Overtourism in a half-empty country

Sri Lanka closed 2025 with 2,362,521 international arrivals, the highest annual figure in its history and 15.1% above 2024. The margin over the old mark is thinner than the word “record” suggests: 2018 brought 2,333,796 visitors, so it took seven years, an Easter attack, a pandemic and a sovereign default to move the record by 1.2%.

The same December the record was set, UN Tourism and the SLTDA published a three-volume Visitor Management Study built on three field missions to ten pilot sites, from Sigiriya and Yala to Mirissa and Ella. Its framing sentence is sweeping: “All the sites selected to be considered in this project are currently facing, at least seasonally, issues with overtourism.”

The two findings only look like a contradiction. Arrivals have barely regained a level the country handled seven years ago, and graded hotels sat 44.5% full in 2024 against 72.8% in 2018. Yet the study documents 500 jeeps in one block of Yala, a Sigiriya staircase that can safely pass only about 600 people an hour, and, during a rare relic exposition in Kandy, a single April day that drew a reported 450,000 people and left four dead.

The record that wasn’t a boom

Arrivals crossed the 2018 peak by a hair. The money did not come with them.

The record is real, and so is the recovery behind it. Every month of 2025 beat its 2024 counterpart, led by September at +30.2%. The number is watched because tourism earns above its weight, the third-largest source of foreign exchange in 2023 at 8.7% of the total, even though it is only about 2.5% of GDP and directly employs a little over 200,000 people. It is a dollar tap first. But the dollars did not follow the crowd.

SLTDA puts 2025 tourism earnings at US$3,219 mn. In 2018 the same series read US$4,381 mn. A record number of people produced 26.5% fewer nominal dollars, and every component of the yield moved the same way.

One caveat cuts both ways. SLTDA reset its spend-survey base mid-2025, a switch that by itself removed about US$188 mn from second-half receipts, so part of the fall is definitional rather than behavioural. The direction, though, is the one the Central Bank’s 2025 review already named the yield problem.

The second series is occupancy. Registered rooms grew to 59,578 in 2025, an addition SLTDA puts at 6,200, or 11.6%, in a single year, while graded hotels averaged 44.5% occupancy across 2024. The build is not just running ahead of the guests; it is running toward the crowds rather than away from them. More than three-quarters of the roughly 4,100 rooms added in 2025 went to the Western, Southern and Central provinces (Western and Southern alone already hold three-fifths of the national room stock), while the Northern Province took 106. Nuwara Eliya added 40.8% more licensed rooms between 2019 and 2024 against flat demand; in Ella’s Badulla District, 468 licensed establishments sit beside more than 700 listings on Booking.com, a shadow supply the state cannot even see.

So the national picture is a country with spare capacity almost everywhere, in its bed stock, its shoulder months, and the two-thirds of the year outside the peaks. On this evidence overtourism should be impossible. The rest of this piece is about why it is not.

Ten sites, ten different pressures

The label is one word. The study’s own chapters describe ten different problems, and only some of them are crowds.

The study is the most serious look at crowding Sri Lanka has commissioned, built on three field missions, smartphone geolocation data for three of the sites, a stairway flow calculation and a volume of international benchmarks. Read its ten chapters one by one and the single word in its framing sentence opens into ten different diagnoses, from jeep congestion to reputation damage to towns that simply built too many rooms.

Pinnawala shows how varied the ten diagnoses are, because its numbers point down, not up. The study records visitation about 40% below its 2015 high, a foreign share fallen from 33.0% to 26.6%, and a TripAdvisor page carrying a 3.5 rating under an animal-welfare warning banner. Its chapter reads less like crowd control than reputation repair; the recommendations are a rebrand as a conservation and research centre, an end to elephant bathing in its current form, and a rebuilt online presence.

