Every few months, AI search creates a new cottage industry of technical superstition.
The file changes. The acronym changes. The pitch changes. The promise stays the same: "there is a special machine-facing trick that will make Google trust you."
Google has now answered that pretty directly.
Why this question keeps coming up
The anxiety is understandable. AI Overviews, AI Mode, and other answer-driven interfaces make site owners feel like traditional publishing logic is dissolving. So people reach for new artifacts: llms.txt files, aggressive content chunking, or custom schema theories that promise a clean shortcut into the answer layer.
But most of that energy is aimed at the wrong layer of the problem.
What Google actually says
In May 2026, Google published its AI optimization guide. The tone is much less mystical than the broader SEO discourse.
- Google says optimizing for AI features is still standard SEO.
- Google says there is no special schema required for AI features.
- Google says
llms.txtis not used for Google Search. - Google says site owners should not overthink "content chunking."
That does not mean machine-readable structure is useless. It means the industry's favorite new hacks are being oversold.
Search Engine Journal's May 2026 coverage got the core takeaway right: Google is effectively saying AEO and GEO are still SEO, not separate religions.
Where llms.txt can still help
This is the important nuance. llms.txt can still be useful in some workflows. It may help other tools, custom agents, internal retrieval layers, or documentation-heavy systems where you want to expose a clean machine-readable map of your site.
I would treat it as a workflow convenience, not a Google visibility lever.
Analysis: this is a recurring pattern in AI tooling. Something can be operationally useful without being a ranking factor.
What to fix instead
If the goal is better visibility in Google AI search, the higher-leverage fixes are still boring and structural:
- Write pages that answer directly. Reduce vague positioning copy.
- Keep visible text aligned with structured data. Do not say one thing in the markup and another on the page.
- Improve proof density. Put case studies, examples, and specifics close to claims.
- Make navigation and internal links cleaner. Google still needs a coherent site, not just a clever file.
- Publish original, non-commodity material. Google explicitly emphasizes unique value.
This is less exciting than a hack, but it is also more defensible.
CTA: If your team is spending more time debating llms.txt than rewriting vague money pages, you are probably optimizing the wrong layer first.