How to Rank in AI Overviews: Do the Opposite of Google’s Guide

To rank in Google’s AI Overviews, build genuinely distinct content for each real customer segment, earn genuine brand mentions on sources AI already cites like Reddit, and follow Microsoft’s advice to add schema markup and question-and-answer blocks.

Google’s new AI search guide warns against the first two, and I read those warnings as hints about what is quietly working.

I have been doing SEO and AEO for 19 years, and Google’s public guidelines and the way Google actually ranks pages have always been two different things. One tactic this guide warns against drives AI traffic that converts 31% higher than non-branded organic search.

Below is what Google published, the two warnings that reveal what works, the AI search myths you can safely ignore, and the step-by-step guide Microsoft published.

Key takeaways

• Google’s AI search guide is vague on purpose. When an AI summary appears, people click a regular link only 8% of the time, versus 15% without one.
• Read the warnings as hints. Google threatens penalties for tactics that work, like distinct content for each customer segment.
• AI traffic is worth chasing. Shoppers from AI sources like ChatGPT convert 31% higher than visitors from non-branded organic search.
• Reddit drives AI citations, making up roughly 40% of citations across major AI engines and about 21% inside Google’s AI Overviews.
• Skip llms.txt and content chunking. Google says neither helps its AI features.
• Microsoft’s AI search guide gives the real steps: schema markup and modular question-and-answer blocks.

What Google’s AI Overviews guide actually says

Google’s guide to generative AI features mostly repeats the advice every Google guide has given for years: write helpful, people-first content for humans instead of algorithms.

On May 15, Google published an official document called Optimizing Your Website for Generative AI Features in Google Search. It is the first time Google has put in writing how to show up inside AI Overviews and AI Mode, the AI answers that now sit at the very top of the results page.

The guide opens the way every Google guide has opened for years. Write helpful, people-first content for humans instead of algorithms, and good things will happen to you.

That sounds reasonable until you see what it implies: that Google’s systems can read your page, understand it the way a human would, and decide whether it is genuinely helpful. A machine cannot taste your content and know it is good, because it reads signals, patterns, links, and behavior.

How Google’s link building rules contradicted what actually ranked

For years Google warned that building links to your own site was a fast track to a penalty, while backlinks stayed one of the most powerful ways to rank. Its rules called artificial links, link schemes, and paid links all grounds for punishment.

Meanwhile, friends of mine who do SEO for a living, like Jeff Oxford at 180 Marketing, were building backlinks for their clients the entire time. It was one of the most powerful things you could do to rank back then.

Google said one thing in public while building links kept working in private. That pattern has never changed.

Why AI Overviews send fewer clicks to online stores

AI Overviews send fewer clicks to stores because shoppers get the answer on Google’s page and rarely click through. I saw this in my own store over the last year, when customers started showing up already knowing my handkerchiefs and linens by name.

Brand searches for Bumblebee Linens shot up dramatically. Shoppers were finding us inside AI answers and then Googling the brand, and the AI Overviews themselves rarely sent clicks anywhere past my homepage.

The data says this is happening to everyone.

Pew Research found that people click a regular link only 8% of the time when an AI summary appears, versus 15% when there is no summary.

Only 1% of users ever click a source link inside the AI summary itself.

The AI traffic that does come through is exploding. Traffic to US retail sites from generative AI sources jumped 1,200% in a single year, according to Adobe.

Why distinct content for each customer segment works in AI search

Distinct content for each customer segment works in AI search because every guide carries information that only applies to that buyer, and Google’s threat to penalize the tactic is the proof that it works.

The guide says you should avoid creating separate versions of your content for every different way people might search, such as reworking the same guide for a different customer or use case.

According to Google, doing this to influence rankings violates its scaled content abuse spam policy. If the tactic did nothing, Google would have no reason to build a punishment for it, and the threat tells you it works.

There is a wrong way to hear this. Spinning up hundreds of near-identical articles, or ten pages targeting the same keyword with the same information rephrased, really is spam, and it will get you buried.

The version that works serves different buyers properly. In my store, the wedding market, the corporate gifting market, and the memorial and funeral market all buy handkerchiefs for completely different reasons.

So I write a genuinely distinct guide for each of those buyers, packed with information that applies only to that customer. It is the same product category, with twelve real audiences getting twelve articles that each say something unique.

The opportunity keeps growing. AI Overviews appear on roughly 16% of all searches and have hit nearly 25% during peak periods, according to Semrush.

Why Reddit brand mentions still get stores cited in AI answers

Reddit brand mentions still get stores cited because Reddit is the single most-cited source feeding AI answers.

It makes up roughly 40% of citations across the major AI engines and about 21% inside Google’s own AI Overviews.

Google warns against seeking out inauthentic mentions and says its systems are excellent at spotting spam and will ignore low-quality mentions you work hard to get.

That claim does not hold up for well-placed mentions on sources Google already trusts, even though spraying your brand across garbage sites really is a waste of time.

