AI search will not reward your clever hack. It will reward your proof.

Google has updated its guidance on how to show up in generative AI search features.

Bad news for the AEO grift economy: there is still no magic trick.

No special file that makes Google love you. No secret markdown copy of your website. No “LLM schema stack” that lets a thin page outrank a useful one. No neat little hack that turns beige AI content into a trusted source.

Google’s line is much more boring, which usually means it is worth listening to.

Its generative AI features in Search still run on core Search systems. Crawling. Indexing. Ranking. Quality. Retrieval. Query fan-out. The same old machinery, with a different answer surface bolted on top.

That does not mean nothing has changed. It means the useful work is not where the hacks are.

Most businesses asking about AEO do not actually have an AEO problem. They have a proof problem.

Their service pages say the same thing as every competitor. Their case studies are vague. Their product pages are missing details. Their blog posts are safe summaries of things everyone already knows. Their YouTube videos, if they exist at all, have lazy titles and no useful transcript. Their claims sound impressive right up until someone asks, “Can you show me?”

Now put that business into AI search.

Why would an answer engine cite it?

If your page could have been written by any competitor with the same prompt, it is not a source. It is wallpaper.

This is where the AI search conversation gets more interesting than another “SEO is dead” panic post.

Search is changing, yes. Google talks about RAG and query fan-out. In plain English, that means the system is not only looking at the exact words someone typed. It can break the question into related questions, pull supporting pages from across those related searches, and use that material to build an answer.

Ahrefs’ recent work on AI Overview citations points the same way. A lot of the cited sources are not simply the exact-query top ten wearing a new hat. YouTube is turning up as a source too. That matters, because it means old-school rank tracking is not enough on its own anymore.

But the answer is not to spray more AI content across more long-tail keywords.

Please do not do that. The internet is already full enough of synthetic porridge.

The better answer is to become a source worth using across the questions your buyer actually asks. That is a different kind of marketing work, and it starts with boring evidence.

What do you actually sell? Who is it for? What problem does it solve? What does it cost? What is included, and what is not? What have you done before, and what changed because of it? What does the customer need to know before buying? What objections come up on sales calls? What mistakes do people make in this category? What proof can you safely show?

Most websites dodge half of that.

They hide behind polish. Nice gradient, weak substance. Big promise, no receipts. Five sections of “tailored solutions” and not one specific example a human can trust or a machine can extract.

AI search does not fix that. It exposes it.

A useful AEO baseline should not start with “write 50 AI articles.” It should start with a harder audit. Which pages are actually crawlable and indexable? Which claims have evidence behind them? Which offers are specific enough to understand? Which questions does the buyer ask before they trust you, and which of your pages answer those questions with something only you can say? Which videos, transcripts, images, reviews, product feeds, case studies, and third-party mentions support the same story? What does Search Console show, if the new generative AI reports are available to you yet? What are AI systems citing now, and what are they getting wrong?

That is not as glamorous as selling an “AI search domination framework.”

Good. Glamour is usually where the nonsense hides.

For service businesses, this means turning real delivery work into proof pages. Not fake case studies. Not inflated screenshots. Real before-and-after notes where the client has approved it, the claim is accurate, and the lesson is useful even if the reader never buys.

For ecommerce, it means product truth. Clean feeds, proper attributes, availability, images, compatibility, FAQs, delivery details, return policies, reviews, comparison data. A buying agent will not care how pretty the hero section is if the product data is a mess.

For B2B, it means source depth. Technical explainers, implementation notes, real customer language, expert commentary, video clips with transcripts, and pages that answer the awkward buying questions your competitors avoid.

For founder-led brands, it means having an actual point of view.

That bit matters more than people want to admit. Google’s guidance keeps talking about unique, useful content. Not unique in the “we shuffled the words” sense. Unique as in this came from your experience, your customers, your data, your process, your scars. That is hard to fake at scale, which is exactly why it is worth something.

The cheap version of AEO will be the same as the cheap version of SEO was. A pile of pages, a pile of promises, a pile of tools, and a dashboard that looks impressive until someone asks whether any of it created trust or revenue.

The better version is smaller and sharper. Map the buyer’s questions. Find the sources AI systems already lean on. Check what is wrong, missing, or weak. Build pages and media that add something real. Structure them properly. Measure what can be measured. Mark the rest honestly as inferred or unknown. Keep updating the proof as the business changes.

That is not a one-off content job. It is an operating layer.

And this is where agencies need to be careful. If AI makes content cheaper, an agency does not become more valuable by selling more content. It becomes valuable by building the system that decides what should exist, proves it, publishes it, checks it, updates it, and learns from what AI search actually does with it.

The output is not the moat. The receipts are.

AI search is not asking businesses to become louder. It is asking them to become more useful, more specific, and easier to verify.

Annoyingly, that means doing the work.

Probably a good thing.

---

Jason Sibley is the founder of Cleo, a post-agency marketing and AI company. JasonVsTheNoise is where he writes about what is actually happening with AI, marketing, and how businesses should be thinking about both.