Discussion
  • Entity SEO

Fewer properties marked up accurately, or maximum coverage with some inferred values? Have you seen consequences either way?

Published 1 reply

Discussion starter: an entity SEO question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.

Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.

1 reply

  • admin
    Admin
    Original poster

    Fewer properties, marked up accurately, and Google says so in almost those words. The more useful distinction is not few versus many but sourced versus guessed. An inferred value that is derived reliably from a system of record, and matches what the page shows, is fine. A value that is a default, a pattern-fill or a model’s guess is a liability, and the risk grows with the size of the catalogue because nobody checks it after launch.

    The practical rule: include every required property for the features you want, then add recommended properties only as fast as you can source them. Missing a recommended property costs you some presentation. A wrong one can cost you eligibility, Merchant Center approvals and, in the worst case, a manual action.

    What Google’s documentation actually says

    Google’s introduction to structured data acknowledges that more recommended properties can make enhanced display more likely, then adds: “it is more important to supply fewer but complete and accurate recommended properties rather than trying to provide every possible recommended property with less complete, badly-formed, or inaccurate data.” Required properties are a different matter: “You must include all the required properties for an object to be eligible for appearance in Google Search with enhanced display.”

    The general structured data guidelines add the quality side: markup must represent the page accurately, “Don’t mark up content that is not visible to readers of the page”, and misleading markup can be treated as spam. The same page is explicit that “Google does not guarantee that your structured data will show up in search results, even if your page is marked up correctly.” A structured data manual action removes rich result eligibility; Google says it does not affect how the page ranks in regular web search.

    Important distinction:

    Structured data eligibility is not ranking impact. Most of the downside of bad markup lands on rich results, merchant listings and Merchant Center, not on blue-link rankings. That makes the damage easy to miss in a rankings report and easy to see in Search Console’s enhancement reports and Merchant Center diagnostics.

    Three kinds of “inferred” value, with very different risk

    Type Examples Typical failure Risk
    Derived from a system of record Brand from the PIM brand field; breadcrumb from the category tree; condition “new” where the business only sells new stock Pipeline bugs, usually caught because the same data drives the visible page Low
    Defaulted InStock hard-coded; priceValidUntil set to year end on every product; one shipping rate or returns window applied site-wide Correct for most items, silently wrong for a subset (oversized goods, clearance, hygiene exclusions) Medium to high
    Guessed or generated GTINs filled by pattern; sameAs matched to a Wikidata entry by name; backfilled publish dates; ratings with no on-page reviews; attributes extracted by an AI model Plausible-looking values that are false, with no visible counterpart to catch the error High

    The first category is not really “inferred” in the sense that worries people. It is the same fact the page already shows, expressed in a second format. The trouble starts when the markup says something the page does not, or says it with more confidence than the business can back.

    The consequences you can actually observe

    These are the documented, visible outcomes. None of them require speculation about hidden quality scores.

    Consequence What triggers it Where you see it Scope
    Item not eligible for a rich result Missing a required property, such as a price on a Product Offer for merchant listings Search Console enhancement reports as errors That item
    Fewer annotations Missing recommended properties such as shipping or returns details Warnings in enhancement reports; plainer listings Presentation only
    Feed values overwritten or disapproved Landing page data disagreeing with the feed Merchant Center diagnostics Shopping surfaces and free listings
    Manual action for structured data Markup that misrepresents the page or marks up hidden content Manual actions report Rich result eligibility, per Google
    A false public claim Wrong price, delivery or returns values surfaced to shoppers Customer complaints, returns disputes Compliance and trust, not only SEO

    The Merchant Center row deserves attention because it turns markup errors into feed errors. Google’s automatic item updates use structured data and other landing page data to update price, availability and condition. Its own example: if the feed says $4 and the product page says $3, the product is updated to $3. When mismatches are frequent, or the data cannot be extracted reliably, products may be disapproved instead, and excessive mismatches can stop the updates altogether. A defaulted InStock value on a page that visibly says “sold out” is exactly the kind of conflict that feeds this system bad information.

