Barry's Tea scored 33.55 out of 100 for structured data

Fifty pages, zero markup

Scan #218, standard depth, target https://barrystea.ie. Of the 109 findings, 4 passed, 46 warned, 50 failed and 9 were informational. The dominant finding was the same one, repeated 49 times: "No schema.org structured data (JSON-LD, Microdata, or RDFa) found on this page." Across the entire crawl the audit found three markup blocks in total. The format split was JSON-LD 0, Microdata 3, RDFa 0. You can open the audit yourself and read every finding.

Zero JSON-LD is the number that matters. JSON-LD is the format the rest of the web has settled on: per the Web Data Commons corpus, 70% of websites that annotate structured data use JSON-LD, against 46% for Microdata and 3% for RDFa. It is also the format Google recommends, in its words "the easiest solution for website owners to implement and maintain at scale", and the one every generator, plugin and CMS emits by default. A site with no JSON-LD at all in 2026 is not a site with a configuration problem. It is a site that was built before machine-readable description became part of publishing, and never revisited.

The only markup in the crawl was written by X

One page in the crawl set did carry structured data, and it produced an odd result: the audit's coverage category scored 100 and reported 15 distinct schema.org types, including ProfilePage, Organization, ImageObject, WebSite, SearchAction, Person, VideoObject, SocialMediaPosting and InteractionCounter. None of that belongs to Barry's. The page was https://x.com/barrysteatweets, the brand's linked X profile, which the crawler followed from an outbound link on the site. The markup on it was authored by X, describing an X profile, using X's own templates.

Read plainly: every page on barrystea.ie itself carried zero schema.org markup, and the only machine-readable description of Barry's Tea anywhere in the crawl was written by a social network. That is not a credit to the brand. It is the clearest possible picture of who currently gets to define the entity. The site coverage finding puts it at "1 of 50 pages (2.0%) carry schema.org structured data" - and that one page is not the site.

"Your site, barrystea.ie, has significant gaps in its structured data implementation, with 50 pages completely lacking schema.org markup."
- the audit's own generated headline, scan #218

You cannot have a completeness problem with nothing to complete

A schema audit normally spends most of its effort on quality: are the required properties present, do entities contradict each other, is the data stale, is it shaped the way an AI system can consume it. Four of those modules returned nothing at all here. ai_readiness reported "No JSON-LD entity nodes detected - AI readiness checks skipped". entity_graph reported "No JSON-LD entity nodes were detected". conflicts and freshness both reported "No JSON-LD entities detected". There were no entities to check.

That is why the presence category scored 0.66 / 100 and og_twitter_xref scored 0.0. The site is not badly marked up. It is unmarked. Everything downstream of that - rich results, entity relationships, freshness signals - describes things that do not yet exist, so there are no gaps to close, only work to start.

The site is not silent to machines in every respect. Open Graph and social metadata were found on 45 of the 50 pages, and microformats on 10. The audit is explicit about what that is worth: "This is not schema.org structured data and is not counted toward the schema score." Open Graph tells a social platform how to render a link preview. It does not tell a machine that Barry's Gold Blend is a product, what it costs, or who sells it. Even the social layer is incomplete: 45 pages warned that OG tags were present but no twitter:card meta tag was, so X falls back to a generic link preview. The audit also recorded no hreflang alternates anywhere on the site.

The irony is that the facts are already written down. The about page states, in prose, that Barry's holds 38% of Irish tea value sales, that the market is worth an estimated €85 million at retail, and that the company employs 65 people. A person reads that and knows exactly what kind of organisation this is. A machine has to infer all of it, because there is no Organization block saying any of it out loud.

What a tea brand forfeits with no Product markup

Start with the part that is documented and unambiguous, because it is the part that pays. Google splits Product markup into two eligibility classes: product snippets, for pages where the product cannot be bought directly, which allow more review information; and merchant listings, for pages where the customer can actually buy, which allow more detail such as sizing, shipping and return policy. Merchant listings are eligible for the shopping knowledge panel, Google Images, popular product results and product snippets, and require an Offer object because, in Google's words, "the merchant has to be the seller of the product in order to be eligible for merchant listing experiences."

