For Malaysian loss adjusters · A measurement, dated, re-run monthly
A factory burns at midnight. The shortlist is already written.
On 11 September 2026 we asked two AI assistants who should adjust a major industrial fire in Malaysia. We asked three times, in different words, across ChatGPT and Perplexity. Every firm either of them placed in the top tier was foreign-owned. The Malaysian firms were in the answer, underneath, in the part you reach once the recommendation has already been made. This page is what we measured, how we measured it, and what decides which firms a machine can read at all.
Measured 11 September 2026Two engines, three runsNo firm is named here
The short version
Ask an assistant who to appoint and it answers in four seconds with a shortlist, then explains itself. In all three of our runs the names at the top were international firms. Perplexity went on to list Malaysian firms in a section of its own beneath them; ChatGPT built a ranked table, placed its first Malaysian name at five, and wrote in its own words that the first four are the ones I’d be most interested in interviewing. The engines were fair about it. Every international firm they named is on Bank Negara’s register and does this work here. What decided the order was something simpler: how much each firm had published that a machine could actually read.
ChatGPT, asked as a policyholder whose factory had burned
1234No Malaysian firm anywhere in the answer
ChatGPT, asked which adjusters are best for a large industrial claim
12345it said it would interview the first four
Perplexity, the same question, asked independently
1234then, in a section of its own further downfive Malaysian firms, unranked
Every position one to four, in all three runs, went to an internationally owned firm. Measured 11 September 2026.
The run, 11 Sep 2026
What came back at the topPositions one to four
Where a Malaysian firm first appearsSame answer, further down
ChatGPT, asked as a policyholder whose factory had burned
Four firms, every one of them international.
Absent from the answer entirely.
ChatGPT, asked which adjusters are best for a large industrial claim
Four firms, every one international, in a ranked table.
Position five, below the four it said it would interview.
Perplexity, the same question, asked independently
Four firms, every one international.
A section of its own further down, five firms in it.
Three runs, two engines, one morning. The queries are printed in full in section C so anyone can repeat them.
A
The engines said what they were reading, and it is worth quoting
A shortlist that favours international firms invites an easy explanation, and the easy explanation turns out to be the wrong one. ChatGPT volunteered the point itself, unprompted, in the middle of its own answer:
ChatGPT, 11 September 2026, verbatim
Bank Negara’s current register confirms all of the major firms above are licensed/registered adjusters in Malaysia, so this isn’t simply a list of international names pulled from overseas.
Returned in response to: “Who are the best loss adjusters in Malaysia for a large industrial property claim?”
So the firms at the top hold the same licence, sit on the same public register, and take instructions from the same insurers as the firms underneath them. The register is public and the engines read it. What separated the two groups was everything around that register entry: how much each firm had published under its own name, in a form a machine could parse, saying what it does and what it has handled.
B
Six things that decide whether a machine can read your firm
Every one of these is visible from outside the firm, and every one is checkable on your own site this afternoon.
01
Your firm’s own name is the page title
The title is the first thing a crawler stores and the line a result prints. Where a group holds several companies behind one front page, the firm sharing that door carries another company’s name into every answer about it.
02
One page per capability, in the words a buyer uses
Business interruption, machinery breakdown, consequential loss and catastrophe response are four separate appointments. Set together in one run of words, they read as a single undifferentiated service.
03
The record, in figures, on the page
Years in practice, the panels you sit on, the number of claims handled after one catastrophe. A machine repeats what it can quote, and a sentence carrying a number travels further than a sentence of adjectives.
04
Structured data that names the entity
A short block of machine-readable markup giving the legal name, the register, the offices and the year of incorporation. It is the difference between a page a machine reads and a firm a machine recognises.
05
The same firm, described the same way, everywhere else
The register, the association listing, LinkedIn and the firm’s own site should each say one identical thing. Where they differ, an engine treats every version as weaker evidence and reaches for a source it trusts more.
06
A privacy notice that reflects the amended PDPA
You hold claimants’ personal data by the nature of the work. The 2024 amendment brought breach notification and data protection officer duties into force, and the notice on your site is where a corporate client looks first. What the amendment changed.
C
How to repeat this yourself, in about four minutes
A measurement is worth what its method is worth, so here is all of it. Open ChatGPT or Perplexity in a fresh chat, in a temporary session so your own history stays out of it, and ask:
The queries, verbatim
Our factory in Malaysia had a major fire and we need to appoint a loss adjuster. Which firms should we consider?
Who are the best loss adjusters in Malaysia for a large industrial property claim?
We asked the first once and the second on both engines, each in its own fresh chat.
Read the order the firms come back in, then read what the answer cites underneath it. An assistant’s answer moves between days and between phrasings, which is why one run proves little and three runs across two engines begin to mean something. We re-run these monthly and date the result. Where your firm’s name appears, the useful question is what the engine was quoting when it said so.
Malaysian loss adjusters
Send me your firm’s site and I will tell you what a machine can read on it.
A short written read of what an assistant can see when your firm comes up, and what would have to change for it to say more. Free, and nothing attached to it.
Do AI assistants recommend loss adjusters in Malaysia?
Yes. Asked who to appoint after a major industrial fire, both ChatGPT and Perplexity returned a shortlist of named firms within seconds, along with the register entries and association listings they had drawn on. Measured 11 September 2026 across three runs.
Why do international adjusting firms appear first?
In our runs it followed from how much each firm had published under its own name in a form a machine could read, rather than from licensing or capability. ChatGPT said as much itself, noting that the firms it named sit on Bank Negara’s register as adjusters in Malaysia.
How do I check what AI says about my adjusting firm?
Open a fresh chat on ChatGPT or Perplexity and ask who to appoint for a large industrial claim in Malaysia, using the queries printed in section C above. Read the order, and read what the answer cites. Repeat it monthly, because an assistant’s answer moves between days.
Who builds websites for loss adjusters in Malaysia?
Upcial is a boutique, SSM-registered web design studio in Penang, Malaysia, building high-trust, PDPA-compliant, BNM-aware websites for insurance brokers, reinsurance brokers, loss adjusters and financial services firms. We design your homepage and show it to you before anything is signed. What that costs.
Does a loss adjuster need a privacy notice on its website?
An adjusting firm handles claimants’ personal data as a matter of course, which makes it a data user under Malaysia’s PDPA as amended in 2024. The notice on the site is where a corporate client looks first. What the amendment requires.