Case Study 18.2 — The Stars That Weren't There: Spammy Markup and a Manual Action

A complementary angle. Case Study 18.1 showed a feature taken away from sites that did nothing wrong. This one shows the opposite failure: markup used to assert things that aren't true, and the penalty Google hands out for it. It teaches the chapter's hardest limit — that structured data works only when it is honest, and that faking it is one of the more reliably caught deceptions in SEO.

This is a labeled composite, built from Google's documented structured-data policies and its real, named manual-action type, "Spammy structured markup." It is not one named company; it is a pattern Google's guidelines describe explicitly and that its Search team has warned about for years. Every specific number below is illustrative. What is entirely real is the policy, the manual action, and the mechanism.

Background: the shortcut that looked free

Imagine a mid-size services business — call it a home-and-garden contractor — that has watched competitors' listings show gold review stars and wants the same. Its marketing agency proposes what sounds like a clever, cost-free win. Rather than do the slow work of earning reviews on the business's Google Business Profile, the agency will simply write the stars into the site's code:

  • add AggregateRating markup to the business's LocalBusiness schema, hard-coded to a flattering 4.9, citing testimonials the business selected itself;
  • add a block of FAQPage markup whose questions and answers do not appear anywhere on the visible page — invisible content, present only in the JSON-LD, written purely to look comprehensive to Google;
  • label several thin category pages as Product to try to pull in product-style rich results.

On a spreadsheet, it looks like pure upside: more rich-result eligibility, more SERP real estate, no waiting. The agency ships it, and for a short while a couple of the pages do show enhanced listings. Everyone is pleased.

This is exactly the trap the chapter warns about in §18.2 and §18.7: because JSON-LD is decoupled from the visible page, it is trivially easy to make the markup say things the page does not. Easy — and, in Google's guidelines, forbidden.

The issue: markup must match a visible, honest page

Google's structured-data guidelines are blunt on the points this business just violated. Three rules, all documented, all broken here:

  1. Don't mark up content invisible to users. Structured data must describe content that is actually on the page and visible to the person reading it. The invented FAQ block — present only in the code — breaks this directly.
  2. Don't mark up irrelevant or misleading content. The markup must be representative of the page. Thin category pages labeled as Product misrepresent what the pages are.
  3. Don't use fake or self-serving reviews. Review markup must reflect genuine reviews, and Google does not show self-serving business reviews at all (the 2019 rule from §18.3). A hard-coded 4.9 the business assigned itself is precisely the "self-serving" pattern the policy names.

Google enforces these with a specific, human-reviewed penalty. In Search Console's Manual Actions report, there is an entry type called "Spammy structured markup." When a reviewer determines that a site's markup violates the guidelines, Google can apply it.

FIGURE C18.3 — "What a spammy-markup manual action looks like"       [constructed teaching example]
  IN SEARCH CONSOLE   Manual Actions → "Spammy structured (data) markup" → Affects: parts of the site.
  WHAT GOOGLE SAYS    Markup on some pages may use techniques outside the structured-data guidelines
                      (e.g., markup for content invisible to users, or irrelevant/misleading markup).
  IMMEDIATE EFFECT    Rich results for the affected pages are removed; the pages can be demoted or their
                      enhancements suppressed until the issue is fixed and reviewed.
  THE FIX PATH        Correct the markup (remove invisible/fake/mismatched data so it matches the visible,
                      honest page), then file a reconsideration request and wait for a human re-review.
  (Illustrative rendering of a real, documented manual-action type — "Spammy structured markup.")

Here is the cruel efficiency of this particular deception: the violation is mechanically checkable. Google does not have to guess whether the FAQ content is invisible — it can compare the rendered page to the markup and see that the questions appear only in the JSON-LD. It does not have to judge whether the reviews are self-serving in some subtle sense — the pattern (a business's own AggregateRating on itself) is exactly what the policy already disallows. A mismatch between markup and page is one of the easiest things for an automated system or a reviewer to detect, which is why faking structured data is such a poor bet.

What it shows

This case is §18.7 made concrete, and it carries three transferable lessons.

  1. The decoupling that makes JSON-LD convenient makes it dangerous. The same property — markup separate from visible content — that lets a plugin inject clean breadcrumbs across a whole site also lets a careless or dishonest team assert things the page doesn't contain. Convenience and risk are the same feature seen from two sides.
  2. Faking a signal is self-defeating, not just against the rules. This is the book's first theme in its sharpest form. Structured data works because it is supposed to be true; the moment it asserts something false, it stops helping Google understand the page and becomes a manipulation attempt — and a mechanically detectable one. The "win" is temporary; the manual action is not free to undo.
  3. Recovery is real but costly. Unlike the Case Study 18.1 sites (which lost a feature through no fault and needed no cleanup), this business must repair the markup and petition to be trusted again — the reconsideration process (Chapter 26). It traded a slow, honest path (earn real reviews on the Business Profile) for a fast, dishonest one that ended up slower.

Outcome

In the version where someone catches it, the fix is unglamorous: strip the invented FAQ content out of the JSON-LD (or add it to the visible page if it's genuinely useful), remove the self-serving AggregateRating, relabel the category pages honestly, confirm every remaining piece of markup describes something a human can see, and file a reconsideration request explaining what was corrected. Rich-result eligibility returns after review. The business then does the thing it should have done first — build a real review presence on its Google Business Profile (Chapter 25) — which is the only legitimate route to the stars it wanted.

In the version where no one catches the manual action for months, the affected pages quietly lose their enhancements and some visibility, the agency blames "an algorithm update," and the business pays for a fix and a delay it never needed.

The lesson

Only mark up what is true and visible, and use the most specific accurate type — nothing more. The whole ethic of structured data compresses into that sentence. Every shortcut in this case — invisible FAQ text, self-assigned stars, mislabeled pages — was an attempt to make markup assert what the page could not honestly support, and each is specifically named in Google's guidelines and enforceable by the "Spammy structured markup" manual action. The honest path is slower and it is the only one that lasts: build a page worth describing, describe it truthfully, and let Google decide what to show.

Discussion questions

  1. The chapter calls a markup-versus-page mismatch "mechanically checkable." Explain what that means and why it makes faking structured data a worse bet than, say, other manipulations that are harder to detect.
  2. The agency's plan looked like "pure upside" on a spreadsheet. Where exactly did the risk live, and why is it the kind of risk that spreadsheets tend to hide?
  3. Compare the recovery in this case with the sites in Case Study 18.1. Why did one group need no cleanup while this business faces a reconsideration request? What does the contrast teach about honest versus dishonest routes to the same visible result (stars)?
  4. Reframe the business's goal (gold review stars in the SERP) as a legitimate project. What is the honest path to it, which chapter owns that work, and why is it slower?
  5. A junior colleague argues, "It's just a caption for machines — who's really harmed if we round our rating up a little?" Answer them using both the ethics guardrail and the self-defeating-signal argument.