Buyers used to search Google and click through a list of links. Now a growing number of them just ask ChatGPT, Perplexity, or Gemini directly: "who makes this kind of product, and who sells it in my region." The AI picks one answer and hands it over. That means a website now has two different readers, Google and the AI chat tools, and they don't read the same way. What Google Wants Google's own guidance is direct: its AI features (AI Overviews, AI Mode) run on the same ranking system as regular search. No special files needed. No special markup. Google says plainly that llms.txt files, which many sites now add specifically for AI, are ignored by Google entirely. Google also warns against writing a separate version of content just for AI, or chopping it into unnatural fragments. The advice is the same advice good SEO has always given: clear, original, well-sourced content written for people. What the Other AI Tools Want ChatGPT, Perplexity, Claude, and Copilot are different. They don't use Google's system, and they do reward specific things: Quotes from named experts Statistics with a named source Citations to reliable sources Content structured so a single passage can stand alone and answer a question A Princeton and IIT Delhi study (presented at the KDD conference in 2024) tested this directly, measuring how much of an AI-generated answer got credited to a source after each change. Adding a quote from a named expert was the strongest single lever: a 41% lift. Adding statistics came in at 33%, adding citations at 28%. Two things standard SEO advice would predict didn't hold up: writing in a more confident, persuasive tone showed no significant effect, and stuffing in extra keywords, the old-school trick, made things slightly worse. These tools also read things Google ignores: an llms.txt file, a real pricing page instead of a "contact us" form. This matters more every month, because AI agents are starting to compare suppliers automatically, on a buyer's behalf. If pricing and specs aren't readable on the page, the agent moves on to a competitor whose page it can read. What a New Regional Site Needs, Concretely A visible pricing page (not hidden behind a form) An llms.txt file at the site root, a short plain-text summary of what the company does Basic schema markup: Organization, FAQPage, Article, Product A robots.txt file that allows the AI crawlers through: GPTBot, PerplexityBot, ClaudeBot, Google-Extended, Bingbot Content that leads with a direct answer (40-60 words), before the supporting detail Comparison tables instead of paragraphs, for anything shaped like "X vs Y" A dated FAQ section with real questions Why the Language Matters All of this has to exist in the local language, not just the original. A perfectly optimized English page does nothing for a buyer who asks a question in Russian. The AI answering that question needs a source that answers in Russian. So the checklist above isn't a one-time job. It has to be repeated for every market, in that market's language, or the site only ever gets found by half its actual audience. Contributed by Lina Sorokina, CEO at Artifact Agency, which works with foreign manufacturers entering CIS and Central Asian markets.