Google changed how it extracts JSON-LD on August 21, 2026, and the practical recommendation is simple: audit the rendered structured data on important website templates and fix any double-escaped values. This is a technical parsing change, not proof that schema became a stronger ranking factor. Still, it matters because Google will no longer compensate for certain sloppy implementations.
For established plumbing, HVAC, electrical, roofing, restoration, and other home-service businesses, the risk usually sits inside a WordPress plugin, theme, template, or custom field. The website may look completely normal to a customer while the machine-readable version of a business name, service, author, product, or location is wrong.
Google’s JSON-LD parsing change in plain English
Google Search Central said its JSON-LD extraction now applies only one pass of HTML unescaping so the parser better follows JSON and related standards. Double-escaped entities such as & and ✔ will no longer be repeatedly decoded to the intended character.
The official announcement recommends standard JSON escapes or Unicode hexadecimal escapes, such as \u0026 for an ampersand. Google Search Central’s August 21 announcement describes the change as an effort to bring its parser in line with established standards. RFC 8259, the JSON specification cited by Google’s Gary Illyes, defines how characters must be escaped inside JSON strings.
A simple example shows the difference. If a template is supposed to output “Heating & Air” but emits a value that has been HTML-encoded twice, one decoding pass may leave the literal entity text instead of the ampersand. The page can still display the business name correctly while Google’s extracted property contains the wrong value.
This distinction matters: the problem may be an incorrect extracted value rather than a dramatic, page-wide failure. Passing a general syntax check should not be the end of the review. The extracted names, descriptions, URLs, prices, and other properties also need to be read and compared with the visible page.
Why schema started to feel less important in the AI era
Modern search systems are better at inferring meaning from visible content, links, images, business profiles, citations, reviews, and surrounding context. Google also says websites do not need special schema.org markup or new machine-readable files to appear in AI Overviews or AI Mode. Its current guidance for AI features and websites tells publishers to keep following foundational search practices instead.
At the same time, Google has reduced or removed several structured-data-powered search features. It phased out some lesser-used rich result types in 2025 and said those removals did not affect page rankings. In May 2026, Google deprecated the FAQ rich result, then removed its documentation in June. That history reasonably caused many people in search marketing to treat schema as less strategically important.
I think that conclusion goes too far. AI has reduced how much a system must depend on schema to understand a webpage. It has not eliminated the value of a clean, explicit machine-readable statement.
AI can infer meaning, but structured data explicitly declares it
The useful 2026 model is that visible content provides evidence a machine can interpret, while structured data provides explicit assertions about entities, attributes, and relationships. Google describes structured data as a standardized format that supplies “explicit clues” about a page’s meaning in its introduction to structured data.
Consider a regional plumbing company with two legitimate offices. A search system may infer the organization, locations, phone numbers, service areas, authorship, and offered services from the page. Valid Organization, LocalBusiness, Service, Person, Article, and BreadcrumbList markup can state those relationships directly. The structured data does not replace the page, the Google Business Profiles, the reviews, or local authority. It supplements them.
That is also why structured data must match what customers can see. Google’s general structured data guidelines prohibit marking up information that is not represented on the visible page. A hidden block stuffed with extra services, cities, ratings, or claims is not a shortcut. It creates a second, conflicting version of the business.
Revved’s earlier article on using sameAs and mentions in JSON-LD explains two ways entity relationships can be expressed. The new parsing rule adds an important condition: those relationships only help when the output is accurate and standards-compliant.
If there is uncertainty about how Google reads a contractor website, a technical visibility audit can identify broken schema, local-search gaps, and conflicting business information before a site-wide template change is made.
Google’s change does not make schema a ranking factor
There is no public evidence that the August 21 parser change increased the ranking weight of structured data. Google framed it as an extraction and standards change. Fixing double escaping may restore the intended information or eligibility for a supported rich result, but it does not guarantee higher rankings, an AI citation, or inclusion in an AI answer.
Structured-data validity and rich-result eligibility are separate questions. Schema.org supports a broad vocabulary, while Google supports specific rich result experiences with their own required properties and policies. Markup can be technically valid without producing a special search appearance. A formerly supported rich result can also disappear without making every underlying entity declaration useless.
The same restraint should apply to AI search claims. Emerging research, including a March 2026 paper on structured linked data as a retrieval memory layer, explores how structured facts can support retrieval systems. That is useful direction, not proof that adding schema to a local business page directly causes commercial AI systems to cite it.
For a broader view of signals beyond markup, Revved’s guide to AI search visibility for local businesses covers content, reviews, geographic relevance, and entity consistency.
What home-service businesses should audit now
A focused template audit is more useful than checking hundreds of individual URLs. Most contractor websites generate structured data through a small number of repeatable templates, so one defect can affect every service page, location page, article, or review-enabled page.
- Inspect the rendered output. Open representative pages from each important template and locate every
application/ld+jsonblock. Do not assume the plugin settings reflect the final output. - Look for repeated encoding. Search for patterns such as
&and&#, then check names and descriptions containing ampersands, apostrophes, quotation marks, symbols, or imported HTML. - Read the extracted values. Use Google’s Rich Results Test for supported Google features and the Schema.org validator for general markup. Google recommends both tools on its structured data testing page.
- Compare markup with visible content. Confirm business names, addresses, telephone numbers, hours, service areas, author names, dates, breadcrumbs, ratings, and service descriptions agree with the page.
- Fix the generator, not each page. If a theme, plugin, custom field, feed, or template escapes a value twice, correct that source and retest every affected template type.
- Monitor after deployment. Recheck representative URLs after caches clear and Google recrawls them. Treat warnings and changes in rich result reporting as diagnostic signals, not ranking guarantees.
Priority should go to markup connected to real business identity and high-value pages: Organization, LocalBusiness, Service, Article, Person, BreadcrumbList, Product, and legitimate Review or AggregateRating implementations. A contractor does not need every possible schema type. The business needs accurate markup for facts that are visible, current, and operationally true.
Structured data is not dead; forgiving implementation is fading
Google’s stricter JSON-LD parsing does not reverse the broader shift toward systems that can interpret webpages without special markup. It clarifies the standard publishers should follow when they choose to make explicit machine-readable declarations: valid JSON, correct values, visible supporting content, and no dependence on a forgiving parser.
The actionable takeaway is to test one representative URL from every important website template this week, inspect the extracted property values, and repair the shared generator if anything is double-escaped or inconsistent. Established trades companies that need the website, local visibility, tracking, and lead flow handled as one system can review Revved’s connected implementation program.
