Taprave
All insights
AI search

Schema markup for ChatGPT and AI search

Structured data is useful description, not a ranking switch. Here is where it helps and where agencies oversell it.

26 August 2026 9 min readBy Emmanuel OlorunsholaUpdated 26 August 2026

Schema is one of the easiest AI-search services to make sound mysterious. It is code using a shared vocabulary. Its job is to identify things and relationships more explicitly than page design alone.

That is useful, especially for entity consistency. It still cannot rescue thin content, manufacture authority or make an unverified claim true.

Start with Organisation or LocalBusiness

Use Organisation for the company identity and an appropriate LocalBusiness subtype when the business has a genuine local establishment. Include the canonical URL, name, logo, contact details and sameAs links you control or can verify.

Give the organisation a stable @id and reference it as publisher or provider on other relevant markup. Keep every field aligned with visible content.

Describe services and content accurately

Service markup can identify the provider, service area and offer described on a genuine service page. Article markup should include the headline, author, publisher and accurate publication and modification dates.

BreadcrumbList helps describe site hierarchy. Product, Event and other types should only be used when the page and business genuinely fit their requirements.

FAQ markup needs restraint

Questions and answers must be visible to users and match the markup. FAQ schema is not a licence to insert hidden keywords or claims.

Google has restricted FAQ rich-result visibility to particular authoritative sites, so do not sell FAQ markup as a guaranteed search enhancement. The visible answers can still help users and machines understand the page.

Validate the implementation

Use Schema.org’s validator for vocabulary and syntax. Use Google’s Rich Results Test for Google-supported experiences. Inspect the rendered page because valid code can still describe the wrong content.

Check for duplicate entities, conflicting addresses, invented ratings, future dates and types that do not match the page. Validation passing means the syntax is acceptable, not that the strategy is sound.

What schema cannot do

It cannot guarantee indexing, rankings, citations or recommendations. It cannot replace customer evidence or third-party corroboration. It cannot make invisible or misleading content acceptable.

The right order is visible evidence first, clear page structure second, accurate markup third, then ongoing validation when facts change.

The takeaway

Use schema to connect accurate visible facts into a coherent entity graph. Do not treat it as a hidden ranking layer. If the claim would be weak without the markup, the schema will not fix it.

Questions people ask

Use the most appropriate LocalBusiness subtype when the business has a real local establishment; otherwise Organisation may be more accurate.

No public requirement says a page must have schema to be cited. Accurate structured data can reduce ambiguity but does not guarantee selection.

No. Add it only where genuine questions and answers are visibly present and appropriate.

Only when the page and review data follow the relevant guidelines. Invented or self-serving aggregate ratings create compliance risk.

Use Schema.org Validator for general markup and Google’s Rich Results Test for supported Google search features, then inspect the visible page.

Sources and further reading

Want us to think about your business like this?

Free 20-minute audit. Straight talk, no proposal-deck theatre.