Google did not ban the machine. It banned the fake person.

On October 1, 2026, Google added one paragraph to its guidance on helpful content, and most of the marketing world read it in about four seconds: Google is cracking down on fake authors, so go add real bylines and bios. That reading is not wrong. Read what Google called deceptive, then notice what it did not mention. Writing with a machine is not there.

What the paragraph says, and what I am adding

The page is Google's documentation titled "Creating helpful, reliable, people-first content" (developers.google.com/search/docs/fundamentals/creating-helpful-content). Its section "Who (created the content)" has long told site owners to make authorship clear. Search Engine Journal reported on October 2, 2026 that the version dated October 1 carries a new paragraph in that section, and that a saved copy from June 8 does not have it (searchenginejournal.com/google-fake-author-warning-site-owner-guidance/591806/).

Anyone who reads the paragraph carefully will see that it is about fake people and not about AI writing. That observation is not mine and I am not claiming it. What I am adding is two things. One is the cost comparison: what a disclosed machine costs you against what an exposed fake costs you, with a documented case and a published study. The other is the case Google's wording does not reach: a real name on pages that person never opened.

The new wording tells site owners to "avoid using deceptive authorship information." Its examples of fabricating a creator profile are "AI-generated headshots, made-up names, or false credentials." It goes on: "Any form of deception makes a page untrustworthy to both users and our automated quality systems," and it calls that deception "a signal of a low-quality page."

A face that belongs to nobody. A name that belongs to nobody. A qualification that nobody earned. Every one of them is a lie about a person. None of them is a statement about how the words got onto the page.

The line Google did not draw

If Google wanted to say "do not publish what a machine wrote," this was the paragraph to say it in. It did not. The next section of the same page, "How (the content was created)," asks a different question: "Is the use of automation, including AI-generation, self-evident to visitors through disclosures or in other ways?"

That is not a ban. That is a request to say so. Google's separate guidance on generative AI content (developers.google.com/search/docs/fundamentals/using-gen-ai-content) goes the same direction: "Sharing information about how a piece of content was created can help give your readers more context." It also says: "It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing."

So the two halves sit side by side on Google's own pages. Using a machine: tell people. Inventing the human: deception. A lot of site owners have had those two backward. They hid the machine because they assumed it was the embarrassing part, and they covered it with a smiling expert who has twenty years of experience and has existed since Tuesday (haha).

This was never a gray area. Search Engine Journal notes that section 4.5.3 of Google's Search Quality Rater Guidelines already treats fake creator profiles and false expertise claims as deception, and that raters are told to give such pages the Lowest rating. The rule existed. It was written for raters, and now it is written for site owners.

Why the fake expert is the expensive choice

A page that says a machine drafted it and a named person reviewed it has nothing left to find out. The worst thing a reader, a competitor or a search engine can discover about it is already printed on it.

A page with an invented expert is carrying a debt, and there is a documented case of the debt coming due. On November 27, 2023, Futurism published "Sports Illustrated Published Articles by Fake, AI-Generated Writers" (futurism.com/sports-illustrated-ai-generated-writers). The reporter traced the headshot of a bylined writer named Drew Ortiz to a website that sells AI-generated portraits. A second byline, Sora Tanaka, had a headshot from the same site.

Look at what it cost. The Arena Group, the magazine's publisher at the time, said the articles were licensed from a third party, AdVon Commerce, and that AdVon had writers use pen names. It removed the content and ended the partnership. The words "fake, AI-generated writers" are now attached to a title that spent decades building its name on reporting. In December 2023 the Arena Group's board terminated its CEO. The company did not tie that decision to the story, and I will not either. The part that is not in dispute is enough: pulled pages, a severed vendor, and a headline that does not expire.

Notice the size of the thing that started it. Two product review bylines. The damage did not stay on two pages, because a reader who catches one fake person has no reason to believe the real ones.

A made-up expert was never a credibility play. It was an admission that nobody at the company wanted their own name on the work.

The argument against my own argument

There are three fair objections, and I cannot wave any of them away.

First: can Google catch it? I do not know, and neither does anyone writing confident posts about this. Search Engine Journal pointed out that the paragraph does not say how automated systems would identify a fabricated profile. Moving the rule from the rater document to the site owner page is evidence of intent. It is not evidence of enforcement. The Sports Illustrated bylines were caught by a reporter with an image search, not by an algorithm. Google is silent on the how. Convenient (haha).

Second: does admitting a machine was involved cost you readers? Yes, and there is independent research on it. Oliver Schilke and Martin Reimann published "The transparency dilemma: How AI disclosure erodes trust" in Organizational Behavior and Human Decision Processes in 2025. Across 13 experiments, people who disclosed their AI use were trusted less than people who did not. That cuts against me and I am not going to soften it. The same authors, summarizing the work for the University of Arizona, report that the steepest drop in trust came when AI use was hidden and uncovered later. So the cost of saying so is real and it is paid up front, which is why I call disclosure the cheaper option and not the free one. You pay a small known price now and avoid a large unknown one later.

Third: a real name can be just as fake as a made-up one. If a company pastes its founder's byline on five hundred pages the founder has never opened, the name is real and the accountability is fiction. That passes the letter of Google's paragraph. It fails the point of it. This is the objection that changes my argument, so it gets its own section.

A real name is not the fix

If the lesson of Google's paragraph is "swap the fake author for a real one," most companies will do exactly that, and nothing will be different. The headshot will be genuine. The bio will be true. And the person in the photo will have no idea what the page says. That is still a lie about a person. It is a lie about what that person did.

So the test is not "is there a byline," and it is not "is the byline a real human." It is whether a specific person would pick up the phone about this page.

Can a reader find out who that person is, somewhere other than the page that is asking for their trust? Did that person read it before it went out? If a claim on it turns out to be wrong, is there someone whose job and reputation it is to fix it? If the answers are yes, the drafting tool matters about as much as which word processor was used. If the answers are no, a genuine bio does not help.

The Sports Illustrated case shows what no looks like. When the bylines were exposed, the publisher's first answer was the name of a vendor. Whatever the contract said, the reader heard that nobody inside the building had read those pages. The fake headshots were the evidence. The absence of an owner was the problem.

The standard applies here, so here is how this page was made. A machine drafted it from the sources named above. A second pass scored the draft against a list of faults, and it was revised, more than once. The Google quotations were checked against the live pages on October 7, 2026, when the helpful content page showed a last updated date of October 5, 2026. I read it and approved it before it published, which is the only way anything goes out under my name on this site. Every claim on it is mine to answer for, including the ones you disagree with.

A made-up expert was never a credibility play. It was an admission that nobody at the company wanted their own name on the work.

So if you have an invented author on your site, do not start by writing a bio for a real one. And if you already have a real name on everything, do not assume you are clear. Pick ten pages carrying that name and ask the person whether they have read them. If they have, say so on the page, and say how it was made. If they have not, you have the same problem as the company with the AI headshot. Yours just has a real face on it.