Skip to content
// method

How every answer
is made.

AI helps draft the answers on FAQs.HK, and we say so on the page. Every claim is checked against the source it cites, and nothing publishes until a named editor has signed off. Once an answer is published, its sources are never silently removed.

That is the whole promise: not confident, unsourced text scraped from the open web, but a reference you can check and trace back to its origin.

editor-approved every claim cited AI disclosed by name
// our roots

20+ years of Hong Kong questions, asked by real people.

FAQs.HK comes from the team behind GeoExpat, GeoBaby and GeoSchools, the Hong Kong communities where expats and parents have asked, answered and argued about daily life for more than 20 years. Those databases are a record of what people here actually need to know: visas and tax, flats and leases, schools and admissions, pregnancy, childcare and everything in between.

That record shapes what we write. The questions we answer come from it, and the human experience in it tells our editors where the official line and real life part ways. But a forum post is not a source: every fact in an answer is still checked against, and cited to, the primary source.

GeoExpat

The Hong Kong expat community: moving here, working here, living here.

GeoBaby

The Hong Kong baby and parenting community.

GeoSchools

Schools and education in Hong Kong. Returning as a read-only archive.

// the pipeline

Four steps from question to answer

AI speeds up the drafting. Verification and a human audit decide what gets published.

  1. 01

    Asked

    human

    Every entry starts with a real question from a reader or an editor, not a keyword chasing search traffic.

  2. 02

    Drafted

    AI

    AI research assistants read the primary sources and draft an answer, with every claim tied to its source.

  3. 03

    Verified

    AI + checks

    Every draft is stress-tested before anyone relies on it. Each claim is checked against its source, every link is opened, and anything unsupported is cut.

  4. 04

    Approved

    human editor

    A named editor reads the answer against its sources and signs off. Nothing publishes without that.

// guarantees

What you can trust

editorial_guarantees.txt
  • ✓ Every answer is dated — you see when it was written and when it was last revised.
  • ✓ Every answer is sourced — each load-bearing claim links to its origin.
  • ✓ Every answer is authored — a named human editor stands behind it.
  • ✓ AI disclosure is mandatory — an answer drafted with AI cannot publish without the byline.
  • ✓ Corrections are logged — sources only grow; revisions are recorded, never silent.
// follow a subject

Follow the subjects, not the staff

You don’t follow individual writers here — you follow subjects: the categories and topics you care about. When an editor publishes or revises an answer in a subject you follow, it surfaces for you. Bylines are name-only — enough to know who stands behind an answer, without turning the site into a personality contest.

The unit of trust is the answer and its sources, not a profile. That keeps the focus where it belongs: on questions getting answered properly, and staying answered as things change.

// ai_policy

Our AI assistance policy, in plain language

We use AI as a drafting and editing tool, the way an editor uses a research assistant — never as the author of record. Here is exactly what that means.

// how we use it
  • + Drafting a first pass an editor then rewrites and verifies.
  • + Tightening and restructuring a draft once its sources are checked.
  • + Surfacing candidate sources — which a human then reads and confirms.
// what we never do
  • − Publish AI text unread, unedited, or unverified by a person.
  • − Let AI invent sources, or cite a source no human has checked.
  • − Hide that AI was involved. If it helped, the byline says so.

The rule, in one line: AI helps draft, every claim is checked against its source, and a named editor signs off, in the open.

// corrections

When we get it wrong

Every answer carries a “report an error” link. When you report one, it’s reviewed by the assigned editor against the sources — not waved through, and not quietly buried. Corrections are footnoted on the entry, never silent: the entry shows what changed and when, and the source trail only ever grows.

Is this just ChatGPT output?
No. AI helps draft, but every claim is checked against the source it cites, and a named editor reviews the answer and signs off before it publishes. Where AI helped, the answer carries an Editor-verified · AI-drafted byline so you know exactly what you’re reading.
Can I suggest a question?
Yes, please. If there’s something we haven’t answered, tell us — an editor will pick it up, research it, and cite it. Suggest a question →
How do I report an error?
Every answer has a “report an error” link. Flag it and the assigned editor reviews it against the sources. Corrections are footnoted on the entry with a revision date — we fix things in the open, never silent, and the source trail only ever grows.
Why do you cite so much?
Because a claim you can’t trace is just an opinion. Citations let you verify the original yourself — and they keep us honest. If we can’t source it, we don’t say it.
// your turn

Have a question we haven’t answered?

Suggest it. An editor will research it properly, cite the sources, and publish it in the open — like everything else here.

Suggest a question Browse the FAQs →