keep, and what it means to assert facts instead of hoping a crawler guesses.
20:26–25:06: Schema Is a Delivery System for Ranking Factors
The first myth to dispel, how schema feeds E-E-A-T, and the problem with letting a machine translate your human words for another machine.
25:06–35:20: Where the Plugins Stop
Cookie-cutter defaults, WordPress’s blogging history turning every page into an “article,” branded search, and the mole problem. Also the Far Side cartoon that explains how this gets out of hand.
35:20–43:30: The Simpsons vs. the Plumber, and How to Test Your Own Site
Scope and access at Fox and Wondery, JSON-LD in plain English, the two free validators, and why improperly written schema is worse than none. “If you’re amplifying incorrect data, that’s a lie.”
43:30–1:03:00: Schema in the AI Age
Why LLMs need more structured data, not less. Plus other vocabularies, improv, refusing to niche down, selling patience, and the time AI invented three schema classes that don’t exist.
Schema in Plain English
Start with the like button.
In 2010, Facebook freed it from Facebook. You didn’t need a page for your business anymore. You could put the button on your own site, and it worked because of a set of terms you added in the back end to describe your content. That set of terms is the Open Graph protocol. A vocabulary.
A year later, schema.org was announced from the stage of Eric’s conference. Google, Microsoft Bing, Yahoo, and Yandex, standing up together. As Eric puts it, the fact that they were on that stage at all was huge. It meant four competitors had each gone through their C-suite and their legal department to get there.
What they announced was a bigger version of the same idea. A formal set of terms. A dictionary. Their pitch to publishers: put these terms in your site to describe your content, and we will each do interesting things with it. We’re still competitors, so we won’t tell you what.
Schema describes three things:
- Classes. The things. People, places, products, services, organizations.
- Properties. The characteristics of those things. Size, color, price, hours, author.
- Relationships. How the things connect. This business has a founder. That founder is a person.
Jake asked him to run it on the guitar hanging on the wall behind him. Depending on what the page is for, that guitar is a product, a musical instrument, or a decoration. It has a size, a color, a wood grain, maybe a price. It has accessories: a strap, a case, extra strings for the campfire. All of that can be encoded.
At launch, schema.org had about 300 classes and 187 properties. Today it’s over 900 classes and around 1,500 properties.
It’s a dictionary. Somebody still has to write the sentences.
The Kansas City Plumber
Say you own a plumbing company. Here’s what Eric wants to know.
What are your differentiators? Are you woman-owned? Veteran-owned? Family-owned? Where is your business located, and separately, what area do you actually serve? Do you have a fleet that runs a few hundred miles, or do you work two neighborhoods well?
None of that shows up in your daily decision-making. You don’t think about being veteran-owned while you’re unclogging a drain. But somebody out there is typing it into a search bar, and with AI in the mix, that long tail keeps getting longer. People don’t just search “plumber near me” anymore. They tell an assistant what kind of business they want to hire.
What schema does with all of it is assert facts.
Instead of relying on a bot to crawl your pages and match some strings of characters, you say it outright: this is a business, it has a logo, it has a founder, that founder is a person, that person has these credentials, this business offers these services, those services are available to this audience at this price.
Two things come from that. You get more control over what’s said about your business. And the search engines get a far more efficient read of your site.
Efficiency matters more than it used to. As Jake pointed out on the recording: in AI, more efficient means fewer tokens.
Schema Is a Delivery System for Ranking Factors
This is the myth Eric wanted to kill first.
Schema is a delivery system for ranking factors. It is not a ranking factor on its own.
Reviews are a ranking factor. Schema is how you hand a search engine a very specific review: this is a review, of this business, it has five stars, it has an author, that author is a person, that person has a name and a photo. A human glances at a review block on screen and absorbs all of that instantly. A computer sees a JPEG it knows how to render and text it has to guess at.
Two rules go with it. The content has to appear on the human-facing side of the page, and it has to be true. Beyond that, you choose what to emphasize. If a customer sends three paragraphs and one sentence is the one you want in front of people, that’s the one you encode.
Do this consistently and you’re building toward E-E-A-T: experience, expertise, authoritativeness, trustworthiness. The thing Google keeps saying it’s looking for.
You’re not gaming anything. You’re labeling your own building so the inspector doesn’t have to guess.
Where the Plugins Stop
Yoast, Rank Math, SEO Press. Eric is a WordPress shop and he isn’t against any of them. Plugins do real work.
But here’s the tension at the center of this episode:
“If you use a plugin, you’re having a machine translate the human words for a machine.”
Schema exists to translate human knowledge so machines can use it. Hand that translation to a machine and you’ve removed the only part that understood the meaning.
What a plugin gets you is the general business layer. Name, address, hours. Those core basics are genuinely worth having, and if that’s all you have, you have something. What it can’t do is say this business is a plumber who specializes in the line running from the house to the street.
The deeper issue is WordPress’s own history. It started as a blogging tool, so the early default was that everything is a post, and every post is an article. Your founder’s bio page becomes an “article” whose title happens to be a human name and whose body happens to be a biography. You and I read that page and see a person. The computer sees text.
