llms.txt: does your Malaysian SME website actually need one?

An llms.txt is a plain text file at your site root that briefs AI agents on your business. It is cheap and harmless to add. But as of mid-2026 it is not the thing that gets you cited by ChatGPT or Google, and no honest vendor should sell it to a Malaysian SME as if it were.

Key takeaways

• llms.txt is a plain-Markdown file at your site root (/llms.txt) that summarises your business for AI agents. It was proposed by Jeremy Howard of Answer.AI on 3 September 2024 as a community proposal, not an official standard.
• As of mid-2026 no major engine (OpenAI, Google or Anthropic) has publicly confirmed using llms.txt as a ranking or recommendation signal. Google's Gary Illyes stated in July 2025 that Google does not support it (per Search Engine Land).
• Adoption is concentrated in developer and documentation sites, not the general web. The documentation host Mintlify switched it on across its hosted docs in November 2024; most SMEs that get cited do not publish one.
• It costs nothing and takes minutes to add, so adding one is fine. It is just not the lever.
• What actually moves AI visibility is schema.org structured data, answer-first crawlable content, and being cited across the web. The Princeton GEO study found statistics, quotations and citing sources lift AI-answer visibility by up to roughly 40 percent.

What is llms.txt, and what goes in it?

llms.txt is a single plain-Markdown file at your site root, reachable at yourdomain.com/llms.txt. It is a short brief, written for AI agents rather than people, telling a model what your business does, what you offer, and where your key pages live. Think of it as robots.txt aimed at large language models, not crawlers.

Per the specification at llmstxt.org, the format is deliberately simple: an H1 with your business name, a one-line summary in a blockquote, then optional sections of links with short notes. The stated purpose is to "provide information to help LLMs use a website at inference time", because, as the spec puts it, "context windows are too small to handle most websites in their entirety." So instead of an AI agent wading through your full HTML, ads and JavaScript, it reads a clean summary you wrote. It is a courtesy file, not a magic switch.

Who created it, and is it an official standard?

It is not an official standard. llms.txt was proposed by Jeremy Howard, co-founder of the AI research lab Answer.AI, in a post published on 3 September 2024. The post is explicit about its status. It opens, "We propose that those interested in providing LLM-friendly content add a /llms.txt file to their site", and calls itself a "proposal".

A real web standard is ratified by a body like the W3C or IETF and is something engines commit to honour. llms.txt has neither. It is a community idea, open for input on GitHub, that website owners may voluntarily adopt. The proposal itself notes it "does not include any particular recommendation for how to process the file, since it will depend on the application." Nobody is obliged to read it, and the author never claimed they would. So a file no engine is required to use cannot be guaranteed to do anything for you, which is the right lens for any "you must have llms.txt" pitch.

Do ChatGPT, Google and Anthropic actually read llms.txt?

As of mid-2026, no major engine has publicly confirmed using llms.txt as a ranking or recommendation signal. The clearest position is Google's: at Google's Search Central Deep Dive APAC event in July 2025, Gary Illyes stated that Google does not support llms.txt and is not planning to, as reported by Search Engine Land.

The proposal itself never promised engines would adopt it. Jeremy Howard framed llms.txt as an opt-in courtesy, not a mandate:

"We propose adding a /llms.txt markdown file to websites to provide LLM-friendly content." — Jeremy Howard, The /llms.txt file (llmstxt.org), September 2024

Google's John Mueller has separately compared the file to the long-deprecated meta keywords tag, a signal engines stopped trusting because the site owner controls it. Be honest about that source, though: the Search Engine Land write-up of Mueller's comparison actually argues llms.txt is not the new meta keywords and defends the file, so the comparison is contested, not a verdict against the file.

OpenAI and Anthropic have not made the same flat public commitment either way. Some observers report seeing AI agents fetch the file, but observation is not confirmation, and a site-controlled file is exactly the kind of signal engines tend to weight lightly. Where llms.txt has clearly caught on is documentation sites. The documentation host Mintlify announced support in November 2024, saying it "now automatically generates and hosts all docs in a plain text format for easier ingestion by large language models", which switched the file on across the developer-doc sites it hosts. That is a useful pattern for technical documentation. It is not evidence that a Klang Valley accountant gets recommended because of a text file.

We made the same point in getting recommended by ChatGPT in Malaysia: treat llms.txt as informational only, and never let it be sold to you as the thing that gets you into ChatGPT.

What should be in a good llms.txt for a Malaysian SME?

