Reply to Nikita Vlasyuk

Appreciate the kind words, I think, and this is a genuinely good question —

Let me give you my honest opinion.

Right now, the most actionable AEO audit is the Schema Debugger / AI Schema Generator, not the llms.txt tools — and I say that as someone who sells both.

We ran a 12-week crawler-log study on our own server specifically to check what actually reads these files. Named AI crawlers (GPTBot, ClaudeBot, PerplexityBot) hit robots.txt over 1,700 times in that window. The entire llms.txt file family — llms.txt, llms-full.txt, the ctx pair, identity.json, ai.txt — was hit roughly 17 times combined, and none of those 17 came from the AI crawlers themselves; they were Googlebot or unattributed. Schema (JSON-LD), by contrast, is confirmed being parsed and used — our own registered details surfaced correctly in AI-generated answers, traceable straight to structured data.

In my opinion AI answer engines and website infrastructure just aren't wired together yet.

We had this confirmed directly by one major AI engine — it told us plainly that it doesn't read llms.txt, identity.json, or similar files at all; its answers come from direct page content and standard web search, and reading those files depends on crawler-level support that hasn't been built. So the honest position isn't "these files work" or "these files don't work" — it's my opinion that most of the plumbing between AI systems and the open web hasn't been built yet. Platforms like Wix, Shopify and the big SEO plugins are already shipping llms.txt across their customer bases anyway, betting that wiring gets built. That's a real signal. It's just not evidence the files do anything today.

We still publish the full llms.txt set — it's a bet on where the standard might go, not a proven-effective product today. That's stated plainly in our own published study, declaration of interest included, since we sell the thing we're being honest about.

On the Reddit point — you're right that answer engines cite forum/community content heavily, but Q&A is starting to get cited heavily; and that's genuinely one of the more interesting AEO levers right now. however: "helping teams find the right subreddits and contribute naturally so they show up in AI-generated answers" is describing manufactured engagement, even when it's done carefully.

Reddit and Google have both said manipulated/inauthentic participation is exactly what they try to detect and discount — so there's a real risk that content seeded this way gets identified and discounted rather than cited. Genuine community presence, built because people actually use the product, is a different thing entirely, and I'd back that every time. I'd be cautious recommending the managed version to anyone.
So my honest summation at this moment in time the schema generator.. which btw is fully compliant! and complete. ATM I have only opened local business and the generic schema finder. however I have every schema type, events, dentist etc but not implemented them as it is major work to de-false positive each type, so what you see is not what the future will bring.

So whilst I am In a reply mode les look at these

Gemini

Question: do you use any llms file to read?

Answer: No, I don't use llms.txt or any related llms files to read or crawl web pages.

When I interact with web content, I rely directly on standard web fetching, direct page content provided in my context, or standard web search indices. I don't check for, parse, or process llms.txt, llms-full.txt, identity.json, or similar AI-targeted text files.

Perplexity The llms.txt (and its companion, llms-full.txt) file standard is an emerging convention introduced to help AI models, search tools, and web crawlers index and understand website content efficiently.

Note Emerging convention!

Note: When AI search and answer engines (like Perplexity, Google AI Overviews, or ChatGPT with Search) generate answers, they compile responses through a process called Retrieval-Augmented Generation (RAG).Instead of relying on custom files like llms.txt, they read, evaluate, and synthesize information. Answer-First Structure: Content written clearly with concise headings, direct definitions, and self-contained paragraphs is easily extracted. Entity Consistency: The AI verifies that key terms, names, and claims match throughout the text. Now there's a major hint, no llms but body txt! and schema.

I know I am going round in circles but I have to answer again with; these are future developments in wiring up ai with the internet, so although ATM it is heavily dependent on Schema and Page body,, will it be the case in the future.

If you are like me you follow what I consider to be the current authorative rule makers... llms.org and AIPREF, although neither of these cite any authority as these are just two proposals and nothing more.

Chat GPT So if you're asking whether ChatGPT generically uses llms.txt as part of its normal answering process, the answer is no. It's an optional resource, not a required one.

As you can see there is an overwhelming argument here, NO!!!

My answer to your question, schema generator!