You have probably been told that people now search in full questions, so your pages should be rewritten as answers for AI. We counted the question queries in our own Search Console to see if that holds for a real Singapore site.

5 of 300: that is how many of the queries bringing singrank.com the most Google impressions between 2 July and 29 September 2026 were written as questions. The other 295 were plain keywords such as geo agency singapore and seo pricing singapore. We fixed the counting rule in writing before pulling the data, and the full list is free to download.

But three of the five were full sentences ending in a question mark, and all three sat on page one, at positions 6.8 to 9.7. The keyword queries had a median position of 43.5. That gap, and Google's explanation of why so few questions show up, is what matters for planning pages.

Key takeaways

5 of 300 top queries were questions. They carried 194 of 28,762 impressions, which is 0.67% (see the dataset).
3 of 5 questions ranked in the top 10. Only 7 of 295 keyword queries did.
The three full-sentence questions averaged position 7.6 on 78 impressions. Keyword queries averaged position 46.3, weighted by impressions.
0 clicks came from any question. Our 300 queries produced 3 clicks in total, and the whole site produced 48 clicks in the window.
The median query was 3 words long. 259 of the 300 queries were 2 to 4 words.

How did we count question queries in Search Console?

We used singrank.com's own Google Search Console figures for 2 July to 29 September 2026: the 90 days ending on the latest date in our reporting system when we ran it on 2 October. The site recorded 32,822 impressions and 48 clicks over the 89 days with data. The top 300 queries by impressions hold 28,762 of those impressions, or 87.6%, but only 3 of the 48 clicks (every query is in the dataset). No client data is used.

Before looking at any query, we wrote the rule into a short script. A query counts as a question if it contains a question mark, starts with a question word such as how, what, who, which or are, or contains a phrase such as how much or how to.

A query of 8 or more words without those markers counts as long. Everything else is a keyword, such as seo agency singapore or search engine optimization pricing, both rows in the dataset. Search-operator queries, such as site: checks from tools, were set aside; the top 300 had none.

The result: 5 of 300, which is 1.7% of rows and 0.67% of impressions, all reproducible from the dataset. No query was 8 or more words long without also being a question. We wrote down in advance what a result below one query in 20 would mean: keyword pages stay the main job, and answering questions is a habit, not a rebuild. That is where this lands.

Group	Queries	Impressions	Clicks	Position (impression-weighted)	Median position	In top 10
Questions	5	194	0	25.7	9.7	3
of which full sentences with "?"	3	78	0	7.6	7.3	3
of which short "how much"	2	116	0	37.9	44.4	0
Keyword-form	295	28,568	3	46.3	43.5	7
of which 3 words or fewer	172	19,424	2	46.7	40.9	7
All 300	300	28,762	3	46.2	43.0	10

The split into full sentences and short how much questions is ours, made after we saw the numbers.

Where did the questions rank, and what do they look like?

A young site cannot win the keyword queries yet. Our biggest, geo agency singapore, drew 1,162 impressions at an average position of 49.65. That is normal for a site whose oldest article is dated 11 June 2026, as our study of impressions without clicks on our own site shows. The full-sentence questions were already on page one:

Query	Impressions	Position
how much does seo cost in singapore	90	32.66
who offers geo marketing for singapore smes?	43	6.81
how much does it cost for seo	26	56.12
are there any singapore specialists for aio and aeo implementation?	18	9.67
how do aeo agencies in singapore typically price and structure their engagements?	17	7.35

Google defines average position as a relative ranking where 1 is the top result, averaged across the times your link appeared.

Our reading, which the numbers don't prove: a long question names several things at once, such as a service, a city, a business size and a buying concern. Few pages match all of them, so a small site that does can rank where it never would for a two-word phrase. The two short how much questions behave like keywords and sat at 32.66 and 56.12, which fits that reading.

Three things stand out in these conversational search queries:

They ask for a provider, not a definition. Who offers, are there any specialists. No question in our top 300 asked what GEO or AEO means.
They carry qualifiers. Singapore smes, price and structure, implementation. Each word narrows who the right answer is.
They read like messages. All three end in a question mark and one is 12 words long. Search Console does not say whether a person typed them, pasted them from a chat tool, or a tracking tool sent them, so we do not claim they came from Google's AI Mode.

Other markets show the same shape. In our preschool search study, 43 of 112 suggestions asked what it costs.

Why does Search Console show so few questions?

Google says anonymised queries are searches typed by fewer than a few dozen people across a two-to-three-month span. They are hidden from the query table but still counted in the chart totals.

