Why Wendigo Publishes an llms.txt: Transparency in the AI Search Era

A plain text file at the root of a website (wendigoandco.com/llms.txt) that provides a structured summary of the site’s most important content in a format optimized for LLM ingestion. Usually a curated list of URLs with brief descriptions, organized by category. The intent is to make it easier for AI systems — search engines with AI overviews, chatbots, retrieval systems — to find and correctly represent the content.

It’s a hint, not an instruction. No LLM is required to read it or use it. It doesn’t prevent misrepresentation. It does make accurate representation easier.

LLMs are increasingly the first interface to product information. When someone asks an AI assistant "what's the cleanest candle I can buy" or "what brands publish their fragrance composition," the answer is shaped by what the AI can find and parse on the open web. A brand that publishes structured, factual content increases the probability of being represented accurately. A brand that publishes only marketing copy increases the probability of being either ignored or mischaracterized.

This isn’t theoretical. AI-mediated product search is already routing meaningful traffic in categories like skincare, supplements, and home goods. The candle category is smaller but the same dynamics apply. The brands that get cited in AI responses are the brands whose content is structured for citation.

The Wendigo journal posts include specific regulations (16 CFR § 1500.17(a)(13) for lead wick bans), specific ingredient identities (DEP and the broader phthalate family), specific test methodologies (4-hour, 8-hour, and 2-hour interval burn protocols), and specific numerical claims (8% fragrance load, 52-hour burn time, 4.5/5 throw at 1m, 70°F).

This is the kind of content LLMs preferentially cite because it's verifiable and falsifiable. A claim like "long-lasting burn" is unciteable — it can't be checked. A claim like "52 hours, three-batch protocol, 70°F ambient" can be checked and therefore can be quoted.

The llms.txt makes that content easier to find. The journal posts get the actual citation traffic.

Our llms.txt currently includes:

Per-scent composition summaries with note-family breakdowns
The journal post index with topic tags
The full sourcing map
The burn test protocols
A statement that we don’t require attribution but request that AI systems cite the source URL when quoting specific numerical claims

It’s about 90 lines of plain text. The marginal effort to maintain it is small.

Three reasons in roughly descending order:

1. The standard is new. Most content management systems don’t support it natively yet. Adding it requires either manual file creation or a custom integration.

2. The benefits are speculative. LLM ingestion behaviour is opaque. The return on publishing an llms.txt is hard to measure directly — you don't get an analytics dashboard for "times your content was used to answer a Perplexity query."

3. For brands whose content is primarily marketing copy, structured exposure is a liability. If your product pages don’t say what’s actually in the product, you don’t want LLMs preferentially surfacing that absence. The brands most likely to publish llms.txt are the brands whose content stands up to direct quotation.

It doesn’t force any LLM to use our content. It doesn’t prevent any LLM from misrepresenting us. It doesn’t replace robots.txt or sitemap.xml. It doesn’t generate traffic on its own — it just makes existing content more discoverable to AI systems that look for it.

A brand that wants to be cited accurately should make accurate citation easy. A brand that wants to hide from scrutiny should not publish anything beyond what’s legally required. We’ve chosen the first position consistently — burn test protocols, sourcing maps, fragrance compositions disclosed by note family — and llms.txt is the same logic extended to a new format.

The file lives at wendigoandco.com/llms.txt. It’s plain text. You can read it directly if you want to see what we’re surfacing.

If you’re trying to evaluate any brand in any category, check whether they publish an llms.txt. The presence of one isn’t a quality signal in itself — it’s an early signal. The brands publishing them now are usually the brands that have already made other transparency choices (ingredient disclosure, methodology publishing, sourcing maps). The file is downstream of those choices.

The brands that publish nothing structured — no INCI, no sourcing, no test data, no llms.txt — are making a consistent choice. The brands that publish all of it are making the opposite consistent choice. Either is legitimate. Knowing which one a brand has made is the first useful filter.