The Kandy number needs the opposite qualifier. The April 2025 crush came during a special exposition of the sacred tooth relic, ten days in April, the first public showing since 2009. Roughly 450,000 people came on the peak day, queues ran close to ten kilometres, and press reports counted four dead and more than 300 hospitalised, most from heat exhaustion, an early toll that rose as the ten-day exposition continued. That is a once-in-a-generation religious event, not a tourism season, and the crowd was overwhelmingly Sri Lankan devotees. What it exposed is the absence of event crowd management, and here the study gets specific, recommending real-time crowd monitoring, smart barriers, mobile medical units, a festival app with live alerts and rehearsed evacuation routes, none of which were in place that April. The next exposition may be a decade away. The Esala Perahera comes every year.

So the real question is not whether Sri Lanka has overtourism everywhere. It does not. It is where the genuine pressure concentrates, and why the same handful of gates take it every time. The next sections take the concentration apart on three axes, space, season, and the hour of the day.

Concentration in space

Ten million tickets a year, and in every category between a third and a half of them pass through a single gate.

Set the zoos aside; they are almost entirely a domestic outing (the Dehiwala zoo’s 2025 crowd was 1.3% foreign; see the methodology note) and sit outside the visitor economy this piece examines. Across the five remaining ticketed categories the SLTDA publishes, wildlife parks, heritage sites, botanical gardens, museums and conservation forests, Sri Lanka sold 8.07 million site tickets in 2025 against 2.36 million international arrivals. About 2.86 million of those tickets went to foreigners, roughly 1.2 ticketed visits per arriving tourist; the other 5.2 million were Sri Lankans. The load is large. Its distribution is the problem.

Sigiriya took 55% of all Central Cultural Fund heritage visits in 2025, 1.09 million tickets. Yala took 29.7% of all national-park visits and 37.4% of the foreign ones. Peradeniya took 54.4% of botanical-garden visits, Colombo National Museum 44.4% of museum visits, Hurulu 33.2% of forest visits. And concentration repeats inside the sites: more than 90% of Yala’s visitors enter Block 1, one of five blocks; at Galle Fort the study’s geolocation heatmap shows the densest crowds not at the heritage assets but in the commercial centre, “suggesting that these commercial establishments are more important for many visitors, than the actual heritage assets of the area”.

Spatial concentration is what makes national spare capacity irrelevant. The 44.5% occupancy average and the empty eastern parks do not relieve Block 1 at dawn; they merely coexist with it.

Concentration in season

The recovery changed who comes. It did not change when.

The market mix behind the 2025 record is a different country from 2018. India is up 25.1% to 531,511, now more than double any other market; Russia is up 189%, Bangladesh 468%, Poland 145%. The markets that anchored 2018 have not come back. China has halved, the United Kingdom is down 16.5%, Japan down 23%.

The deeper change is structural. The new arrivals come cheaper, shorter and older. India supplies 22.5% of all arrivals but, at a five-night average stay, only about 13% of the nights, while the long-staying northern Europeans who once anchored the calendar are a shrinking share, and travellers over 60 are now 17.5% of arrivals, the fastest-growing band. The pipeline is concentrated too. Three Gulf hubs (Dubai, Doha and Abu Dhabi) funnel 30% of all arrivals, while sea and cruise together carry under 1%. A market this short-staying and this funnelled does not disperse into regional room-nights. It compresses into day-trips to the marquee gates.

What did not change is the calendar. The ratio between the busiest and quietest month was 1.96 in 2018 and 1.95 in 2025; December remains the peak, May and June the trough, and the whole curve lies almost on top of its pre-crisis self.

At the sites, that mild national wave amplifies. Yala’s 2024 range ran from 28,543 visitors in June to 84,734 in December, a factor of three. Polonnaruwa’s peak months double its low ones. Hurulu is the extreme case. Because its season follows the elephants that move between Minneriya, Kaudulla and the Hurulu reserve, its daily load swings from about 80 visitors in September to more than 1,250 in December, a factor of nearly sixteen. A site sized for its December crowd is empty in September; a site sized for September breaks every December.