If you choose to add one, keep it short and factual. A good SME llms.txt names your business, says what you do and where, lists your core services and any price guidance, and links to your most important pages. The point is to give an agent the cite-worthy facts in one clean place. Here is a usable skeleton:

# Acme Trading Sdn Bhd

> Family-run customs clearance and haulage firm in Port Klang, Selangor,
> serving importers and exporters across the Klang Valley since 2014.

## Services
- Customs clearance (import and export)
- Container haulage and trucking
- Sea and air freight forwarding

## Key facts
- Location: Port Klang, Selangor, Malaysia
- Founded: 2014
- Contact: WhatsApp +60 12 345 6789

## Important pages
- [Services](https://example.com/services): full service list and coverage
- [About](https://example.com/about): the team and licences
- [Contact](https://example.com/contact): enquiry and quote requests

Notice what makes it work: specific facts, real numbers, plain language. Those same qualities make your actual web pages cite-worthy too, which is the more important place to put them.

Is it worth adding? The honest verdict.

Yes, add one if it is free and quick, but spend zero budget chasing it and expect no miracle. For a Malaysian SME deciding where to put limited time and money, llms.txt is a tidy-up, not a growth lever. The cost-benefit is simple:

• It is cheap. A basic llms.txt takes minutes to write and costs nothing to host. There is no downside to a clean, accurate one.
• It may help the agents that do read it. Some tools and crawlers fetch it. A tidy brief cannot hurt and might save an agent the work of parsing your site.
• It is not the lever. No major engine has confirmed it as a ranking signal, and Google has said it does not use it. If your site is invisible in AI search, llms.txt will not fix that.
• Never pay a premium for it. If a vendor's whole AI-search offer is "we add an llms.txt", they are charging you for a five-minute file and skipping the work that matters. It is a harmless bonus, not a strategy.

How do you create and publish one?

Publishing an llms.txt is a short technical job, the kind a developer or a managed website service does in minutes, and you can do it yourself with a plain text editor. Here is the whole process, start to finish:

1. Open a plain text editor and write the file using the structure above: H1 business name, a blockquote summary, then Markdown sections of services, facts and key links.
2. Save it as llms.txt (plain text, no formatting).
3. Upload it to the root of your website so it is served at yourdomain.com/llms.txt, the same level as your robots.txt and sitemap.xml.
4. Visit yourdomain.com/llms.txt in a browser to confirm it loads as plain text and the links are correct.
5. Keep it accurate. When your services or contact details change, update the file, the same way you would update any page.

That is the entire job. On a managed platform, ask whoever runs it to drop the file in; on a static or server-rendered site like the ones we build, it sits alongside the other root files automatically.

What actually moves AI visibility instead, and as well?

If you want to be cited by ChatGPT, Perplexity, Gemini and Google's AI answers, the work is in your pages and your reputation, not a root file. Three moves carry almost all the weight, and they are the same ones that have always built trust online.

First, structured data. schema.org JSON-LD tells engines what your business is, where it is, what it offers and what it costs, in a language they actually parse. We walk non-developers through it in schema.org for non-developers, and it is the single highest-leverage technical move a Malaysian SME can make.

Second, answer-first crawlable content. The Princeton GEO study (Aggarwal et al., KDD 2024) tested nine tactics across 10,000 queries and found that adding statistics, quotations and citing sources produced the strongest gains, lifting a page's visibility in AI answers by up to roughly 40 percent. Vague copy is invisible; specific, factual, well-structured pages get quoted. Crucially, those pages must be readable as plain HTML, because engines do not run your JavaScript.

Third, presence across the web. AI models gain confidence in a business named consistently across independent sources: your own site, a directory, a LinkedIn presence, a review profile, a press mention. A single mention on your own site is weak corroboration. The broader GEO vs SEO picture for Malaysian SMEs covers how those signals stack up. With Malaysia at 35.4 million internet users and 98.0 percent penetration (DataReportal, Digital 2026), the audience asking AI for recommendations is nearly everyone.

Where to start

If you have been told you need an llms.txt, the honest read is: add it if it is free, but do not mistake it for AI-search strategy. The real work is a fast, crawlable, server-rendered site with schema.org baked in, answer-first content, and presence across the web.

Every site we build at Wiz Studio Labs ships server-rendered HTML, schema.org structured data on every page, FAQ blocks, and yes, an llms.txt at the root, so you are covered on the cheap bonus and on the parts that actually count. If your current site cannot do that, see our templates or start a brief. RM 399 a year, one edit included, and you pay only if you keep it.