Hi everyone, this is Neo.

While scanning the SEO news this week, I read Search Engine Journal founder Loren Baker’s session recap, Schema For AI Citations: How To Become A Trusted Source. The biggest takeaway hit me between the eyes: most of us are still thinking about structured data like it’s 2019.

Back then, we added schema for star ratings, rich snippets, and prettier search results. In 2026, those priorities have shifted. The real value of schema is now about making AI trust you — and cite you.

In this post I’ll combine Loren’s framework, a few industry data points, and my own judgment to explain the whole thing: why schema is a trust builder for the AI era, what “four-source consistency” means, the traps ecommerce sellers fall into, and how to roll it out step by step.

Let’s Get Real: Schema Isn’t a Ranking Switch, It’s a Trust Builder

Loren opened with a blunt truth: schema markup won’t make AI systems cite you directly — but it does help engines like Google, Bing, and ChatGPT understand who you are, verify your claims, and decide whether to feature you.

Microsoft has said schema helps its LLMs understand your content. Google’s guidance notes structured data isn’t required for AI features but recommends it as part of a comprehensive SEO strategy. And OpenAI is even more direct: structured product data feeds what ChatGPT presents to shoppers.

One mindset fix matters here: don’t treat schema as a magic “add this and rank” switch. Engines compare your markup against your actual page, your business profile, your reviews, and what other people are saying about you online. Markup without substance behind it just makes the gaps obvious.

Loren’s analogy is perfect: think of your schema as your online dating profile — your first chance to introduce yourself to the engines. If what your profile says doesn’t match reality, that’s catfishing. Nobody wants to deceive Google, and definitely not ChatGPT.

The industry data backs the shift. Per Nudgenow’s analysis, 71% of pages cited by ChatGPT include structured data — yet only 18% of ecommerce product pages carry complete schema markup. That gap is the opportunity window.

The Core Framework: Four-Source Consistency

Loren kept returning to one idea that applies to every business model — four sources of information that must tell the same story:

1. Your website pages (the visible content)
2. Your business profile (e.g., Google Business Profile)
3. Your data feed (e.g., Merchant Center product feed for ecommerce)
4. Third-party information (reviews, directories, what people say about you elsewhere)

When these four sources corroborate each other, you give the engines a single, reliable reference point. When they contradict each other, even sophisticated markup undermines your credibility.

His example: the website says “Suite” but the markup says “Ste” — a tiny mismatch with a potentially big cost. Business name, phone number, opening hours, SKU, price, job title, author name… any field that changes across surfaces without explanation is a demerit in the machine’s eyes.

The Nine Pillars: Not One Can Be Missing

For local businesses, Loren broke the whole system into nine pillars:

1. The entity and a stable ID — everything hinges on this; without an entity, no other property has a reference point
2. NAP — name, address, phone
3. Geo coordinates
4. Opening hours
5. Service area
6. Conversion actions
7. Reviews
8. sameAs links — your official profiles on other platforms
9. Overall consistency

Here’s the most common point of confusion: “service area” on Google Business Profile is not the same as areaServed in schema. The GBP field describes where you dispatch or deliver; areaServed covers every area you serve. In a city with multiple locations, reusing generic markup across all of them wipes out exactly the distinctions that help an engine recommend the right one.

Loren demonstrated live: he asked ChatGPT for a physical therapist in Hoboken open that day, and it returned a client’s brand — with hours and services (orthopedic rehab, sports injuries, post-op rehab) that weren’t on the location’s landing page at all. They lived only in the schema and the Business Profile. Two out of three sources agreed, and that was enough for ChatGPT to decide.

This doesn’t mean you should hide useful info in structured data — visible content remains the foundation. But when sources align, an engine can identify your business accurately and match it to specific needs without guesswork.

Ecommerce Sellers, Read This: Your Merchant Center Feed Is the Source of Truth

For ecommerce sites, Loren’s advice is to treat your Merchant Center feed as the source of truth, then align your page schema with it field by field.

Key fields: product name, description, brand, images, SKU, MPN, GTIN, a stable product ID — then offer details like price, currency, availability, shipping cost, delivery time, return policy, and product variants.

