Someone has told you that you need “AI Mode optimisation”. Before you pay, read what Google itself says about getting a site into Google AI Mode. It’s shorter than you’d expect.

That’s the whole official position. The rest of this piece shows what those fundamentals look like in practice, what Google says you can safely ignore, and which claims from outside Google have nothing official behind them.

How Google AI Mode finds pages

Google AI Mode is a Search experience built for questions that need exploration, reasoning or comparison^[1].

Google describes one technique behind it: query fan-out. Instead of running only the exact words the person typed, the system may issue several related searches, then pull the results together into one answer^[1]. So a single question can lead to pages that wouldn’t rank for the original wording.

There’s a second detail worth knowing. AI Mode and AI Overviews may use different models and techniques, so the links they show can differ^[1]. Appearing in one tells you nothing certain about the other. It also means nobody can honestly promise you a fixed position.

Eligibility: the two gates

Google’s documentation sets out the conditions plainly. A page has to be indexed, and it has to be eligible to show in Google Search with a snippet^[1][2]. If a page is blocked from indexing or can’t show a snippet, it isn’t a candidate.

Your site also needs to be included in generative AI features in Search, which you can check in Google Search Console^[2]. Meeting these conditions makes a page eligible. It doesn’t guarantee the page will be indexed or served, and Google says so^[1][2].

What the fundamentals look like in practice

Google lists the basics for generative AI features, and they read like standard SEO housekeeping^[1]. Here’s each one in plain language.

• Allow crawling. Check your robots.txt file (the file that tells crawlers which areas of a site they may visit) doesn’t block pages you want found. Google also mentions any CDN or host in front of your site, so check they aren’t blocking Google’s crawlers either^[1].
• Use internal links. Link to important pages from other relevant pages so crawlers and visitors can find them^[1].
• Give a good page experience. Pages should load and work well for the people using them^[1].
• Put important content in text. If a key detail only exists inside an image or a script-driven widget, put it in written text on the page too^[1].
• Use quality images and video. Where they support the content, make them good ones^[1].
• Match structured data to visible text. Structured data is code that labels what’s on a page, such as a price or a review rating. What you mark up must agree with what visitors can see^[1].
• Keep Merchant Center and Business Profile details current. More on that below^[1].

Nothing on that list is new. If your site has been neglected on any of it, fix that first. Don’t open a separate AI workstream.

What Google says you can ignore

This is the part that can save you money. Google’s documentation says you don’t need any of the following for generative AI search^[1][2]:

• An llms.txt file or any other special file or markup for AI
• Chunking your content (breaking pages into small, bite-sized pieces for machines)
• Rewriting your content just for AI
• Seeking inauthentic mentions of your brand around the web
• Over-focusing on structured data. No special schema.org markup is needed^[1][2]

Google’s own framing is that optimising for generative AI search is still SEO^[2]. It also points readers to its advice on evaluating third-party SEO guidance before they buy services sold as “AEO” or “GEO”^[2]. GEO (generative engine optimisation) is the label many agencies use for work aimed at AI answers. As a label, it’s fine. As a reason to buy tactics Google has told you to skip, it isn’t.

Write content only you could write

Google says unique, non-commodity content will likely matter most^[2]. Commodity content is what anyone could produce by restating what already exists. Non-commodity content brings something new.

Google’s own contrast is a good one: a first-hand review gives a unique perspective, while a summary only restates what’s out there^[2]. Applied to your site, that means:

• Write product pages that say what you’ve seen from using, testing or shipping the product, not just the manufacturer’s spec sheet.
• Have the person who does the work write the guides, with their actual opinion on what to choose and why.
• Add original photos and video of your products, premises or process.

Don’t build a page for every query variation

Fan-out tempts people into a bad idea: if AI Mode searches many related questions, why not publish a page for each one? Google addresses this directly. Separate content for every query variation, made mainly to manipulate rankings or AI responses, violates the scaled content abuse policy^[2][4].

The policy sits within Google’s spam policies for web search^[4]. One thorough page that covers a topic and its natural follow-up questions is a better plan than a pile of thin near-duplicates. Create a new page only when it has its own purpose for a real reader, not because a keyword tool shows a slightly different phrasing.

Keep Merchant Center and Business Profile details up to date

Google lists up-to-date Merchant Center and Business Profile information among its fundamentals^[1]. It doesn’t say more than that, so don’t read hidden rules into it. If you sell products, make sure your product data matches your site. If you serve local customers, make sure your business details are correct.

If you’re not sure which one does what, our guide to Google Merchant Center vs Google Business Profile explains the difference: https://oxfordimpulse.com/google-merchant-center-vs-google-business-profile/ . If you sell online and haven’t built your product data properly, the complete guide to Google Merchant Center covers the setup: https://oxfordimpulse.com/complete-guide-google-merchant-center/ .

How to see your results

Google gives you reporting for this. Activity from these features is counted in the Performance report in Google Search Console, and there’s also a Generative AI performance report^[1][3]. Open Search Console, find those reports and review them on a regular schedule.

Use them to see which pages get impressions, then compare against the pages you’d expect to perform. A strong page that never shows up is a prompt to check indexing and snippet eligibility. A weak page that does show up is a hint about what’s working.

Spend your time on the pages and fundamentals, not on special files. For the wider picture of how Google’s AI search experience fits into your visibility plan, read our overview of Google AI Mode: https://oxfordimpulse.com/google-ai-mode/ .

FAQs

Sources

• Google Search Central: AI features and your website developers.google.com
• Google Search Central: Optimizing your website for generative AI features on Google Search developers.google.com
• Search Console Help: Generative AI performance report (Search) support.google.com
• Google Search Central: Spam policies for Google web search developers.google.com