BLOG / does-llms-txt-or-sku-md-help-seo.mdx
Does llms.txt or SKU.md help SEO? Measure the right claims
Separate crawlability, content quality, agent readability, platform adoption, and ranking—and define honest 30/60/90-day measurements.
Publishing llms.txt or SKU.md can create measurable technical access and editorial discipline. It cannot, by itself, prove better rankings, AI citations, recommendations, or sales.
SKU.md is merchant-hosted product discovery and stable product knowledge: agents can find products, understand sourced facts, and know where to recheck live commerce. It is not a new checkout protocol.
Five claims that must stay separate
| Claim | Can the publisher measure it? | What it does not establish |
|---|---|---|
| Crawlability | Status, robots access, logs | Indexing or ranking |
| Content quality | Review coverage, evidence, contradictions | Search-engine judgment |
| Agent readability | Parse success, task evaluations | Real platform use |
| Platform adoption | Documented fetches or explicit official support | Universal adoption |
| Ranking or conversion | Search/analytics trends with controls | Causation from one file |
llms.txt may make a curated set of readable resources easier to fetch. SKU.md may make product evidence easier to inspect. Neither file tells a crawler that it must index or trust the content.
Google’s guidance is explicit
Google’s AI features and your website says there are no additional requirements or special optimization for AI Overviews or AI Mode. It also says publishers do not need new machine-readable AI text files or special Schema.org markup to appear.
Pages still need ordinary Search eligibility and helpful, reliable, people-first content. Meeting requirements does not guarantee that Google will crawl, index, or serve a page. Product structured data should match visible content; it remains the documented product-search surface rather than evidence that SKU.md affects ranking.
A 30/60/90-day measurement plan
| Window | Measure | Honest interpretation |
|---|---|---|
| 30 days | Exact 200/MIME, indexing state, server fetches, broken links, baseline queries | Technical availability and baseline only |
| 60 days | Evidence freshness, agent task accuracy, Search Console impressions/clicks, referral labels | Association and content operations |
| 90 days | Query cohorts, landing-page trends, assisted conversions, controlled comparisons | Directional evidence, not automatic causation |
Record deployment dates, unrelated site changes, seasonality, and platform updates. Compare similar product groups when possible. A before/after chart without controls cannot isolate one machine-readable file.
What to improve even without an SEO promise
Make important content visible in HTML, keep internal links coherent, use correct canonical URLs, maintain Product JSON-LD that matches the page, and publish accurate source-backed product explanations. Validate public delivery and correct contradictions promptly.
These practices help people and make the site easier to audit. They remain useful even when ranking, citation, recommendation, and conversion are unproven.
Further reading
Read What is SKU.md? for the category boundary and the llms.txt history for implementation context.