That’s a real cost. A lot of people search for the owner or the salesperson by name, something you notice fast once you take Google Search Console seriously. Get the person markup right and you’re feeding the knowledge panel. Leave it as an article and you’ve diluted the meaning of the page.
Eric’s own analogy landed better than ours: schema is plumbing. You can put anything you want down the drain. That doesn’t mean it looks good coming out the other end.
Wrong Schema Is Worse Than None
Jake told on himself for this one.
Early on, he asked ChatGPT for schema for his blog posts, pasted it in, ran the validator, and got a green light. What he didn’t catch was six different types stacked on top of each other, each one technically correct, all of them fighting. He eventually stripped the whole thing out.
Eric’s version is worse and funnier. He asked an AI to write schema. The vocabulary has 900+ classes. The AI decided that wasn’t enough and invented three more, plus properties to go with them. Search engines will never use any of it, because none of it exists.
And the code was pristine. Syntactically perfect. It would sail through any linter. It took a human who knew schema.org to look at it and say: I have never seen that class in my life.
“If you’re amplifying incorrect data, that’s a lie. You’re lying. And Google doesn’t like it when sites are lying.”
Which is why the last step is the one people skip. Your plugin’s green light means schema exists on the page. It says nothing about whether the meaning is right.
Pick three pages. The homepage, the about page, a top service or product. Drop each URL into validator.schema.org: raw HTML on the left, what it found on the right. Then run the same page through Google’s Rich Results Test, which shows you the subset Google actually acts on. Schema.org isn’t owned by Google, so those two tools tell you different things.
Then compare it to your Google Business Profile. Same hours? Same phone number? Inconsistency is its own signal, and not a good one.
Fifteen minutes, three pages, two free tools. That’s the whole starting move.
Schema in the AI Age
Eric opened this section by pointing listeners back to Episode 15 with Michelle Bourbonniere on how AI search actually works. His philosophy and hers intersect, and he wanted her to get the credit.
His position is straightforward. In an AI world we need more structured data, not less.
Start with the part everyone forgets: LLMs don’t have their own search engines. They’re relying on other platforms to find things. So the signals you send still matter, and they matter in the same places they always did.
Then the part that should get your attention. As LLMs evolve, hallucinations are going to get more frequent and harder to detect. The counterweight is asserted fact. This is our official logo. These are our hours. These are the services we offer, and why. These are our real reviews. The more a business states plainly and consistently, the less room a model has to invent something.
Schema also solves a problem that predates AI: disambiguation. Type M-O-L-E into an old search engine and you get a small furry animal, a spot on your skin, an ABC show about spies, a Mexican chocolate sauce, and a unit of scientific measurement. Structured data lets you say which one you meant. Bring it back to the Simpsons, as Eric did, and the question becomes which Springfield you’re talking about. There are a lot of them.
Schema isn’t the only vocabulary that can do this, either. Open Graph is still in use. Dublin Core predates schema.org by years. MusicBrainz describes recordings down to the track and the performers. Wikidata is a structured version of everything in Wikipedia, and Eric used it on the artillerymedia.com rebuild to point search bots at a specific city and remove all doubt about which one we meant.
The thing he’s most excited about is the least technical thing he said all hour. All this pressure around AI is pushing everyone back to the same place: you have a website, so let’s make it the best possible website for your target customer. Answer real questions for real people. Then encode those answers so the machines get it right.
Great content first. Structured data second. In that order.
Improv, Patience, and Refusing to Niche Down
Eric has taught improv, blues harmonica, early childhood education, and gender studies. Jake asked how the improv shows up in the schema work.
Two tenets, both about people.
Listen. He can’t understand a client if he isn’t hearing what they aren’t saying. And “yes, and” instead of “no, but.” This stuff is technical enough that people get embarrassed about what they don’t know. Meet them where they are, confirm what they’ve got right, then build on top of it.
He also said no to the advice everyone gets. Niche down, pick an industry, own a silo. He looked at his actual work and found the thing that set him apart traveled across every size and type of business he served. So the niche became a promise instead of a vertical: the web guy you wish you’d hired in the first place.
And then patience, which is the hardest sale in SEO. Eric’s answer is to be honest about the limits of the job:
“I can’t make Google do anything. What I can do is amplify signals so that the likelihood that Google’s going to pick up that signal is stronger.”
He won’t guarantee outcomes. He starts by asking a client what success looks like to them, which is where the myths surface and the expectations get set. What he’ll say is that after twenty-plus years of doing this, he has a strong feeling about what happens when your website, your Google Business Profile, and your reviews all tell a machine the same true story.
Most of your competitors haven’t done it. That’s the opening.
Call to Action
Want help making sure your website is telling search engines and AI the right story about your business?
Start here: https://artillerymedia.com/contact/
And if the question is specifically about schema, email Eric: eric@smartdatawebsites.com. He offered listeners a free 30-minute consult at https://smartdataseo.as.me/slowandsteady.
Show Credits
Hosts: Jake Kramer & John Wooten
Guest: Eric Franzon (Smart Data Websites, https://smartdatawebsites.com/)
Produced by: ARTILLERY, Lincoln, Nebraska
Slow & Steady — A Podcast by ARTILLERY 🎧
https://artillerymedia.com/podcast/
