Product Data Needs A Truth Set Before AI Search.

Google's July 7 Search Central update added clearer merchant-listing guidance for `Product.category` and sale-price effective dates. That sounds narrow. For a showroom, supplier, distributor, builder, or product-heavy remodeler, it is actually a useful operating reminder: product data is not an SEO decoration. It is a truth set.

The practical question is not, "Can we add more schema?" The better question is, "Do the visible page, product feed, internal catalog, sale calendar, and structured data agree?" If they do not, AI Search, Shopping surfaces, browser agents, sales staff, and customers all inherit the same ambiguity.

The update is about parity, not magic

Google's update says `Product.category` can use plain text for a merchant-defined custom product type and `CategoryCode` for a Google Product Category. Google also clarified that sale duration can use `validFrom`, `validThrough`, and `priceValidUntil`, with ISO 8601 dates and correct placement under an `Offer` or price specification.

That is not a new AI-ranking hack. It is a parity requirement. The page should show what the product is, the feed should classify it the same way, the sale should have real start and end dates, and the JSON-LD should describe only what the page and business can support.

Building-industry catalogs are messy by default

Cabinets, flooring, plumbing fixtures, tile, appliances, hardware, slabs, lighting, and installed packages rarely live in one clean system. The website may have marketing names. The showroom may use vendor SKUs. The ERP may use internal categories. Merchant Center may have another taxonomy. A salesperson may know the real constraint only because they have sold the line for years.

That is exactly why the truth set matters. Before AI can summarize, compare, recommend, quote, or route a shopper, the business needs a reviewed record for the product: visible category, Google Product Category where relevant, merchant product type, current price, sale window, availability, source system, owner, last review date, and known caveats.

AI Search still does not need special AI schema

Google's generative AI Search guidance still says structured data is not required for generative AI search and there is no special schema.org markup to add for AI Overviews or AI Mode. The durable work is still useful, original, crawlable content, clean technical signals, and structured data that matches visible page content.

So the Datum rule stays boring: do not add Product markup to a non-product page, do not invent sale dates, do not expose hidden claims, and do not let a vendor feed overwrite what the customer can actually buy. Use schema to clarify the real record, not to decorate a weak one.

Ads agents need an identity gate too

The same day, the Google Ads Developer Blog announced a pilot where pending Google Ads API Basic Access applicants can use Google Cloud project brand verification to speed review. Google's brand-verification documentation ties the developer token to a Cloud project, OAuth consent state, publishing state, and verified brand identity.

That belongs in the same governance bucket. If an AI or automation tool can inspect campaigns, upload conversions, change budgets, or diagnose product ads, the business needs to know which developer token, Cloud project, OAuth app, scopes, owner, approval path, and logs are involved. A loose agent with account access is not a workflow. It is an unreviewed side effect waiting to happen.

The product truth-set checklist

• Visible page facts: name, brand, category, price, sale dates, availability, and purchase or inquiry path.
• Feed facts: merchant product type, Google Product Category, SKU, price, sale-price window, and diagnostics.
• Schema facts: Product, Offer, category, price validity, image, and canonical URL that match the visible page.
• Source facts: source system, source URL or record, owner, last reviewed date, and approval status.
• Failure state: missing category, stale sale date, unavailable product, price mismatch, or feed/schema disagreement.
• Review path: who approves changes before the data reaches Search, Shopping, ads, or an AI workflow.

Where Datum would start

For a supplier or showroom, start with one high-value category, not the whole catalog. Pull ten representative product pages. Compare the visible page, feed, internal source record, and JSON-LD. Mark every mismatch. Decide which system wins when data conflicts. Then build the review queue and validation checks before asking AI to summarize or recommend anything.

For a builder, remodeler, or designer, the same pattern applies to packages and allowances. If a page or proposal says a package includes a product line, finish level, or price window, the source record should prove it. AI can help prepare the comparison, but the business owns the truth set.

That is the real AI Search lesson in a structured-data update: better answers come from better governed facts. Category, price, sale dates, feed parity, identity, permissions, logs, and review are not side quests. They are the operating system underneath trustworthy AI work.

Next step, if this note maps to a problem on your desk: Private Training — a private working session for your team ($1,500+).