Seasonality is the oldest complaint in Sri Lankan tourism, and the standard answer has been marketing the shoulder months. The 2018-to-2025 overlay shows how little that has moved. The study’s answer is different. If the wave cannot be flattened, meter it, with time-slot tickets, quotas and event calendars that shift demand within the peak rather than wishing the peak away.

One more signal sits in the newest data. December 2025, normally the busiest month of all, grew just 4.16% year on year while the eleven months before it ran at about 16.5%; Cyclone Ditwah had closed roads and emptied the hills, collapsing domestic visitation at Knuckles to 92 people for the month. It is one month, and a caveat more than a trend. But it points at a fourth kind of concentration the study does not model: exposure. The calendar that stacks demand into a few months stacks it into the months the monsoon, and its cyclones, are most likely to take away.

Concentration in the hour, one staircase

Sigiriya’s constraint is a metal stairway roughly 300 steps long, and the hour everyone wants to be on it.

Sigiriya is the country’s most visited paid attraction, with 1,086,918 tickets in 2025, of which a record 582,687 were foreign, up 14.6% on 2024. In its peak months the site averages more than 3,300 climbers a day, and the study’s smartphone-geolocation data shows them arriving together. The densest crowding falls between 9 and 10 in the morning, before the heat.

Something else changed at the top of the rock in 2025. For the first time, most of the crowd was foreign. Sigiriya’s foreign share crossed from 30% in 2023 to 45% in 2024 to 53.6% in 2025, as foreign tickets more than doubled over three years and domestic ones fell a fifth. The site the country built for its own schoolchildren and pilgrims is now, by headcount, an international one, which is precisely the crowd the fee schedule and the study are built around.

Everyone on the summit has passed through one bottleneck, the final steel staircase above the Lion’s Paw terrace. A 2022 report put the acceptable load on top of the rock at 8,500 people at any given time; the new study rejects that as physically unreachable and does the queueing arithmetic instead. At observed climbing speeds the staircase safely passes about 600 people an hour. That is the number that matters, because it turns a volume problem into a timing one. Spread evenly across an eight-hour gate, even a peak day’s 3,300 climbers would arrive at roughly 400 an hour, comfortably under the limit. They do not spread. Bunched into the nine-o’clock window, that same crowd is what produces the queue, the heat casualties and the wasp-attack pile-ups the study records. Sigiriya is not over capacity. It is over capacity for sixty minutes.

The mismatch between demand and stairway is a management problem before it is an infrastructure one, and the study’s inventory of the management is unsparing. The online ticketing portal exists but is “almost not used”; it requires a Google account, demands a passport number before showing prices, and offers no time slot, so it confers no benefit over the queue. The QR codes on site do not work. Medical services staff the mornings only, at a site whose risks (heat, overexertion, wasps) run all day. The museum included in every ticket sits hidden on an upper floor and is skipped by most visitors.

In June 2025 Sigiriya became the first destination in Sri Lanka to be certified a Sustainable Destination by Cabinet approval. The study, written the same year, records that its managers still treat tourism as, in its words, a by-product.

Concentration on wheels, the jeep economy

At Yala the study drops its academic hedging and names the cap it thinks the park needs.

Yala received 646,704 visitors in 2024 and, per the study, sometimes holds up to 500 jeeps inside the park at once, nearly all in Block 1, chasing radioed leopard sightings. Licensed jeeps rose from 350 to 550 in a single year; rule-breaking drivers are banned “about once per month”. TripAdvisor holds the park at 3.5, which the study calls “a dangerously low average rating for such a popular tourist attraction”. Its headline prescription is a hard number: no more than 100 to 150 vehicles in Block 1 during the morning and evening safari periods, tightening further over time.

Then the turnstiles moved again. In 2025, the year the study was being written and published, Yala grew another 19.5%, to 772,764 visitors. Every constraint it documented is now carrying a fifth more load than when it was documented.

