AI Search Models Ignore CSS Rendering. Hidden Text Injects Fake Entity Data.

AI search crawlers bypass visual-rendering checks and process the raw DOM text layer directly, reveals Sebastian.

A field test by the agency Klaneo exploited this blind spot by embedding a fabricated "New York office" claim into a homepage using white text on a white background.

The page isolated the injection with zero external citations, backlinks, or local NAP signals.

While Google’s classic indexer detects and penalizes matching font-and-background color values, LLM parsers swallow the hidden HTML strings.

The generative models accepted the unverified text and cited the fake location as factual entity data.