Only 1 in 6 Malaysian SME websites tells AI what and where the business is

Customers can now ask ChatGPT, Gemini, Perplexity or Google's AI answers which clinic, workshop or caterer to use. Those tools can only recommend a business they can read and understand. So we took 500 small and medium businesses from across Malaysia, in nine sectors, and checked their websites for the signals AI tools look for. Then we looked up each business on Google Maps and checked its Business Profile.

Raw data: download the results for all 500 sites (one row per site, anonymised) and the results by sector. Both are free to use with credit to One Search Pro and a link to this page. More about the files.

Part A: the websites

We gave each website one point for each of five basics: it does not block AI crawlers, it has LocalBusiness markup, it has an llms.txt file, it uses HTTPS with a mobile layout, and it has some Bahasa Malaysia content. The average score was 2.2 out of 5. Only one site of the 500 scored 5 out of 5.

1. Few sites tell AI what and where the business is

LocalBusiness markup is a short block of code that states a business's name, address, phone number and opening hours in a format machines can read. Google documents itas the way to tell Search about these details. Only 16.6% of the sites we checked have it, or one of its subtypes such as Restaurant, Hotel or Dentist. Most sites (63.4%) have some markup, but it is usually the generic kind a website theme adds, which says nothing about the location. Only 3.4% haveFAQ markup.

2. About 1 in 11 block AI crawlers, mostly without meaning to

AI companies use crawlers to read websites. OpenAI's are calledGPTBot, OAI-SearchBot and ChatGPT-User, and Perplexity andGooglepublish theirs too. A site can turn them away in itsrobots.txt file, or its firewall can refuse them.

8.8% of the sites block at least one AI crawler. Only 1.4% do it in robots.txt, where an owner would normally make that choice. The other blocks come from the firewall: 7.8% of sites refused our request when it identified itself as GPTBot, while letting a normal browser in. Security services offer this as a setting; Cloudflare, for example, has aone-click option to block AI bots. That pattern suggests many owners may not know their site is closed to ChatGPT. Automotive and beauty and wellness businesses block the most (14.3% each); food and beverage the least (1.7%).

One in five sites (19.8%) have no robots.txt file at all. That doesn't block anyone, but it means the owner hasn't set any rules for crawlers.

"Most of these gaps take an afternoon to fix. Check that your firewall isn't blocking ChatGPT, add your address and opening hours to your site as LocalBusiness markup, and write the description on your Google profile. Those three things tell AI what you do and where you are."

3. Hardly any SME websites speak Bahasa Malaysia

Only 6.4% of homepages have any Bahasa Malaysia content, and 6.2% offer a language switch. That matters: in our early look at AI citations (Part B below), the same question asked in English and in BM often led to different businesses. Healthcare had the most BM content (13.6%). None of the 35 beauty and wellness sites had any.

4. llms.txt is spreading, mostly by accident

llms.txt is a proposed standard: a plain-text file that gives AI tools a short guide to a website's key pages. Not every AI company has said it reads it. 16.8%of the sites (84) already have one, but 52 of those 84 were generated automatically by Shopify, Wix or a WordPress SEO plugin such as Yoast orRank Math, rather than written by the business.

5. HTTPS and mobile are no longer the problem

92.8% of sites use HTTPS and 96.4% have a mobile layout. The gaps are in the newer signals that matter for AI search, not the basics of a decade ago.

Part C: Google Business Profiles

We searched for each of the 500 businesses on Google Maps by name and state and found a matching profile for488 (97.6%). We then ran seven checks: the profile is claimed, and it lists opening hours, a phone number, a link to the business's own website, a business description, 10 or more photos and 10 or more reviews.

6. SMEs look after their Google profile far better than their website

84% of the businesses have claimed their profile, which lets the ownercontrol how it appearson Search and Maps. Compare that with 17% that have LocalBusiness markup on their own site. Most profiles list a phone number (94%), link to a website (88%) and show opening hours (82%). The median profile has a 4.6-star rating and 64 reviews.

7. But fewer than half have written a description

Only 46% have a business description, the one field where the owner explains in their own words what the business does. Only 1 in 5 (21%) pass all seven checks. One in four (26%) have fewer than 10 reviews. Professional services (median 8 reviews) and construction (median 10) have the thinnest profiles; food and beverage (median 661) and hotels (median 618) the richest. And 77 businesses have never claimed their profile, so the owner has no say in what Google shows.

