Structured data (code added to your website that labels your business name, address, hours, and services in a format machines can read) does not guarantee that an AI assistant will quote you. Even perfectly valid schema can be ignored if it conflicts with other signals, triggers a content policy filter, or is simply incomplete. Here is how to diagnose the problem and fix it.

How 'machine-readable' still may not be 'quoteable'

There is an important difference between data that a crawler can parse and data that an AI assistant will confidently surface in a response. AI systems — including the large language models that power tools like ChatGPT, Perplexity, and Google's AI Overviews — apply a second layer of judgement before quoting any source. They ask, in effect: is this information clear, consistent, and trustworthy enough to repeat to a user?

Your LocalBusiness schema might be technically valid and still fail that test for several reasons:

  • Inconsistency across sources. If your schema says your address is "14 Main Street, Galway" but your Google Business Profile (GBP — the free listing that appears in Google Maps and the local panel on search results) says "14 Main St, Galway City", an AI may treat both as uncertain and quote neither.
  • Thin surrounding content. Schema embedded in a page with very little readable text gives the AI no corroborating context. The structured data looks like an island.
  • Stale data. Hours or services that appear to contradict recent reviews or third-party listings (such as your Tripadvisor or TheFork profile) reduce confidence.

The practical implication: schema is a signal, not a guarantee. It needs to be consistent with everything else the AI has seen about your business.

Content policies: what triggers extraction avoidance

AI assistants are trained with content policies that make them cautious about quoting certain types of claims. These policies exist to prevent the spread of misleading information, but they can catch legitimate local businesses in the crossfire.

Common triggers include:

  • Superlative or unverifiable claims. Phrases like "the best plumber in Dublin" or "Ireland's leading beauty clinic" in your schema description or page copy can cause an AI to treat the surrounding content as promotional rather than factual.
  • Medical or health-adjacent language without qualification. If you run a physiotherapy clinic or a pharmacy and your schema description includes outcome language ("we cure", "guaranteed relief"), moderation systems may suppress extraction entirely.
  • Missing business category context. A LocalBusiness schema type is broad. If you do not specify a more precise type — MedicalClinic, HairSalon, Plumber — the AI has less confidence about what you actually do and may skip the quote.
  • No verifiable third-party presence. AI systems typically weight businesses more highly when they appear on recognised platforms. For Irish businesses, that means listings on Golden Pages, Booksy (for salons and wellness), Treatwell, or Tripadvisor, depending on your sector.

None of this is documented in a public rulebook. It is inferred from observed behaviour. The safest approach is to write schema descriptions the way a careful journalist would write a factual sentence: specific, neutral, and verifiable.

Field completeness requirements and fallback behaviour

When an AI cannot find a required piece of information in your schema, it does not always fail gracefully. Sometimes it falls back to whatever it can find — which might be an outdated directory listing or a review snippet that misrepresents you.

The fields that matter most for local business extraction are:

Field Why it matters
name Must match GBP and all directory listings exactly
address (with streetAddress, addressLocality, postalCode) Incomplete addresses reduce location confidence
telephone Should match the number on your GBP and website header
openingHoursSpecification Vague or missing hours cause AI to omit operational details
description Needs to be factual, specific, and free of superlatives
url Should point to the canonical version of your homepage

If any of these are missing, the AI may still quote you — but it will fill gaps from elsewhere, and that elsewhere may be wrong. Completing every field is not optional if you want accurate extraction.

For more on how to structure the pages that sit behind your schema, see Service pages that AI can quote: structure, FAQs, and proof.

How to reduce ambiguity between NAP and schema fields

NAP stands for Name, Address, Phone — the three core identifiers that local SEO and AI systems use to confirm a business is real and locate it precisely. Ambiguity between your NAP data and your schema fields is one of the most common and most fixable causes of extraction failure.

Here is how to audit and tighten it:

  1. Write down your canonical NAP. Decide on one exact version of your business name, address format, and phone number. Include the county if you are outside a major city (e.g. "Ennis, Co. Clare" rather than just "Ennis").
  2. Check your GBP. Log into your Google Business Profile and confirm the name, address, and phone match your canonical version character for character.
  3. Check your schema. Use Google's Rich Results Test or Schema.org's validator to read back what your schema actually says. Compare it to your canonical NAP.
  4. Check your directories. Look at your Golden Pages, Tripadvisor, Booksy, or Treatwell listings. Update any that differ.
  5. Check your website header and footer. The phone number and address displayed on your site should match the schema exactly.

This process sounds tedious because it is. But inconsistency here is the single most common reason a business with valid schema still does not get quoted. For guidance on how your About page can reinforce these signals, see About page that supports credibility and local recommendations.

Also worth reading: Structured data that AI and moderating systems actually accept goes deeper on the technical side of what passes moderation.

A practical test plan: change → wait → re-query

Once you have made changes to your schema or NAP consistency, you need a way to know whether they worked. Here is a simple, repeatable process:

Step 1 — Make one change at a time. If you fix your address format and rewrite your description simultaneously, you will not know which change made the difference. Fix one thing, then test.

Step 2 — Wait. AI systems do not update in real time. Crawl cycles, index updates, and model retraining all introduce lag. Typically, allow at least two to four weeks before drawing conclusions.

Step 3 — Re-query consistently. Ask the same question in the same AI tool each time. For example: "Is there a [your business type] in [your town]?" or "What are the opening hours of [your business name]?" Write down the exact response.

Step 4 — Test across tools. ChatGPT, Perplexity, and Google's AI Overviews each have different extraction behaviours. A fix that works in one may not yet show in another.

Step 5 — Check what is being quoted. If the AI quotes you but gets a detail wrong (old phone number, incorrect hours), trace where that wrong data is coming from. It is almost always a directory listing that has not been updated.

If you want a starting point before you begin this process, the free check at suggestedby.ai shows whether AI assistants are currently naming your business in response to relevant local queries — which tells you whether you have a baseline to improve from or a gap to close.

For further reading on the content side of this problem, FAQ pages for Irish local services: examples AI can answer directly covers how well-structured FAQ content can support extraction even when schema alone is not enough.