We Scanned 45 Singapore Business Websites. The Technical Basics Are Mostly Fine.
We expected to find a lot of broken Singapore SME websites. We did not. Here is what 45 real sites actually score, what is genuinely still missing, and why that changes where you should spend your effort.
- 01We scanned 45 Singapore SME websites across six industries. Median technical score: 92 out of 100.
- 02The basics are effectively solved. HTTPS, canonicals and contact details are near universal, so none of them differentiate anyone any more.
- 03The real gap is answer structure. 36% have no question headings or FAQ at all, and that is exactly what AI answers extract.
We built a free technical SEO checker earlier this year, so we pointed it at 45 real Singapore SME websites to find out what the average small business site actually looks like under the hood.
We expected a mess. We were wrong, and the way we were wrong is the useful part.
Method, and its honest limitation
On 19 July 2026 we ran 45 Singapore business websites through our own free checker, which scores 15 on-page signals across three groups: findability, AI and GEO readiness, and local signals. 42 returned a result. Three failed, one of them because the site blocked our request with a 403.
The six industries: aircon and home services, physiotherapy, renovation and interior design, law firms, dental clinics, and accounting and corporate services.
The limitation you should hold onto while reading: we collected these sites from the first page of Google for ordinary buyer queries, things a real customer would type like "aircon servicing singapore". That means every site here is already visible enough to rank. This is a sample of what your visible competitors look like, not a sample of the average Singapore SME. The true average is certainly worse.
That bias is exactly why the results are interesting.
The headline numbers
| Measure | Result |
|---|---|
| Sites scanned | 42 |
| Median score | 92 / 100 |
| Mean score | 88 |
| Range | 45 to 100 |
| Scored 90 or above | 25 of 42 (60%) |
| Scored below 70 | 4 of 42 (10%) |
Six in ten of the businesses you are competing against for a page-one buyer query have essentially clean technical setups.
The full dataset, including per-industry medians and the pass rate for every individual check, is published as a live benchmark page that you can compare your own score against.
What is already solved
These are the things people still sell as "SEO audits", and in this sample they are close to universal:
- HTTPS: 100% pass. Every single site.
- Canonical tags: 100% pass.
- Contact details on the page: 98% pass. Singapore SMEs put their phone number up. Local signals scored a mean of 93.
- Content depth: 98% pass.
- Structured data: 93% pass. Only 3 sites of 42 had no JSON-LD at all.
If your SEO proposal leads with "we will add SSL and canonical tags", it is describing work your competitors finished years ago.
What is genuinely still missing
| Gap | Sites affected |
|---|---|
| No question headings or FAQ content | 36% |
| No og for link previews | 36% |
| H1 problems (missing, or several on one page) | 31% |
| Meta description too long, too short or absent | 24% |
| Title tag problems | 19% |
| robots.txt or sitemap problems | 10% |
The two that actually cost you something are the first and the fourth.
The finding that matters: nobody is writing answers
36% of these sites have no question-style headings and no FAQ content anywhere on the page. Not a schema problem. There is simply nothing on the page shaped like a question a customer would ask.
Broken down by industry, the pattern is sharp:
| Industry | Sites with no Q&A content |
|---|---|
| Law firms | 3 of 4 |
| Dental clinics | 5 of 10 |
| Physiotherapy | 3 of 8 |
| Accounting | 3 of 6 |
| Renovation and interior design | 1 of 7 |
| Aircon and home services | 0 of 7 |
The aircon companies have solved this completely. Every one of them answers "how much does aircon servicing cost" somewhere on the page, because that is the question every customer asks and they have been competing on it for years.
The law firms have almost entirely not. Neither have half the dental clinics.
Why this is the expensive gap: AI answers do not rank your page, they extract a passage from it. A model looking for the answer to "how much do braces cost in Singapore" needs a passage that answers that question directly. If your page is a wall of practice-area prose with no question anywhere in it, there is nothing clean to lift. The research backs this up: across 75,000 AI answers and over a million citations, listicles and articles are the most-cited formats, because both are structurally full of extractable answers.
This is also the cheapest gap to close. It is writing, not engineering.
Meta descriptions are now an AI problem, not just a Google one
24% of the sample had a meta description that was missing, truncated or padded past the useful limit.
That used to be a minor snippet issue. It is worth more now. Investigations into ChatGPT's retrieval layer suggest that what the model reads for many results is the description, not the full page. If that holds, an over-long meta is not a cosmetic problem, it is your value proposition getting cut off before the model ever sees it.
We found this on our own site while running this study. Our homepage description was 187 characters. We cut it to 141.
The one thing nobody has, that does not matter
74% of the sample had no llms.txt file. Ignore that number.
There is no public evidence that any major AI search system reads third-party llms.txt files. Google's John Mueller has compared the file to meta keywords. We wrote up the evidence separately. Our own checker gives it the lowest weight of any check and labels it unproven, which is why it barely moved these scores.
We mention it because llms.txt is currently being sold as a ranking lever, and this sample is a reminder that the sites already winning these queries do not have one.
What we would actually do with this
If your competitors are sitting at a median of 92, then technical hygiene is the price of entry, not an advantage. Getting to 92 stops you losing. It does not make you win.
- Get to technical parity, then stop. It is a weekend of work, not a retainer.
- Write the answers. Take the ten questions your customers actually ask, make them headings, and answer each one directly with a real number. A third of your competitors have not done this.
- Fix your meta descriptions to around 150 characters. Cheap, and it may be what the model reads.
- Then go and earn authority, because that is the part none of this touches. We wrote honestly about that here, including our own site being technically near-perfect and still sitting on page five.
Frequently asked questions
What is a good technical SEO score for a Singapore small business?
In this sample of 42 businesses already ranking on page one for buyer queries, the median was 92 out of 100 and 60% scored 90 or above. Treat 90+ as parity with visible competitors rather than as an achievement. Below 70, which was 10% of the sample, you are losing points your competitors are not.
What do most Singapore SME websites get wrong?
Answer structure. 36% had no question headings or FAQ content at all, which is the single most common real gap. Then link-preview images (36%), H1 problems (31%) and meta description issues (24%). The classic technical items, HTTPS, canonicals, contact details, were effectively universal.
Does my business website need an llms.txt file?
Probably not for search. No major AI search system has been shown to read third-party llms.txt files, 74% of the sites in this sample do not have one, and they rank fine. It is genuinely useful if you publish developer documentation. Otherwise publish it for optionality, not for citations.
How was this study run?
We ran 45 Singapore SME websites through rankr's free technical SEO checker on 19 July 2026 and 42 returned results. Sites were collected from page-one Google results for ordinary buyer queries across six industries, which biases the sample toward already-visible businesses. The checker scores 15 on-page signals and the same tool is free for anyone to run on their own site.