llms.txt on Squarespace

The function is narrow on purpose. A standard web crawl returns thousands of fragments — paginated indexes, tag pages, search results, outdated drafts — and an AI engine summarising a site has to guess which version of a topic is the canonical one. llms.txt removes the guessing. A photographer's llms.txt might point an AI to one services page, one pricing page, one investment guide, and skip the dozens of gallery permalinks that surface in a standard sitemap. The engine ends up with the authoritative source it needed, not the loudest one.

The spec¹ is short. One required section, three optional. That is the whole protocol. The file format is markdown, so a human can read it, but the audience is machine. The location is the site root, because that mirrors robots.txt and sitemap.xml — patterns AI clients already know to look for.

The publisher list is verifiable. Anthropic's developer-docs llms.txt⁵ opens with # Anthropic Developer Documentation, a one-line summary of what the docs cover, and an H2 structure mapping API areas to URLs. Perplexity's version⁶ sits at docs.perplexity.ai/llms-full.txt with the same template. Cloudflare's⁷ is the cleanest marketing-side example — H1, blockquote describing Cloudflare Workers, then H2 sections for Products, Solutions, Other Pages, and a Full Context Version link.

The consumer side is murkier. OpenAI's bot documentation⁹ lists GPTBot, ChatGPT-User, and OAI-SearchBot and does not mention llms.txt anywhere. Anthropic's robots policy describes ClaudeBot, Claude-User, and Claude-SearchBot with no llms.txt commitment. Perplexity has not publicly stated that PerplexityBot uses llms.txt at answer time. Where llms.txt unambiguously gets read today is by IDE coding assistants — Cursor, Continue, Claude Code — and by MCP-server integrations that point an AI at a documentation entry-point.

The reason to ship llms.txt now is not 2026 traffic. It is 2027 traffic. The publisher-side adoption pattern — Anthropic, Perplexity, Cloudflare, and a long tail of documentation sites — is what happens before the consumer-side commitment lands. The protocol is the same one that would have looked optional for sitemap.xml in 2005. Ship it because the install is three minutes and the file is markdown, not because Perplexity Search confirms today that it changes anything.

The forum thread³ documents the constraint plainly: there is no built-in way to create a file at /llms.txt on a Squarespace 7.1 site. Kenny Kane's post⁴ confirms the same and adds the detail that helped — Squarespace uploads land under /s/, so the workaround chain is: upload (or build) the file content somewhere Squarespace will serve, then redirect /llms.txt to it.

Two install paths emerge from that constraint. The cleaner one is a regular Squarespace page at a slug like /llms, containing the manifest body as markdown, with a URL Mapping that 301-redirects /llms.txt to it. The alternative is a static file uploaded to /s/llms.txt with the same redirect. Both reach the same destination from a crawler's point of view — the difference is editability and Content-Type, both covered below.

The install order matters. The redirect points at the destination page, so the destination must exist first. If the URL Mapping is saved before the /llms page is published, the redirect will land on a 404 and some crawlers will cache that result for hours before retrying.

The slug must be exactly llms. Any other slug works mechanically, but /llms is the convention community write-ups use⁴ and it is what the validators below test against by default. Avoid /llms-txt or /s/llms-txt; both work but neither matches the convention.

 # Asheville Family Photography > Family and lifestyle photography studio in Asheville, NC. This file lists the canonical pages an AI assistant should read first when summarising the studio.
## Core pages - [Services and packages](https://example.com/services): Sessions offered, deliverables, turnaround.
- [Pricing and what is included](https://example.com/pricing): Investment ranges and inclusions.
- [About](https://example.com/about): Founder bio and approach.
- [Contact](https://example.com/contact): Inquiry form and response time.
## Optional - [Client guide PDF](https://example.com/guide.pdf): Pre-session preparation reference.

The spec is permissive about everything except the H1. The single rule: open with # Site Name. The blockquote and the H2-delimited file lists are conventions that humans and parsers both find easy to read, so use them — but the file remains valid without them.

 /llms.txt -> /llms 301 

Squarespace² enforces strict syntax: each rule on its own line, the arrow as -> with no surrounding whitespace adjustments, the status code last. External redirect targets are allowed if they begin with https://, but the install above keeps the destination on-site so the page stays editable from inside Squarespace.

 # Inspect the headers and follow redirects
curl -IL https://yoursite.com/llms.txt
# Fetch the body
curl -L https://yoursite.com/llms.txt

The expected response chain: a single 301 Moved Permanently with Location: /llms, followed by a 200 OK for the destination page. A double 301 means the URL Mapping is correct but the destination page itself is set to redirect — open the page settings and clear any forward.

The spec¹ is silent on Content-Type, which is part of why client behaviour varies. The validators⁸ check structural compliance against the markdown spec, not the served MIME type. The IDE assistants known to consume llms.txt today — Cursor, Continue, Claude Code — fetch the response body and parse markdown from it without strict header enforcement.

The format inside each H2 file list is the markdown link plus an optional short note after a colon, per the spec¹. Cloudflare's public file⁷ follows this pattern exactly: H2 for the section name, then dashed-bullet markdown links with one-line descriptions.

The templates leaf ships four full-length patterns tuned to Squarespace 7.1 collection URLs: a service business, an ecommerce store, a wedding/family photographer, and a therapist or solo-practice professional. Each template lists the specific pages a Squarespace site of that type actually has — Services collection URLs, Store product permalinks, Project gallery items, blog post structure — instead of the generic "/about /pricing /contact" pattern.