Mostly, yes, for Google's own surfaces. In May 2026 Google formalized its position in an official guide, "Optimizing your website for generative AI features on Google Search," which states plainly that optimizing for its AI features is search optimization, and that you don't need markdown, llms.txt, or any AI-specific files to appear. That's accurate for AI Overviews and AI Mode. It's incomplete as a whole-ecosystem strategy, because Google's guidance stops at the edge of Google's own products.
What did Google actually say?
Google's guide makes two blunt claims. First, on the discipline: "From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO." Second, on tooling: you "don't need to create machine-readable files, AI text files, markup, or Markdown to appear in generative AI search." The reasoning is that Google's AI features are rooted in its core ranking and quality systems. AI Overviews and AI Mode pull from the same index that ranks classic results, so the same signals apply. Google representatives have repeated the message consistently since 2025, including that Google does not crawl or use llms.txt.
What does Google's guide say actually helps?
The guide's recommendations read like a disciplined SEO program, not a new playbook. Google emphasizes non-commodity content: material with unique insight beyond common knowledge, and lists concrete technical fundamentals.
| Google's recommendation | What it maps to |
|---|---|
| Unique, non-commodity content | Evidence density, first-hand data |
| Follow crawling best practices | Reachability. Don't block the bots |
| Use semantic HTML | Extractability. Clean, parseable chunks |
| Follow JavaScript SEO guidance | Content visible without executing JS |
| Strong page experience, minimize duplication | Quality and clarity signals |
| Merchant Center feeds + Google Business Profile | Structured product/local data |
Every row is something answer-engine optimization already treats as table stakes. Google isn't negating the work; it's renaming it. Notably, "it's still SEO" raises the bar. Structured data and answer-first structure aren't optional extras, they're how the same index understands you well enough to synthesize an answer.
"Still SEO" is not "do nothing new." It means the durable levers. Crawlability, semantic structure, structured data, and entity strength. Are the levers, on both classic and AI surfaces. There is no shortcut file.
Where does Google's guidance stop?
At Google's property line. The guide speaks for Google Search's AI features. It says nothing about how ChatGPT, Perplexity, or Claude retrieve and cite, and those engines behave differently. They crawl with their own bots, weight sources differently, and some fetch files Google ignores. Citation overlap between engines is low, so a page ChatGPT loves may be absent from Perplexity's answer. That's the gap GEO names: not a magic technique Google is debunking, but the reality that visibility now spans several engines you don't own, each measured separately.
So the honest reconciliation is this. For Google, do excellent SEO with an answer-first, well-structured, strongly-corroborated site. Exactly what the guide describes, and exactly what wins AI Overviews. Skip the llms.txt and markdown mirrors; Google won't read them and the evidence they help elsewhere is thin. Then treat the wider ecosystem as its own problem: the same fundamentals plus entity strength and per-engine measurement. Google saying "it's still SEO" is a useful correction to file-based snake oil, not a reason to stop measuring where you're actually cited.
What to do next
Read Google's guide as a checklist you can mostly already tick: crawlable, semantic, structured, non-commodity, no special files. If a "GEO" vendor is selling you an AI-specific file as the lever, Google just told you it isn't. The lever is the same durable work, and knowing whether your pages actually pass it, across every engine and not just Google, is what Buffy Intel measures. Start with the fundamentals Google names; verify the result across the engines Google doesn't speak for.