How to Rank in Google AI Overview: 6 Content Rules Most Sites Ignore

Getting cited in an AI Overview is a large umbrella — technical setup, links, brand signals all sit under it. This post covers one part of it properly: content.

Because content is where most sites lose before anything else gets a chance to matter.

A quick illustration of what is possible first. Search "best permanent makeup salon in Grapevine, Texas" and you will find The-M-Effect, a client of mine, cited at the top of the AI Overview. Their competitors appear too, and Yelp sits above them — but they are recommended, their website appears, and they hold a spot in the local pack as well.

That is a small business appearing on three surfaces of one result page. Not from a trick. From following the rules below.

Rule 1: Unique means unique perspective

Google's guidance asks for unique content. Everyone reads that as "not copied" and moves on, because plagiarism has been a solved problem for years.

Read the elaboration and it says something else: unique means your point of view.

Take "how to make coffee at home." That is a generic angle. Every result says roughly the same thing.

Now take: "I tried 5 coffee methods for a month — here's the one I stuck with."

Same subject. Completely different value. The second shares an experience nobody else can reproduce. Google prefers real first-hand experience over general advice, consistently and by design, because the whole system runs on human-generated material and it is trying to identify which of it is genuine.

So before writing anything, answer one question: what do I know here that is mine?

Rule 2: Create non-commodity content

This overlaps with the first rule but adds something specific.

"7 tips for first-time home buyers" is a commodity. It applies to anyone, which means it belongs to no one. Thousands of pages already say it.

Non-commodity content is that same topic with your own experience and proof attached. Not "get an inspection" but "we negotiated the inspection waiver and saved $8,000 — here's the exact sequence."

That is specific. It contains no general advice at all. It is a case study inside an article, and there is far less of it published than there is listicle content — which is exactly why it earns attention.

The practical additions:

• Show proof. Screenshots, numbers, dates, before and after.
• Show the person. Photos, name, credentials. Long-form articles that read as written by a real human with a real history outperform anonymous ones.
• Use relevant images, tied to the actual scenario being described rather than stock photography.

You are trying to deliver the best experience available on that topic. Proof is what separates an experience from an opinion.

Rule 3: Vary the format

People used to write long unbroken paragraphs. You still can — nothing is wrong with prose. But prose alone is a wasted opportunity.

Within text content you have:

• Bullet points, where a list is the honest structure
• Tables, where the information is genuinely comparative
• Cards or blocks, where content breaks into discrete chunks
• Headings that mean something, so the shape of the page is readable at a glance

Two things happen when you do this. Search engines parse the content more easily. And readers can skim — which is what readers actually do. Nobody sits down for ninety minutes with your article. They scan headings, read first lines, and extract what they need.

Put a solid wall of text in front of someone and their first thought is "when does this end." That is a bounce, and it is entirely avoidable.

Rule 4: Add visual content, not just text

If we are discussing content, there is no reason to limit it to writing.

Images, infographics, video, diagrams — all of it belongs in the same page. It adds variety, it makes the page more appealing, and it gives search engines more to work with.

This matters more for AI Overviews than for classic organic results, because AI Overviews are recommendations. Recommendations go to quality. Offer more genuine quality across more formats, and you become eligible for more kinds of citation. It is a straightforward give and take.

Rule 5: Stop building a page per keyword

This one causes real, measurable damage, and it is everywhere — especially on affiliate sites.

The old pattern: pull keywords from a research tool, find ten that mean the same thing, build ten pages.

Look at these:

• best coffee maker
• top coffee machines
• coffee maker guide
• how to pick a coffee maker
• coffee machine tips

Those all mean the same thing. My name is Naveed. Naveed is my name. I am Naveed. Three phrasings, one fact.

Five pages for those five phrases is five weak pages. One page — "Best Coffee Makers: A Complete Guide" — covers all of it: the machines themselves, how to choose, the quick tips, the comparison. One topically authoritative long-form page that Google prefers and readers actually use.

Build the five separate pages and you get cannibalisation: your own pages competing against each other for the same query. Google understands they mean the same thing, so it picks one and credits only that. Sometimes worse — the pages look duplicative, relevance suffers, and rankings drop across the group.

You did five times the work for a worse outcome. Do less, and do it well. Doing more and confusing things is worse than doing less and doing it properly.

The rule: if several keywords share one intent, they get one page.

Rule 6: Hold AI content to real standards

Writing with AI is not a problem. Let me be precise about why.

Google's own position: AI-generated content can rank well. Google does not penalise content simply because AI produced it — and it does not automatically reward it either. Content is ranked on whether it is helpful, original, and demonstrates E-E-A-T: experience, expertise, authoritativeness, trustworthiness.

Note "original" again means original point of view, not original wording. If the perspective is genuinely yours, the wording follows naturally.

What does get punished is explicit: unedited, mass-produced AI content fails these standards and risks penalties under Google's spam and scaled content abuse policies.

Generating fifty or five hundred articles is trivially easy now. Publishing them without editing, without fact-checking, with topics overlapping each other and no purpose behind any of them — that is what gets penalised. Not the tool. The absence of care.

The workflow that works:

1. Let AI draft. That is a legitimate use of it.
2. Check it properly. Verify every factual claim. AI is confidently wrong often enough that this is not optional.
3. Add your own insights. Tell it your specific experience and have it woven in — the case study, the number, the thing that went wrong.
4. Then publish.

AI is impressive and it will not replace your personal statement. Your point of view has to be in there, because your point of view is the only part a competitor cannot generate too.

The short version

Unique perspective, not unique wording. Specific experience with proof, not commodity advice. Varied formats, text and visual. One page per intent, not per keyword. AI as a drafting tool, never as a publishing pipeline.

None of this is a shortcut, and that is deliberate. Practices that follow Google's guidelines survive the next update. Practices that exploit a gap do not.