Explainer · Tech made clear

Structured data: why the green light is no longer enough

Listen to the articleRead by your browser
Person seen from behind at a laptop, reading a report on screen, the way you read the extracted values in Google's structured data test

On 21 August 2026, Google quietly stopped repairing an error your website may have been making for years. Nothing appeared on screen, no alert went out, and the validation test can still show green. That is exactly what makes this change dangerous: it does not break your code, it changes the value Google records. This article explains what happened, who is concerned, and how to check your site in two minutes, without installing anything.

The invisible code that speaks to Google on your behalf

Beneath the visible part of your website lives a block of code called JSON-LD, the "structured data". Its job: telling Google, unambiguously, "this is the company name, this is a price, this is a customer rating". It is what allows your stars, opening hours or prices to appear directly in the search results. You probably never wrote it yourself: your CMS, your theme or a plugin generates it for you, on every page, automatically. Keep that word, automatically: it is what turns one small mistake into a series.

What changed on 21 August

Some tools encode special characters twice. The "&" of "Home & Garden" becomes & instead of &, a checkmark becomes a half-resolved code. For years, Google applied several decoding passes and quietly caught this kind of slip. The announcement of 21 August, clarified by Gary Illyes with reference to the JSON standard, ends the safety net: one single decoding pass, and Google records whatever remains, as is.

The consequence is surprising in its discretion. In most cases the block stays technically valid: Search Console reports nothing, the official test can show its green light. What changes is the stored value. "Home & Garden" instead of "Home & Garden", in the name Google associates with your business. Silently. This is neither a penalty nor a ranking update: it is a change in how Google reads, and it will not warn anyone.

The three traps, from most common to most treacherous

Trap 1, the doubled ampersand. The "&" character encoded twice by a theme or a plugin. Result: the recorded value contains code instead of the character. It is the most widespread case, because "&" lives in thousands of company names, product names and street names.

Trap 2, the double-encoded entity. A symbol, a checkmark, an arrow, inserted as an HTML code and then re-encoded along the way. It stays half-resolved in the final value. The same symbol pasted directly as a Unicode character passes without a problem: the double encoding is the trap, not the symbol.

Trap 3, the deceptive green. The most treacherous of the three, because it disarms your only checking reflex. The validation test verifies that the block is well formed; it does not verify that the values are the ones you intended. A block can therefore be perfectly valid and tell an error. The conclusion fits in one sentence: check the values, not the light.

The check, two minutes on the clock

Reading what Google really records

  1. Open the Rich Results Test, the official tool, free, no sign-up.
  2. Test your home page, then your most visited page.
  3. Do not stop at the light. Unfold the detected block and read the extracted values, character by character: the name, the price, the rating.
  4. If you still see & or an unresolved apostrophe code where you expected "&" or an apostrophe, that residual value is what Google now reads.
  5. The error almost always comes from the theme or plugin that generates the block: that is where it gets fixed, once, for every page at the same time.

Who is concerned, without the catastrophism

This does not mean half the web is broken, and this article will not tell you that. But the arithmetic deserves to be laid out: when the error comes from a theme or a plugin, it repeats on every page built with it. And roughly one website in two runs on WordPress, Shopify, Wix or Squarespace, with WordPress alone accounting for close to 41% of all websites (W3Techs, August 2026). That is exactly the terrain where JSON-LD is generated in series, with no human proofreading. The figure does not say how many sites are affected, nobody knows; it says where this type of error gets manufactured at scale, and why the two-minute check is worth it even if your site "has always worked fine".

This change is one of three that Google delivered in ten August days; the other two, a spam update and a click redirect, are handled with the same discipline of measurement. The full picture and the method are in our guide to measuring Google changes; and if your curves have moved recently, start with the five checks of the anti-panic protocol, a wrongly recorded value and a ranking drop are not treated the same way.

Frequently asked questions

How do I know whether my site uses structured data?

Test any page in the official Rich Results Test: it lists the detected blocks. If your site runs on WordPress, Shopify, Wix or Squarespace, it almost certainly generates some, through the theme or an SEO plugin, even if nobody ever configured it consciously.

The test is green: is my site in the clear?

Not yet. The green light confirms that the block is well formed, not that the values are right. Since 21 August 2026, a badly encoded value can sit inside a perfectly valid block. You need to unfold the detected block and read the extracted values: the name, the price, the rating, character by character.

I found a wrong value: am I risking a penalty?

No, this change is neither a penalty nor a ranking update. The risk is elsewhere: Google records and may display an erroneous value, a disfigured company name, an unreadable price. The fix happens at the source, in the theme or plugin that generates the block, once for all pages.

I can't make sense of the test report: what should I do?

That is normal, it is written for technicians. Two options: send a screenshot of the report to whoever looks after your site and ask them to check the extracted values, or write to us. Reading this kind of report is part of our job; yours is elsewhere.

More questions? Many answers are already gathered on our FAQ page.

Would you rather we looked for you?

We test your key pages, read the extracted values and tell you in writing what is right, what is wrong and where the source gets fixed. No jargon, screenshots included.

Information current as of 27 August 2026. Main sources: Google Search Central announcement of 21 August 2026 on structured data processing, clarified by Gary Illyes with reference to RFC 8259 · Rich Results Test, Google's official tool · W3Techs, usage statistics of content management systems, August 2026. Cover photo: Katrin Bolovtsova, Pexels.

Portrait of Maryna Skachek, founder of MariCleo Studio

Written byMaryna Skachek holds a PhD in biology, is a qualified e-commerce project manager and the founder of MariCleo Studio. She writes about digital strategy, content and the professional uses of AI.

WhatsApp