    The last row is a business point, not a Google one. Price, shipping and returns values in markup can appear in search as statements to shoppers. A returns window that is wrong for some categories is a consumer-facing claim, and in Australia the Australian Consumer Law’s rules on misleading representations do not care whether the claim came from a template default.

    Consequences that are plausible but thinly evidenced

    A wrong sameAs link can associate your organisation or author with the wrong entity. Knowledge panel confusion from this shows up in practitioner reports, but there is no controlled public evidence of how much weight Google gives sameAs against its other entity signals. For AI search, Google’s guide to generative AI features says no special schema is needed. How other AI systems use JSON-LD is poorly documented. The safe assumption is that anything published in markup can be repeated back as fact, so a guessed value is a guessed value wherever it travels.

    What the “maximum coverage” approach looks like at scale

    Hypothetical example:

    A retailer with 40,000 SKUs ships a schema template with availability defaulting to InStock, a single shipping rate, and a 30-day return policy on every product. At any time, about 2,000 SKUs are out of stock, 3,000 are bulky items with a different delivery charge, and 1,500 are clearance items with no change-of-mind returns. Up to one product in six now carries at least one false claim. Nobody notices in the rich results report, because the markup validates. The signs show up elsewhere: Merchant Center availability mismatches on out-of-stock items, listings showing free-looking delivery on items that charge for it, and customer service tickets about returns.

    The same retailer with a “fewer, accurate” template would output availability from the live stock field, omit shipping details for bulky items until the rate table is in the product data, and scope the return policy by category. Coverage is lower on day one. Every value is true, and coverage grows as data work lands.

    GTINs are the other common temptation. If, in the same hypothetical catalogue, a large share of products has no GTIN in the PIM, the honest options are to omit gtin for those items or to source real GTINs from suppliers. Generating values that look right is worse than leaving the field out, because Merchant Center validates GTINs, and a wrong identifier can tie your listing to someone else’s product.

    A decision rule for each property

    1. Is it required for a feature you want? If yes, it must be present and true. If you cannot source it, the item is not ready for that feature.
    2. Does a Google feature or a consumer you care about use it? Properties with no documented consumer add maintenance without a visible return. Several rich result types were retired in 2025 and 2026 (FAQ rich results among them), so check the current documentation before investing.
    3. Is there a system of record? A PIM field, an inventory service, a policy table. If the only source is a default or a model, treat the value as unverified.
    4. Is it visible on the page? If the markup says something the page does not, fix the page or drop the property.
    5. Does it update when the truth changes? Price and availability must update with the page. Values rendered once at publish time go stale.
    6. What does it cost if wrong? A wrong material value is minor; a wrong price, returns window or rating is a claim to a customer.

    How to find out which situation you are in

    • Crawl with custom extraction to compare JSON-LD price and availability against the visible values, and report the mismatch rate per template.
    • Review Merchant Center diagnostics for price and availability mismatches, and whether automatic updates are changing your feed.
    • Read Search Console’s merchant listings and product snippets reports for errors (eligibility) versus warnings (optional properties).
    • Check the Manual actions report, even if you expect it to be empty.
    • Manually open every sameAs target on Organization and Person markup; there are usually few enough to check by hand.
    • List every property whose value comes from a default rather than a data field. That list is your risk register.

    To test whether extra coverage is worth the effort, add one well-sourced recommended property (shipping details are a good candidate) to half of comparable categories and hold the rest back. Compare merchant listing impressions and CTR in Search Console’s search appearance data over six to eight weeks. If the treated half does not outperform, the property is a nice-to-have and data effort belongs elsewhere.

    The position I would take

    If this were my catalogue, I would ship required properties everywhere they can be true, recommended properties only where a field in a real system backs them, and nothing that exists only because a validator suggested it. Then I would measure the mismatch rate between markup and page every release. A catalogue with full coverage and a few per cent false values is worse than one with partial coverage and none, because the first carries its errors into Merchant Center, onto search results and in front of customers, while the second only misses a few enhancements it can add later.

    Need help with this on your own site?

    Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.

    See SEO services

Add a reply

Replies are open to approved members. Log in or apply to join.