Here’s a number that shows why completeness matters: per Nudgenow, Product schema combined with AggregateRating makes a product 3x more likely to appear in AI-generated recommendations than basic markup alone. And ChatGPT confirmed in late 2025 that it uses structured data to decide which products appear in its shopping results.

Deep attributes are becoming more important by the day. AI searchers don’t type keywords — they describe needs: not “brand running shoes” but “size 15, sky blue, delivered by Friday, comfortable for a bad ankle.” Material, weight, waterproofing, color, size, shipping speed, return policy — these deep attributes are how AI connects those detailed queries to your product.

The OutOfStock Trap

Loren flagged one detail that hits revenue directly: when your schema flips to OutOfStock the moment inventory hits zero, Google concludes you no longer sell the item — and you can lose the ranking you built, even if you restock days later.

The fix is to use Schema.org’s values for temporary statuses, distinguishing “temporarily out” from “permanently discontinued.” He said getting this one right often matters more than other property details.

Person Entities Don’t Make Experts — They Verify Them

Loren stressed a point many people misunderstand: adding a Person entity doesn’t make anyone an expert. Schema can only help a machine verify authority that already exists.

But the authority has to actually be there. Two cases:

Case one: the lettuce recall. His client, an indoor gardening tech brand, launched a content campaign around produce recalls three months before a foodborne illness outbreak. When the outbreak hit and recall searches spiked over a weekend, their March post on lettuce recalls and food safety was featured at the top of the AI Overview — alongside the FDA and CDC — ranked second in organic search behind the FDA, and pulled about 1,300 clicks from that news cycle.

No single property did that. It was the whole digital footprint: the author’s credentials, published papers, and prior work were all in the markup, on both the author page and the post. When an engine weighs citing a commercial site next to a government agency, it needs a reason not to discount you.

Case two: the shared company name. A client’s company name was shared with several others, and searching for the CEO returned nothing relevant. His team built a complete Person schema on the executive bio page, linked to the correct profile and company details — within days, the AI Overview identified the right person.

Validate First, Then Scale

Loren closed with pragmatic advice: schema pays off at scale — but scaling before validating just multiplies the same error across a thousand pages.

His rollout path:

1. Pick one page as a model, choosing the most specific entity type that honestly describes it
2. Assign a stable ID
3. Pull the facts from the visible page and your source of truth
4. Cross-check against trusted third-party sources
5. Add only values you can support
6. Validate before anything goes to development

Then segment: a location group, a product family, a set of author pages — units within striking distance move faster than a full-site rebuild, and they’re measurable.

Neo’s Take: What Sellers Should Actually Do

Framework done — here’s my practical advice.

First, treat your existing schema as a health check. Most independent-site schema comes from theme templates: incomplete fields, unstable IDs, mismatches with the real page. Run Google’s Rich Results Test and Schema.org validation, and fix four-source inconsistencies before adding anything new.

Second, brand-information consistency is your highest-ROI task. If AI answers about your brand are wrong, it’s usually not AI’s fault — it’s your NAP, hours, and sameAs links saying different things on different platforms. Spend one afternoon unifying them; it might be the cheapest AI SEO optimization you’ll ever do.

Third, ecommerce sellers: treat your product feed as a first-class citizen. If your products underperform in ChatGPT shopping or Google Shopping, check your Merchant Center feed first: GTIN, MPN, stable product ID, deep attributes (material, color, size, delivery time). OpenAI’s official feed spec requires item_id to be stable, title in sentence case, description in plain, factual language — details that decide whether AI shopping surfaces your product at all.

Fourth, don’t treat schema as a lifeline. It won’t compensate for thin content, invented credentials, or a neglected profile. Earn the authority first, express it clearly, then make it easy to verify — the order matters.

One sentence to close: in 2026, structured data has gone from “nice to have” to “machine-trust infrastructure.” Your site is going to be read by AI sooner or later — you might as well make sure it’s read correctly, and believed.

I’m Neo, and I write about independent site SEO. Hit me with your schema war stories or your own tricks in the comments.