Schema, AI Overviews and what changed for local clients

Which structured data assistants actually read, which of it moves anything, and what to stop selling as an AI service.

Structured data has been quietly useful for a decade and loudly oversold for about two years. For local clients specifically, some of it now does meaningful work in AI-generated answers, and some of it has never done anything at all. Knowing the difference keeps you from selling a service you cannot substantiate.

What actually gets read

For a local business, the load-bearing types are LocalBusiness, Service, FAQPage and — where relevant — Product and Review. These give an assistant unambiguous answers to the questions people actually ask: what this business does, where, when it is open, and what it charges.

The mistake is treating schema as a ranking lever. It is a disambiguation lever. It does not make a weak page strong; it makes a strong page legible.

Consistency matters more than coverage

A client with correct LocalBusiness markup, a correct Google Business Profile and consistent name, address and phone data across listings is in a far better position than one with twelve schema types and three different phone numbers on the web. Conflicting signals are worse than sparse ones, because they force a machine to guess.

What to stop selling

"AI optimization" as a standalone line item, priced as though it were a separate discipline, is difficult to defend. Schema work, entity consistency and clear on-page answers are ordinary technical SEO. They happen to be what assistants read, which is a good reason to do them well — not a good reason to invoice them twice.

What to measure instead

Track whether the client appears in AI-generated answers for their core service-plus-location queries, and track it manually, monthly. It is imperfect, it is not a dashboard metric, and it is currently more honest than anything automated.