A 12-word question is likely to be rare, so real questions are the ones most likely to be missing from your list. Our 300 listed queries carry 87.6% of impressions but only 3 of 48 clicks, as the dataset shows. Read 5 of 300 as a floor, not every question asked.

Google's AI Mode announcement describes AI Mode splitting a question into subtopics and running many searches at once, which it calls query fan-out, and says people are asking "more complex, longer and multimodal questions". Its guide to AI features says AI Overviews may use the same technique, that sites shown in AI features are counted in Search Console's Web search type, and that no special optimisation is needed. Our AI Overviews guide explains what that means in practice, and where AI visibility demand actually shows up covers the buyer's side.

What should a Singapore business do with this?

Our 3 clicks came from seo agency, geo agency in singapore and konsultan pemasaran batam; the three page-one questions drew 78 impressions and no clicks.

So nothing here shows that answering questions brings enquiries. It shows a young site can be visible for specific questions first. For a view across many sites, our Singapore search visibility study covers 59,570 query-months of Search Console data. We will publish the update either way.

If you run a Singapore business and write pages for Google and AI answers:

Keep your keyword pages. 28,568 of the 28,762 impressions in our dataset came from keyword-form searches. A page called aircon servicing Tampines still matches how almost all our searchers phrased things.
Put the specific questions inside those pages. Answer who does this for a small business here, how it is priced and what is included, in plain sentences near the top.
Answer the provider question directly. Say what you do, for whom, and where, in the page's first lines.
Check your question queries in Search Console once a quarter. In the Performance report, filter for question words and a question mark. Track the position of those, not just impressions, as our rank tracking guide explains.

Five questions is a tiny sample. It is still worth acting on, because steps 1 and 2 cost an afternoon and leave what ranks alone. A full rewrite into question pages would reshape your site around the 194 impressions and away from the 28,568.

Want us to sort your queries?

Send us view-only access to your Search Console, or an export of your top 300 queries, through our contact page. We will run the same script and tell you, in writing, how many of your question queries Search Console shows and where they rank. If you then want the pages written and measured, that is our SEO, AEO and GEO programme.

Method

Source and frame: singrank.com's own Google Search Console property, read on 2 October 2026, window 2 July to 29 September 2026 (89 days with data). The 300 queries with the most impressions, from a pull of 500. One row is one distinct query.
Rule, written before collection: question = question mark, a question word at the start (English or Indonesian), or a phrase such as how much, how to, which, who or why; long = 8 or more words and not a question; keyword = the rest. Script and plan were saved with checksums first.
Hand check: all 5 questions (5 genuine), all 11 keyword queries of 6 or 7 words (0 questions), 30 random keyword queries (0 questions), then the full list (0 missed). No counts changed.
Exploratory, after collection: the full-sentence versus how much split; rows 301 to 500 of the pull held 5 more questions and 1 long query.
Redaction: five queries naming another business or tool read [business] in the public file.

The research hub lists our other studies.

Limits

One young site, ours, with few clicks. A site with ten years of rankings may see a very different mix.
Search Console hides rare queries, so questions are undercounted by design. This is a floor.
Average position varies by time, place, device and the searcher's history.
We cannot tell who or what typed a query, so we make no claim about AI Mode, voice search or chat tools.
Impressions are appearances of our pages, not demand in the market.
Five questions are too few for any statistical test. The position gap is a description, not proof.

Cite this

SingRank, "Question Queries in Search Console: 5 of Our Top 300", Khuzayfah Redo, data collected 2 October 2026 from singrank.com's Google Search Console (2 July to 29 September 2026), published on singrank.com. Dataset: singrank-question-queries-census-2026.csv, version 1, one anonymised row per query. Please quote each figure with its base, such as 5 of 300.

Figures are singrank.com's own Google Search Console numbers for 2 July to 29 September 2026 (dataset). No client data appears on this page.

FAQ

How many search queries are written as full questions?

On our own site, few: 5 of the top 300 queries, carrying 194 of 28,762 impressions. Google hides rare queries from Search Console tables, so the true share is likely higher than any site can see.

Do question queries rank better than keyword queries?

On our site they did, as a description, not a proven effect. Three full-sentence questions sat at positions 6.81 to 9.67; keyword queries had a median position of 43.5. Five questions are too few to test this statistically.

Are these queries from Google AI Mode?

We cannot tell. Google's AI features guide says AI feature appearances are counted in the same Search Console report, but the query list carries no label saying which surface a query came from.

Should I rewrite my service pages as questions?

We would not. 28,568 of 28,762 impressions came from keyword-form searches, and Google says AI Overviews and AI Mode need no special optimisation. Keep the page shaped around the service and add short answers to the questions buyers ask.