Does publishing more get you found? Google's own answer

The advice is everywhere: publish more, publish often, get found. It is close enough to true that it survives repetition, and vague enough that nobody checks it. Google has actually written down what happens, and the documentation is more useful than the slogan.

The mechanism is fan-out, not volume

When someone asks an AI-powered search a question, Google runs what it calls query fan-out: "a set of concurrent, related queries generated by the model to request more information." Their own example — ask "how to fix a lawn that's full of weeds" and the machine also goes looking for "best herbicides for lawns", "remove weeds without chemicals", and "how to prevent weeds in lawn" [1].

You did not target those three phrases. One page that genuinely covers the subject can be retrieved by all of them. Google adds that fan-out lets it show "a wider and more diverse set of helpful links … than with a classic web search" [2].

That is the honest engine behind the advice. Depth creates surface area. A thorough page has many doors into it.

One page answering a question well is retrievable by a dozen questions you never wrote down.

Where "more" stops working

The same Google guide addresses the tempting next move — a page for every variation someone might search — and rules it out:

"Doing so primarily to manipulate rankings or generative AI responses in Google Search violates Google's scaled content abuse spam policy. This is also an ineffective long-term strategy, as a high quantity of pages doesn't make a website higher quality or more relevant to users" [1].

The policy itself defines scaled content abuse as "many pages generated for the primary purpose of manipulating search rankings and not helping users," and the first example it lists is using generative AI "to generate many pages without adding value for users" [3].

Google's test for what counts instead is unusually concrete. Commodity content is something like "7 Tips for First-Time Homebuyers" — common knowledge that "could originate from anyone." Non-commodity content is their example "Why We Waived the Inspection & Saved Money: A Look Inside the Sewer Line" — a first-hand account nobody else can write [1].

The difference is not length, keywords or cadence. It is whether anything happened to you.

Five things you can stop paying for

Google's guide carries a section headed "Mythbusting generative AI search: what you don't need to do." Worth reading in full before buying any AEO or GEO package, because it retires five popular ones [1]:

An llms.txt file. "You don't need to create new machine readable files, AI text files, markup, or Markdown … Google Search itself doesn't use them." Keeping one "will neither harm nor help."

Chunking your content. "There's no requirement to break your content into tiny pieces for AI to better understand it."

Rewriting in an AI-friendly voice. "You don't need to write in a specific way just for generative AI search."

Buying mentions. "Seeking inauthentic 'mentions' across the web isn't as helpful as it might seem."

Schema added for AI. "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add" — though it still earns rich results in ordinary search.

On the acronyms themselves Google is blunt: "optimizing for generative AI search is optimizing for the search experience, and thus still SEO" [1].

The fourth reader

Here is the part the SEO conversation has not caught up with. Your pages now have a reader that does not read: an agent, sent by someone to do a job. Google's own guide for them opens by saying that sites built for humans, "with complex hover-states, shifting layouts, and fluid motion," are "functionally broken for agents" [4].

Agents see a site three ways: a screenshot, which is "slow and expensive" and works best as a backup; the raw HTML, for structure and relationships; and the accessibility tree, described as "a high-fidelity map that ignores the visual 'noise' of CSS to focus on pure utility" [4].

What they ask for is specific, and none of it is content work. Prefer real <button> and <a> elements over styled divs. Keep layouts stable, so an add-to-cart button is not somewhere new on every category. Avoid transparent overlays that hide what is underneath. Set cursor: pointer, "a strong signal for actionability." Link every label to its input. Keep interactive elements larger than eight square pixels, or visual analysis discards them [4].

For the first three readers, useful content means readable. For the fourth it means operable. A business can publish beautifully every week and still be unusable to the reader growing fastest — which is the same workflow-versus-agent distinction our types guide draws, arriving from the other direction.

The reassuring part, in Google's words: "Everything we suggest to make a site 'agent-ready' also makes sites better for humans" [4].

So what does consistency actually buy you?

Not ranking, directly. It buys coverage — more subjects on which you have said something specific, and therefore more fan-out questions that can land on you. Ten pages drawn from real work beat a hundred assembled from common knowledge, and Google's summary of the whole thing is a question you can ask yourself without any tooling: is this content that my visitors would find satisfying? [5]

Which leaves a practical problem. Most businesses have the first-hand material — the job that went wrong, the thing customers always ask — sitting in someone's head, because writing it up and posting it in four places is a chore that loses to actual work. Kadres, the agent we build, connects to Instagram, TikTok, Facebook and YouTube and produces the posts from a brand profile you teach it — captions, carousels and video, in your voice, on a schedule.

What it cannot do is have the experience for you. No tool can, and by Google's own definition that experience is the only part that was ever scarce. A publishing agent lowers the cost of getting it out of your head. It does not put anything in there.

AiAgentNook editorial · last verified 17 Sep 2026 · platform and search documentation changes often; corrections are noted here, not silently edited.