Key takeaways
- Google says there is no special schema.org markup a business must add to appear in its AI Overviews or AI Mode.
- Structured data labels information on a page; it does not prove the information is accurate or guarantee a search feature.
- A schema test passing is not evidence that AI visibility has improved. Measure the two separately.
- Before paying for new markup, ask for an inventory of what already exists and which pages have a verified need.
- Google's guidance applies to its own AI features, not automatically to every other AI assistant.
A local business owner hearing about AI-search optimisation could be forgiven for assuming new code is the missing piece. Suppliers sometimes present schema markup as a ticket into AI Overviews, AI Mode or chatbot answers. Before agreeing to pay for that work, it helps to know exactly what the evidence says schema markup can and cannot do.
No special schema is needed for Google AI search
Google's AI-features guidance says no special schema.org structured data is required for AI Overviews or AI Mode. A page needs to be indexed and eligible to appear in Google Search with a snippet to be eligible as a supporting link. Eligibility is not a guarantee of inclusion.
Ordinary SEO practice still matters: make important information available as readable text, and ensure structured data matches what visitors can see. These are sensible checks, not a formula that makes a page appear in an AI answer.
What structured data actually does
Structured data is a standardised way of labelling information on a page: a business name, opening hours or an article headline, for example. The shared vocabulary is maintained at schema.org. Labels can help a search engine classify information already visible to people; they are not a stamp of approval confirming that the information is true.
Accuracy remains the business's responsibility. A label for opening hours does not check that the hours are correct. Google recommends fewer, complete and accurate properties over lots of incomplete or invented ones.
AI visibility is not one Google feature
Google AI Overviews, Google AI Mode, ChatGPT and Perplexity are separate systems with different sources and rules. Google's guidance on schema applies to Google's own features. It should not be presented as a universal law for every AI assistant.
LocalBusiness markup and supported features
Google documents LocalBusiness structured data for describing details such as opening hours in supported Search functionality. Meeting the requirements can help a page qualify for a feature, but Google does not promise that the feature will appear for any particular search.
Audit before you pay for anything
The useful first step is usually to find out what already exists and whether it is working properly. Ask your website provider for an inventory showing each page, the type of markup on it and the system that creates it: a theme, SEO plugin or custom code.
- Flag pages with multiple markup blocks and ask whether they describe different things or duplicate one another.
- Ask for a plain description of a known issue, rather than accepting a blanket recommendation to remove everything.
- Agree which URL will change, why it needs attention and how the finished work will be checked.
Accuracy comes before labelling
Before coding work begins, prepare a factual checklist: real business name, current contact details, actual opening hours and the correct destination for important pages. Where a business has more than one location, keep those facts distinct. Google's structured-data guidelines require information that accurately represents the visible page.
Do not invent information merely to fill a field. An uncertain price, qualification or rating should be clarified or omitted, not guessed at to complete a template.
Proof the work was done properly
Once work is agreed, there should be a clear handover: which URLs changed, which markup types were involved, what system produces them and who will maintain them. Google recommends the Rich Results Test and URL Inspection to check implementation and access issues. Those tests help investigate technical work; they do not test whether an AI assistant will recommend the business.
Schema supports features, not promises
A completed technical correction and an observed change in AI visibility are different outcomes. If an AI mention appears after an update, record the platform, question, answer and source cited. One mention after one update does not prove the update caused it.
Start by requesting an inventory of existing markup, then resolve one verified issue at a time. Confirm the facts, inspect the published result and agree who will keep the information accurate after future changes.