Small-site AI visibility · Specificity over scale
The smallest property here has the highest AI visibility rate
This site earns 1,952 clicks — a rounding error next to the rest of the portfolio. And 8.1% of its search impressions come from Google AI features, five times the rate of our largest property. It also holds 200 of 1,000 queries in the top three. The variable is not size. It is how specific the question is.
Read this before the numbers
This study does not meet our own evidence standard in full
Every other case study here carries dated platform screenshots showing the property, the filter and the window, plus Bing organic and Copilot citation data. This one has neither. No dashboard screenshots were captured for this property, and no Bing Webmaster Tools data was exported.
What it does have is the raw Google Search Console CSV exports, which are the primary source — the screenshots elsewhere are corroboration of exports like these, not a separate measurement. So the figures below are sound, and the finding stands. But the coverage table marks both gaps not audited, and we would rather tell you that at the top than let you notice it at the bottom.
8.1%
of impressions from AI features — highest we measure
200
of 1,000 queries averaging a top-3 position
124
pages appearing in Google AI features
4.92%
mobile CTR, at an average position of 9.89
1,952
organic clicks from 71,572 impressions
Property
MakeMoneyOnline.lk · owned and operated
Type
Online income methods, written for the Sri Lankan market
Window
16 months, ending 14 August 2026
Site size
271 pages ranking in Google
Evidence
Search Console CSV exports only — see the note below
Paid promotion
None
The finding
AI visibility rate runs opposite to site size
Across the properties where both figures exist, we can measure AI-feature impressions as a share of ordinary search impressions. It is the cleanest available proxy for “how likely is an answer engine to use this site”, because it normalises away scale.
The result is the inverse of what anyone would assume. The property with 3.91 million impressions sits at 1.5%. The property with 71,572 impressions sits at 8.1%.
AI-feature impressions as a share of all search impressions
Google AI feature appearances divided by total Google search impressions, per property, over the same 16-month window. Sorted by rate — note that the ordering is almost exactly the reverse of the ordering by size.
Source: Google Search Console, Generative AI features export and Performance export, per property, exported 12–14 August 2026. VelankanniShrine.in is absent because Search Console exposes no Generative AI report for it. Google's AI report only began collecting data on 18 May 2026, so all five rates are computed on the same partial-window basis and are comparable to each other, not to a full 16 months.
The mechanism is competition for the source slot, not a preference for small sites. An AI answer cites two to five sources. The question “how do I earn money online in Sri Lanka without investment” has very few credible pages behind it — so a small, specific site can be one of the handful. The question “Coimbatore bus routes” has a municipal transport authority, three newspapers and Google Maps.
The second finding
200 of 1,000 queries in the top three, on a site this size
The same specificity that earns the citations earns the positions.
Where 1,000 the top 1,000 exportable queries actually sit
80 queries average position #1 and 200 average the top 3 — 20% of the measured set, the second-strongest positional profile in the portfolio, on the smallest property in it. Search Console caps its export at 1,000 rows.
200
Top 3 · 20.0% of the queries measured
472
Page one · 47.2% of the queries measured
Source: Google Search Console, makemoneyonline.lk, Search type: Web, 16 months ending 14 August 2026. Band thresholds are ours: #1 is an average below 1.5, since Search Console reports position as a decimal average rather than an integer rank.
Top queries by clicks · 16-month averages
| Query | Clicks | Impressions | CTR | Avg position |
|---|---|---|---|---|
| earn money online sri lanka | 44 | 990 | 4.44% | 11.21 |
| earn money online sri lanka without investment | 32 | 400 | 8.00% | 7.26 |
| sri lanka money earn website | 29 | 230 | 12.61% | 5.85 |
| earn money online sri lanka sinhala | 28 | 313 | 8.95% | 6.9 |
| online earning websites in sri lanka | 26 | 307 | 8.47% | 6.16 |
| daraz affiliate program sri lanka | 18 | 927 | 1.94% | 6.24 |
| print on demand sri lanka | 13 | 213 | 6.10% | 9.64 |
| reselling business sri lanka | 12 | 89 | 13.48% | 9.97 |
Every one of these carries “sri lanka” or a local platform name. None of them is a query an international site would target, and that is exactly why they are winnable.
Evidence · AI features
5,795 appearances across 124 pages
Concentrated on the homepage, but with a real tail — 123 other pages appear at least once.
Google AI feature appearances per page
Top 9 of 124 pages. The homepage takes 44% — this is a “recommend me a way to earn” market, so the entity page absorbs most of it — but the method pages behind it each pick up their own specific question.