The bar for the basic version is low. For a product snippet, Google's documentation requires name, plus at least one of review, aggregateRating or offers. A price or a rating. That is the entry fee, and barrystea.ie is not paying it on any page. If you go further and want the markup to feed Merchant Center item updates, Google asks for price, priceCurrency, availability and condition on the Offer, and is firm that the values must match what the customer is shown: mismatched prices are a policy violation, not a quality nudge. If you add star ratings, Review objects need author and reviewRating.ratingValue, plus itemReviewed unless the review is nested inside the item it reviews, and AggregateRating needs ratingValue plus a ratingCount or reviewCount. One trap: reviews you control about yourself do not qualify at Organization level, so a wall of testimonials marked up as Organization reviews earns nothing. Product-level reviews are not subject to that restriction.

Rich results are the measurable half of this. Google's own documentation cites Rotten Tomatoes measuring a 25% higher click-through rate on pages with structured data, the Food Network seeing a 35% increase in visits, and Nestlé reporting an 82% higher click-through rate on pages that show as rich results. Those are vendor-supplied case studies, and they are about classic rich results rather than anything AI-shaped. They are still the clearest published numbers on what the markup is worth.

Now the part where we will not overclaim. Structured data is not a confirmed lever for AI Overviews or AI Mode. Google's own documentation says so directly: "You don't need to create new machine readable files, AI text files, or markup to appear in these features. There's also no special schema.org structured data that you need to add." Its AI optimization guide repeats that structured data is not required for generative AI search, while still recommending it for rich results. Anyone selling you schema as an AI ranking hack is going past the evidence. The one first-party signal pointing the other way comes from Microsoft: at SMX Munich, Fabrice Canel of Bing told the audience that schema helps Bing and Copilot's LLMs, and pushed publishers to use IndexNow for freshness, because "Gen AIs value fresh content in particular, partly as a reference check of their LLM training data."

So the honest framing is not "add schema and rank in AI answers". It is this: structured data is the only place on a web page where you state, unambiguously and without a machine having to infer it from prose, that this thing is a Product, it is called Barry's Gold Blend, it costs this much, and it is sold by this organisation. Every other route to that fact is a guess made by a parser. Barry's already does well in AI answers on brand recognition alone, which reflects the marketing done since James J. Barry founded the company in Cork in 1901, not to its website. That is exactly the kind of advantage that erodes quietly when the machine-readable layer is doing the describing. The job of schema has shifted since Google retired FAQ rich results: less about collecting SERP decorations, more about being the definitive machine-readable statement of what your entities are.

And Barry's is not unusual. Across the Web Data Commons October 2024 extraction, 51.25% of crawled pages carried structured data, which leaves close to half that do not, and at domain level only 44.12% of domains had any at all. That is up from the 5.7% the first WDC release recorded in 2010, but the tail is long and it is full of well-run brands with marketing sites that predate the machine-readable web.

The order to fix this in

This is a half-day of work for most brand sites, and it is boring work, which is why it gets deferred. Do it in this order.

1. Put Organization and WebSite JSON-LD in the site template. One block, every page, done once. Name, logo, url, sameAs pointing at the real social profiles. This is the block that stops X being the most authoritative machine-readable description of your brand.

2. Add Product to every product page. name, image, description, brand, and an Offer with price, priceCurrency and availability. If you sell directly, that Offer is what makes merchant listing experiences possible at all. If you only link out to retailers, you are in product-snippet territory instead, and you need a review or aggregateRating to qualify.

3. Mark up the things that already exist as entities. A brand site usually has more than products: BreadcrumbList on every page, Recipe if you publish any, Article on the blog, LocalBusiness or Organization contact points. These are descriptions of content you have already written.

4. Fix the social layer while you are in there. Adding twitter:card to the 45 pages that have Open Graph but no card type is a template edit. It will not move a schema score, and it is not structured data, but it stops every shared link rendering as a bare URL.

5. Re-crawl and check for contradictions. The moment you have entities, the audit modules that returned nothing here start doing real work: conflicting names, prices that disagree with the page, stale dates, orphaned entities. That is the point at which a score becomes a quality measure rather than a presence check.

Audit your structured data

Questions this raises