Results by sector

Construction and engineering firms scored highest on the website checks (2.40 out of 5) and professional services lowest (2.14), but the spread is narrow: no sector is ready. Hotels are the most likely to have a profile description (76.3%), but few list opening hours (10.2%) and only 6.8% pass all seven profile checks.

Beauty and wellness (35 sites) and automotive (49) had fewer eligible businesses than the other sectors, so their figures are less precise. Results by state are in the per-site file; states outside Kuala Lumpur, Selangor, Sarawak, Johor and Penang have fewer than 40 sites each, so treat them as indicative only.

What SMEs can do

• Check that your firewall or security plugin isn't blocking AI crawlers such as GPTBot and OAI-SearchBot, unless you have decided to block them.
• Add LocalBusiness markup with your name, address, phone number and opening hours.
• Claim your Google Business Profile, write the business description and keep hours and photos up to date.
• If your customers search in Bahasa Malaysia, put real BM content on your site, not just a translated menu.
• Ask the AI assistants the questions your customers ask, in both languages, and see who they recommend.

If you want AI search to find and recommend your business, our AI search SEO andlocal SEO teams can help.

Methodology

• Sample: every business in OpenStreetMap(Malaysia) that lists a website, pulled on 13 October 2026 through theOverpass API: 10,558 listings. We removed chains, big brands, government bodies, schools, NGOs, places of worship, malls and supermarkets, and links to social media or marketplaces, leaving 2,373 websites. A manual review then removed listed companies, multinationals, government-linked companies, insurers, banks, franchise chains and property sales galleries.
• Selection: 500 live websites across 9 sectors, with an equal quota per sector (sectors with too few sites passed their places to the others) and sites drawn across states with a fixed random seed. The selection never looked at the check results.
• Website checks (13 October 2026): robots.txt rules for GPTBot, OAI-SearchBot, ChatGPT-User, Google-Extended, PerplexityBot, ClaudeBotand CCBot; a homepage request as GPTBot and PerplexityBot, where a 401 or 403 error for the bot when a normal browser gets through counts as a firewall block; LocalBusiness (or a subtype) and FAQPage markup on the homepage, with generic Organization markup not counted; an llms.txt file that returns plain text, with each one tagged by what generated it; HTTPS and a mobile viewport tag; and Bahasa Malaysia text on the homepage, a lang="ms" tag or a BM language switch.
• Readiness score: one point each for not blocking AI crawlers, LocalBusiness markup, llms.txt, HTTPS plus a mobile layout, and Bahasa Malaysia content (0 to 5).
• Google Business Profiles (14 October 2026): each business searched on Google Maps as "name, state, Malaysia" with the Apify Google Maps scraper, top result only. A result counted as the business when its website matched the sampled site (375 profiles) or when at least half of the distinctive words in the names matched (113). Using only the 375 website-confirmed matches gives similar results: 88% claimed and 51% with a description.
• AI citations (Part B, preliminary): 25 local questions covering all 9 sectors and 15 towns, each in English and Bahasa Malaysia, run through SE Visiblewith a Malaysian location. The tool had processed about 84% of the English and 80% of the BM answers when we pulled the data.

Caveats

• OpenStreetMap listings lean towards places mappers visit and businesses with a web presence. The results describe SMEs that have a website, not all SMEs.
• We checked one homepage per site; inner pages may differ.
• Scans ran from a cloud server outside Malaysia. Some sites may block foreign or data-centre traffic, and response times are indicative only.
• Google Maps matching by name can pick a neighbouring branch or a business with a similar name.

Download the raw data

Every number on this page comes from these files. You are welcome to use them: please credit One Search Pro and link to this page. To protect the businesses in the sample, the files have no business names, domains or URLs. Each site has an ID, its sector and state, and its check results. Review counts are shown in bands.

• Results per site (CSV)500 rows, one per website: sector, state and every check result, plus the Google Business Profile checks. Anonymised: no names, domains or URLs.
• Results by sector (CSV)9 sectors plus the overall total, with every percentage quoted on this page.

Media enquiries: journalists can use these findings with credit to One Search Pro and a link to this page. For more detail, sector or state breakdowns, or an interview with Eric Ang, founder of One Search Pro, pleasecontact us.