5,983
AI appearances across all 124 pages, of which the top 9 are charted
Source: Google Search Console Generative AI features, Pages export, 14 August 2026. This export sums to 5,983 while the Devices export sums to 5,795; Google attributes the two breakdowns slightly differently and both are published here.
AI appearances by device
| Device | AI appearances | Share |
|---|---|---|
| Mobile | 2,976 | 51.4% |
| Desktop | 2,713 | 46.8% |
| Tablet | 106 | 1.8% |
Almost an even split, on a site where 74% of organic clicks are mobile. AI answers are being served to a more desktop-weighted audience than the blue links are.
AI appearances by country
| Country | AI appearances | Share |
|---|---|---|
| Sri Lanka | 5,458 | 94.2% |
| All other countries | 337 | 5.8% |
The tail runs to 63 countries — the Gulf, Japan, Singapore, Italy — but at one to sixty appearances each. This is a single-market property and the AI data agrees.
The constraints
The most polluted category on the internet, in one small market
The category is a scam magnet
“Make money online” is one of the most polluted topics on the internet, and the Sri Lankan version doubly so — pyramid schemes, fake job boards, cloned survey sites. Any page has to be credible in a category where the reader arrives already suspicious.
The income numbers have to be real and local
A global “earn $5,000 a month” page is useless and insulting to somebody in Colombo. Figures have to be in rupees, achievable in this market, and honest about what the low end looks like — which rules out most of the affiliate content that dominates the category.
The audience is on a phone, and on data
74% of clicks are mobile, from a market where mobile data is metered. Long, slow, ad-heavy pages are not a bad experience here, they are a cost the reader pays.
What was actually built
Six decisions, and the one that matters most is the last
A page per method, not a listicle
Selling books, print on demand, Amazon Associates, stock photos, voice-over work, vehicle rentals, data entry. Each earning method has its own page with its own realistic numbers — which is why 124 separate pages now appear in Google AI features rather than one.
Local platforms, not just global ones
Daraz, the dominant Sri Lankan marketplace, has its own affiliate guide. “Daraz affiliate program sri lanka” is a query no international site is going to answer well, and it converts at position 6.24.
Rupees, ranges and the honest low end
Income figures in LKR with realistic ranges, plus explicit warnings about the scams that dominate the category. In a topic this polluted, being the source that says “this one probably will not work for you” is what earns the trust the rest of the page needs.
Written in English, for Sinhala speakers
The audience is Sinhala-speaking and searches in English — “earn money online sri lanka sinhala” is itself an English query asking for Sinhala content. Only four Sinhala-script queries appear in the entire export, and they earned zero clicks. Assuming vernacular search here would have been a costly mistake.
Built light for metered data
Mobile carries 74% of clicks at a 4.92% click rate against desktop’s 1.20%. A page that loads fast on a phone on a metered connection is not a technical nicety in this market, it is the product.
Specific enough that nobody else is answering
Not “make money online” — “earn money online in Sri Lanka without investment”. The narrower the question, the fewer credible sources exist, and the more likely an answer engine is to use yours. That is the whole finding of this study.
A finding that contradicts our own portfolio
Vernacular search matters enormously — except when it does not
On our matrimony property, the Sinhala term mangala yojana converts at
19.4% — better than any English category term on that site, and a finding
we called out as the thing an English-only keyword tool would never surface.
On this property, in the same country, for the same language: four Sinhala-script queries in the entire 1,000-row export, and they earned zero clicks between them.
Meanwhile earn money online sri lanka sinhala — an English query
asking for Sinhala-language content — earns 28 clicks at 8.95%.
We are including this because it undercuts a rule we stated confidently on another page in this portfolio. A rule that holds on one property and fails on the next is not a rule, it is an observation — and treating it as a rule is how agencies end up recommending a Sinhala content programme to a site whose audience would never have used it.
Who it reaches
Eight positions better on mobile than on desktop
By device · Google
| Device | Clicks | CTR | Avg pos |
|---|---|---|---|
| Mobile | 1,440 | 4.92% | 9.89 |
| Desktop | 495 | 1.20% | 18.44 |
| Tablet | 17 | 1.44% | 9.77 |
Mobile at 4.92% from position 9.89; desktop at 1.20% from 18.44. A single blended position of about 13 would have described neither, and would have hidden that desktop is the weak surface worth fixing.
By country · Google
| Country | Clicks | CTR | Avg pos |
|---|---|---|---|
| Sri Lanka | 1,876 | 3.74% | 13.82 |
| Qatar | 11 | 6.79% | 8.09 |
| Japan | 10 | 4.00% | 19.18 |
| United Arab Emirates | 8 | 1.65% | 26.03 |
| Australia | 5 | 1.34% | 38.37 |
| Singapore | 4 | 2.29% | 7.75 |
96% of clicks are Sri Lankan. Qatar converts at 6.79% — the migrant-worker pattern we saw on the matrimony property, on a much smaller base.
