# Using your own AI with this site

> The two ways to copy a report to an AI, what rules and tasks the text contains, the plain-text entry points the site offers, and how to keep the AI’s answers checkable.

- verified_at: 2026-10-31
- volatile: no
- page: https://edu.ik.sg/en/kb/using-with-your-ai

After you generate a planning report, the page has “Copy for my AI”. It gives you a block of text with instructions, which you can paste into any AI assistant. The site does not call a model for you.

## Two ways to copy

1. **With knowledge-base links**: for an AI that can open web pages. The text lists the site’s knowledge-base entry points and the AI reads them through the links. It needs the site to be deployed at a public address.
2. **With the knowledge-base text embedded**: for an AI that cannot browse. The full text of the most relevant articles is attached at the end under “knowledge-base excerpts”, each with its source page and verified date (or “unverified”), so it is longer.

## What the text contains

- **A role**: a careful study-planning analysis assistant whose job is to help you understand the report, spot risks, fill in missing information and give checkable next steps.
- **Six rules it must follow**:
  1. Facts come only from the knowledge base; if it is not there, say “the knowledge base does not have this” and do not fill in fees, cut-offs, dates or policies from memory.
  2. Cite the article title or link and its verified date; remind the user to check anything marked unverified or volatile with the official source.
  3. The fit score is not an admission probability; give no probability, success rate or guarantee.
  4. Do not guess dimensions shown as unknown; list what the user needs to add.
  5. If the report and the knowledge base disagree, point out the conflict and which is more likely out of date, and defer to official sources.
  6. Answer in the user’s language, conclusion first and reasons after.
- **Five analysis tasks**: summarise the main conclusion in 3 to 5 sentences; check the five dimensions one by one, separating tested, self-rated and unknown, and name the one that matters most; compare the paths, costs and cautions with the knowledge base for gaps or out-of-date items; list the 5 most worthwhile things to do in the next 30 days, in priority order; finally list what the user should add and the questions to confirm with MOE, ICA or the institutions.
- **The report itself**, and its raw data as JSON.

## Entry points the site offers

- `/llms.txt`: an overview of the knowledge base, listing every article with its title, summary and plain-text link, plus the two entry points below.
- `/kb-export/index.json`: an index of every article with language, summary, category, tags, sources, verified date, whether it is volatile, and links to the page and the plain text.
- `/kb-export/zh/all.md` and `/kb-export/en/all.md`: the full text of all articles in that language.
- `/kb-export/{language}/{article}.md`: one article as plain text, starting with the verified date (UNVERIFIED if it has none) and whether it is volatile, and ending with its sources.

These entry points are public and need no login, and the Markdown article text is served with a cross-origin read permission. The report page itself is excluded from search engines in robots.txt.

## Keeping the answers checkable

- Ask the AI to give the article title and verified date for every fact.
- When you see “unverified” or “volatile”, confirm on the official source the article lists.
- Where the knowledge base has nothing, have it say so rather than guess.
- In a follow-up you can ask, “Using only the knowledge base, which conclusions have no source?” This is a suggested wording, not a feature of the site.

## Privacy

You can check and edit the text before pasting. It holds only what you entered in the form and the report, with no name, phone number or email, and the site does not store it. What your AI assistant does with the text you paste is decided by its provider, so please find out.

## Next steps

- Generate a report in the [planner](/en/planner/primary).
- To see where the scores come from, read [How we score](/en/kb/how-we-score).

## Sources
- The llms.txt proposal (Jeremy Howard, published 3 Sep 2024): https://llmstxt.org/
- This site’s source code (PromptBuilder, KbExportController, SeoController and lang/en/prompt.php), checked item by item for this article: https://github.com/arrow-town/edu