What Google itself says about getting into AI answers

Is there a separate way to rank in Google's AI answers?

No, and Google now says so in writing. Its guide, Optimizing your website for generative AI features on Google Search, last updated on 10 July 2026, states that AI Overviews and AI Mode are rooted in the same core ranking and quality systems as ordinary results. A page must be indexed and eligible to show a snippet before it can appear in either.

That matters for a founder because a whole industry has grown up selling AI search as a new discipline with new tricks. Google's position is narrower: do search well, and offer something worth quoting.

What does Google say actually helps?

The guide boils down to four things, and none of them is new to anyone who has run a website properly.

• Original, non-commodity content. Google singles this out. Pages that repeat common knowledge give an AI answer nothing to cite. Pages with first-hand detail do.
• Technical basics. Pages must be crawlable, indexable, render properly if they rely on JavaScript, and avoid duplicate content.
• Current business and product data. Google points to Google Business Profile and Merchant Center as ways to help your products and services show in AI responses.
• Measurement. Search Console now has a report for generative AI features, which I cover in the guide to free AI visibility reports.

What does Google tell you to stop doing?

This is the useful half of the guide, because it names tactics that agencies are still selling.

• llms.txt and other AI files. Google Search does not use them. Creating one will not help or hurt your Google visibility.
• Chunking pages for AI. There is no need to cut content into small pieces and no ideal page length.
• Rewriting copy for machines. No special writing style is needed. Write for the buyer.
• Pages for every sub-question. Mass-producing pages to match each variation of a query falls under Google's scaled content abuse policy.
• Manufactured brand mentions. Inauthentic mentions are called out as a tactic to avoid.

Google also tells site owners to be wary of third-party AEO and GEO tools that promise ranking success. Coming from the company that runs the results, that is a pointed line.

So is llms.txt pointless?

For Google Search, yes. Elsewhere the picture is mixed. Lighthouse 13.3, the audit tool built into Chrome, added an llms.txt check in May 2026 under a new Agentic Browsing category. It is aimed at AI agents that browse a site on a person's behalf, not at search ranking, and a missing file is marked as not applicable.

My view: if you already have a clean llms.txt, leave it. If you are paying someone to write one as your AI strategy, that money belongs elsewhere.

What does this mean for an Indian manufacturer's website?

It shifts the question from "what AI tricks are we missing?" to "what does our site say that nobody else's does?" On many factory websites the answer is very little. The product pages repeat catalogue language, the capacity and approvals sit in a PDF, and the application knowledge lives in the sales team's heads.

Google's guide rewards the opposite: plant capacity in plain text, the standards you are approved to, tolerances you hold, industries you supply, and worked examples of problems you have solved. That is also what a purchase manager wants to read. For how assistants beyond Google weigh these signals, see how AI assistants decide which companies to recommend; the AI Visibility section has the rest of the reading path.