Google's AI content guidance now points to the rater guidelines, and calls fake authors deception

Google updated its guidance on using generative AI content on October 1. Its documentation changelog says the page now carries material from the Search Quality Rater Guidelines, to match what Google presents at its developer events. The rule itself is unchanged: using AI to generate many pages without adding value for users may violate the spam policy on scaled content abuse.

What the page now says

• Read two rater sections. Google points to section 4.6.5 of the rater guidelines, on scaled content abuse, and 4.6.6, on main content made with little effort, originality or added value. It adds that the guidelines are not a guide to ranking first and that raters' scores don't directly affect rankings.
• Fact-check everything the model wrote, including the parts readers rarely see. Language models predict likely words rather than retrieve facts, the page says, so all AI-generated content needs a manual check before publishing. That covers title elements, meta descriptions, structured data and image alt text as well as the body.
• Tell readers how the content was made. For automated content, Google suggests explaining how automation was used and adding image metadata.
• Merchant Center rules for shops. AI-generated product images must carry the IPTC DigitalSourceType value TrainedAlgorithmicMedia, and AI-generated titles and descriptions must be submitted separately and labelled as AI-generated.

Made-up authors

Google's guide to creating helpful content encourages accurate bylines and author pages, then draws a line: fabricating creator profiles, "such as by using AI-generated headshots, made-up names, or false credentials", is deception. Google calls any form of deception a signal of a low-quality page, to readers and to its automated systems alike.

What to do

If a site publishes with AI help, check the parts that get skipped: titles, descriptions, schema and alt text produced in bulk. If an author box shows a person who doesn't exist, replace it with a real name or an honest team byline. And if pages were produced at volume, reading sections 4.6.5 and 4.6.6 against a sample of them is a quicker audit than any tool.