Reddit is also being gamed harder than almost anything else online. I have friends who quietly place brand mentions into exactly the right threads for a living, and it works.

I recently had Danny Kirk, founder of the agency ReddiReach, which does this for hundreds of brands, on my podcast. He confirmed what I suspected: a huge amount of what you read on Reddit today was placed there on purpose, and it flows straight into the AI answers your customers see.

Put your effort into earning genuine, useful mentions on the sources Google already trusts and cites, starting with Reddit.

The payoff is real money. Shoppers arriving from AI sources like ChatGPT convert 31% higher than visitors from non-branded organic search.

AI SEO tactics Google says you can skip

Google says you can skip four popular AI SEO tactics: llms.txt files, markdown copies of your pages, content chunking, and special schema markup for AI.

This is the one part of the guide where it did everyone a favor. A flood of so-called experts spent the last year selling complicated tricks for ranking in AI, and Google quietly torched many of them.

• llms.txt files. Google says this special text file does nothing for its AI features.
• Markdown copies of your pages. You can skip publishing a separate markdown version of your site.
• Content chunking. Google’s systems already pull the right passage out of a normal page, so chopping content into tiny pieces is unnecessary.
• Special schema for AI. Google says no special markup or structured data is required to appear in its AI features.

Google also now says AI-generated content is fine as long as it is helpful. That is a massive reversal from the panic it spent two years creating, and I believe it happened because Google cannot reliably tell the difference anymore.

For a store owner, this is real relief. You can stop losing sleep over re-engineering your whole site with machine-readable files.

Why Google keeps its AI ranking advice vague

Google keeps its AI ranking advice vague because specific advice would help you pull traffic out of its AI answers and onto your store. Every click its AI answer keeps is a click that stays on Google, so those low click-through numbers are the business model.

Google also cannot reliably tell which content was written by AI. It spent two years insisting it could spot AI content and reward real human expertise, and its systems cannot do it at the level it claims.

So when Google cleaned up the results, it could not surgically remove the bad actors. It leaned on the one signal it fully trusts, whether a domain is already established and known to be good, and flattened almost everyone else.

How Google’s Helpful Content Update flattened HouseFresh

HouseFresh lost roughly 95% of its Google traffic after the September 2023 Helpful Content Update, falling from about 4,000 daily visitors to about 200. It ran genuine hands-on air purifier reviews, exactly the first-hand expertise Google’s E-E-A-T guidance claims to reward.

The March 2024 core update hit HouseFresh again.

Meanwhile, big media sites running affiliate roundups by writers who never touched the products stayed at the top, because they were already trusted domains. HouseFresh followed the guidelines, and the guidelines did not save it.

What Microsoft’s AI search guide tells you to do

Microsoft’s AI search guide tells you to add structured schema markup to your pages and break your content into clean, modular question-and-answer blocks an AI can lift straight out.

Microsoft published its guidance in October, and it gives you concrete, tactical steps where Google gives you general advice.

In other words, Microsoft tells you to do several of the things Google waved off or warned against. One of these companies is teaching, and the other is managing you.

How to rank in AI Overviews: the playbook for ecommerce stores

The AI search playbook for a store comes down to reading Google’s warnings as hints and Microsoft’s guide as the instructions. Here is how I would put it to work.

• Treat Google’s guide as a confession. Read every warning in it as a hint about what is quietly working.
• Use Microsoft’s guide for the how-to, including schema markup and question-and-answer blocks.
• Build genuinely distinct content for each real customer segment, where every version earns its place.
• Earn mentions only on sources Google already trusts and cites, starting with Reddit.
• Stop spending time on llms.txt files and markdown copies of your pages, which Google says do nothing.

That is the whole playbook, pulled from between the lines of a document written to make sure you never found it.

Frequently asked questions

Do you need an llms.txt file to rank in AI search?

Google says an llms.txt file does nothing for its AI features, so you can skip it for Google. Google also says you do not need markdown copies of your pages or chunked content to appear in AI Overviews.

Does Google penalize AI-generated content?

Google now says AI-generated content is fine as long as it is helpful. That reverses two years of warnings, likely because Google cannot reliably tell AI-written content from human-written content.

Is writing separate content for different customers considered spam?

Writing separate content for different customers is spam only when the pages are near-identical rewrites targeting the same keyword. A genuinely distinct guide for each real audience, such as wedding, corporate gift, and memorial buyers, serves different people and works well in AI search.

Which sources do AI Overviews cite most?

Reddit is the most-cited source in AI answers, making up roughly 40% of citations across major AI engines and about 21% of the citations inside Google’s AI Overviews. Earning genuine mentions there is one of the most effective ways to get a brand cited.

How often do people click links in Google AI Overviews?

Pew Research found that people click a regular link 8% of the time when an AI summary appears, versus 15% without one. Only 1% click a source link inside the AI summary.

Does traffic from AI search convert better?

Shoppers arriving from AI sources like ChatGPT convert 31% higher than visitors from non-branded organic search. Traffic to US retail sites from generative AI sources also grew 1,200% in a single year, according to Adobe.