By Samuel Anan

Structured data is often introduced to business owners with a promise that goes beyond what the technology can establish. Add the right markup, the pitch suggests, and search or AI systems will finally understand and recommend the brand.

I would separate the useful technical work from that promise. Markup can describe information in a defined structure. It does not turn an unsupported business claim into a fact, and it does not guarantee selection in a search or AI answer.

The best starting point is the visible page and the business information it contains.

Markup describes content; it does not replace it

Google's introduction to structured data explains how structured information can help its systems understand page content and support eligible search features.

That is not a reason to hide important facts only in code. A customer should still be able to find and understand the offer, location and relevant details on the page itself.

If the visible service description is confusing, adding a structured version of the same confusion is not the complete solution. If the markup says something the page does not support, the problem becomes more serious.

Google's structured-data policies require relevant, accurate representation of the content. The technical implementation should follow the public facts, not create a second, more promotional version for machines.

Use the type that fits the actual content

A page may describe an organisation, a local business, a product, an article or another supported kind of information. The choice should reflect the page and the relevant implementation requirements.

Do not select a type merely because a marketing article claims it produces better visibility. Check the official documentation for the feature being pursued and the fields that genuinely apply.

An illustrative service page should not pretend to be a product with invented ratings simply to resemble a richer search result. Nor should a general testimonial be converted into a review claim without considering the applicable rules and evidence.

When specialist implementation is needed, involve someone who can inspect the live page, code and testing results. A generated snippet of markup is not evidence that the deployment is correct.

Keep facts consistent across the visible and technical layers

Review names, URLs, prices, availability and other material details. A change in the business can leave structured data stale even after the visible text has been edited.

This is one reason I begin with a business-facts audit. The approved record gives the technical review a standard beyond whatever happens to be in the code.

Check the public result after deployment. A content management system or plugin may generate fields automatically, and those defaults may not match the current offer.

The aim is not identical wording everywhere. It is agreement about the facts the page represents.

Validation is not the same as eligibility or appearance

A technical test can identify certain errors or show that markup is readable. Passing a test does not guarantee that a search feature will appear for the page.

Eligibility can depend on additional requirements, and actual presentation is a separate outcome. Keep those stages distinct in both the implementation report and the sales explanation.

A responsible delivery note might say that the relevant markup was implemented and tested, with the observed test result. It should not say that a particular search appearance has been secured unless that appearance was actually observed, and even then it should not be described as permanent.

Google's structured-data introduction explains the distinction between providing the information and guaranteeing its use in a feature. That limitation should not disappear when the work is repackaged as a service.

There is no special Google AI markup requirement

Google's AI features guidance states that no special structured data is needed specifically for its AI features. Established search practices and useful, accessible content remain relevant.

This does not mean technical quality is irrelevant. It means the claim should be narrower than “add this schema to get recommended by AI”.

Other services have their own access and retrieval arrangements. OpenAI's crawler documentation, for example, distinguishes crawler purposes and controls. Those details concern access; they do not create a universal markup recipe for recommendations across every AI system.

Be cautious when a service combines several different systems under one confident guarantee without showing the basis for it.

What a useful technical report should contain

I would expect the page URL, the purpose of the markup, the source of the facts, the implementation date and the testing method. The report should list unresolved issues and identify who maintains the information when the business changes.

If a visible result is being reported, keep the date and observation separate from the implementation claim. A search appearance observed once is not the same as a contractual outcome that the business controls.

The AI observation method provides the same discipline for generated answers: preserve the exact response and context rather than summarising everything as improved authority.

Put structured data in its proper place

It belongs within a broader effort to make a business's public information accurate, accessible and useful. It should be maintained alongside the content, not treated as a hidden substitute for it.

That is the role I would want technical structure to play in Sparlab: support clear business profiles and content while keeping the limits of any visibility claim visible.

The technology is useful enough without exaggeration. A good implementation describes the right facts correctly and can be inspected. The business still needs a useful offer, reliable public evidence and a customer journey that makes sense after the page is found.