Top pages by clicks · Google, 16 months
| Page | Clicks | Impressions | CTR | Avg position |
|---|---|---|---|---|
| / | 1,198 | 28,848 | 4.15% | 11.36 |
| /method/selling-books/ | 43 | 1,218 | 3.53% | 8.55 |
| /daraz-affiliate-program/ | 39 | 1,773 | 2.20% | 6.62 |
| /method/print-on-demand/ | 37 | 1,573 | 2.35% | 9.85 |
| /method/amazon-associates/ | 35 | 1,778 | 1.97% | 7.68 |
| /data-entry-jobs/ | 28 | 803 | 3.49% | 15.44 |
| /method/selling-stock-photos/ | 28 | 826 | 3.39% | 7.25 |
The homepage takes 1,198 of 1,952 clicks. Behind it, a long tail of method pages each doing 20 to 45 — which is what a page-per-method architecture looks like in the data.
Coverage, stated honestly
Two documented surfaces, three gaps
This is the thinnest evidence base in the portfolio and the table says so.
| Surface | Status | Finding | Evidence |
|---|---|---|---|
| Google organic | Documented | 1,952 clicks from 71,572 impressions at 2.73%. 80 of the top 1,000 exportable queries average position #1 and 200 average the top 3 — the second-best positional profile in the portfolio. | Search Console CSV exports |
| Google AI features | Documented | 5,795 appearances across 124 pages — 8.1% of the property’s total search impressions, the highest rate of any property we measure. | Search Console Generative AI exports |
| Platform screenshots | Not audited | No dashboard screenshots were captured for this property, so unlike every other study here the export figures are not corroborated by a visible interface showing the property, filter and window. | Not captured |
| Bing organic and Bing Copilot | Not audited | No Bing Webmaster Tools data was exported for this property, so there is no citation figure, no citation share and no Bing ranking data in this study. | Not exported |
| ChatGPT, Gemini, Perplexity | Not audited | No per-answer citation export exists for these surfaces. | No export available |
Why this is the most useful study here for a small team
You do not need scale. You need to be one of the few credible answers.
The flagship study in this portfolio has 241,500 AI citations and is genuinely hard to replicate — it took 500+ hand-tested products and years of work. This one is the opposite, and for most companies reading this it is the more actionable of the two.
A 271-page site in one small market, with no budget, gets a higher share of its impressions from AI features than any other property we run. Not because it is clever. Because the questions it answers have almost no other credible answer.
The second transfer is the one in the language section above. We stated a confident rule on another property — vernacular search carries real intent — and this property contradicts it outright. Read your own export before you act on anybody's rule, including ours.
Who did this work
An owned property, published with its gaps showing
Alston Antony
Owner & SEO lead
MakeMoneyOnline.lk is owned and operated by Alston Antony, who is Sri Lankan and built the site for the market he grew up in. He has worked in search since 2010, holds an MSc in Computer Software Engineering (Distinction) from the University of Greenwich, and is a professional member of BCS, The Chartered Institute for IT.
In a category full of schemes, the editorial line is that income figures are in rupees, stated as ranges, and honest about the low end. That is also the reason the pages are usable as a source.
How to verify every number on this page
- No screenshots and no Bing data exist for this property. Stated at the top, in the coverage table, and here. Every figure comes from Google Search Console CSV exports dated 14 August 2026.
- Site totals use the Devices export, which is complete. The query export hit Search Console's 1,000-row cap, so position bands are shares of that sample. The pages export returned 271 rows, under the cap, so that is a real total.
- Two AI totals are published because they differ: the Devices export sums to 5,795, the Pages export to 5,983. Google attributes the breakdowns slightly differently. The 8.1% share figure uses the Devices number, the smaller of the two.
- The cross-property comparison divides each property's Generative AI export total by its Performance export impressions. Google's AI report only began collecting on 18 May 2026, so every rate is computed on the same partial-window basis — comparable to each other, but not to a full 16 months.
- Position bands were recomputed from the raw export. “#1” means an average below 1.5, because Search Console reports position as a decimal average.
- The language finding is a null result, and those are easy to get wrong. Four Sinhala-script queries appear in the 1,000-row export with zero clicks between them. A query set truncated at 1,000 rows could in principle hide more, so the honest claim is “not material in the measured set”, which is what the page says.
- The site is live. Open makemoneyonline.lk and search any of the queries in the table above.
Case study FAQ