Who pays, and who gets counted

Foreigners are half the heritage crowd and nearly all of the heritage money. Sri Lankans are most of the crowd at every gate, and almost none of the management.

Why any of this matters is a revenue question before it is a comfort question. The five ticketed categories collected about Rs 14.3 bn at the gate in 2025, roughly US$48 mn, led by the wildlife parks (Rs 6.7 bn) and the heritage sites (Rs 5.5 bn). And the gates anchor far more than their own tickets; these sites are much of the reason visitors board the plane at all, which puts them at the base of the US$3.2 bn tourism economy and the country’s third-largest source of foreign exchange. That stream is only as durable as the experience at the turnstile, which is what visitor management exists to protect.

Start with the crowd, not the money. Of the 8.07 million site tickets sold across the five categories in 2025, 5.2 million, almost two in three, went to Sri Lankans. They are the majority in every category and the overwhelming majority in most, 82% of museum and botanical-garden visits, two-thirds of forest visits. The foreign visitor the whole apparatus is built around is, at the turnstile, the minority.

Dual pricing is what flips the picture. The Central Cultural Fund’s sites took 1.97 million visitors in 2025, 48.5% of them foreign, yet foreigners supplied 98% of the Rs 5.48 bn ticket income. At the national museums, foreigners were 17.9% of visitors and 79.8% of revenue. On the CCF’s published split, a foreign visitor is worth roughly fifty times a domestic one at the gate, which is why every management decision reads the foreign column and none reads the crowd.

Pricing follows the same lopsided pattern, and no coherent crowd logic sits under it. Each institution sets its own gate fees, at its own time, for its own purposes. The result, as the study records it, is that Hurulu, the least regulated gate in the busiest safari cluster, charges foreigners a fraction of what the national parks beside it do, while Polonnaruwa carries one of the heritage estate’s premium fees at a site the study itself says has no strong crowding problem. Whether price causes either outcome, this data cannot say. What is visible is that nowhere in the system is a fee doing crowd management.

The domestic side of the ledger is moving too, and in the opposite direction. Sri Lankan visits to the legacy paid attractions have slid for years; domestic botanical-garden visits are down 28.6% on 2018, with Peradeniya alone losing nearly a third. Where domestic leisure still grows it has moved to the wildlife parks, up 16.8% in a single year, and to conservation forests and sites like Horton Plains that are now almost entirely a Sri Lankan weekend destination. It is a cost-of-living story hidden inside a tourism dataset. The households priced out of the gardens are still filling the parks.

And the study never looked. Its ten pilot sites are the foreign-facing ones. The country’s actual largest crowds are domestic and almost all absent from it, Peradeniya with 781,000 local visitors, Knuckles with 246,000, Jaffna Fort with 235,000, of whom 94% are Sri Lankan. Manage the foreign minority and you have managed the money. You have not managed the crowd.

The sites nobody counts

Before the fix comes a measurement problem. For half of the study’s ten flagship destinations, the best available number is a partial proxy, or nothing at all.

The study’s most basic table carries a finding of its own. For the Temple of the Tooth, Nuwara Eliya and Ella, the visitor-number cell reads, in effect, unknown. Mirissa is legible only through its whale-watching permits (119,618 passengers in 2024). Galle Fort has no count either; the study falls back on the three museums inside the fort, which logged 155,794 visitors in 2024, 83.8% of them Sri Lankan, a fraction of the fort’s actual footfall. These are not obscure places. The country’s crowd-management debate rests on turnstile data that exists for only half the sites being debated.

The study is candid about the reason. No formally established destination management organisation exists for any of these sites, or anywhere else in Sri Lanka; most managing agencies have no dedicated tourism staff; and at all ten sites, private operators told the researchers there is “virtually no relationship or exchange” with the public institutions in charge. Hurulu Eco Park, which absorbed roughly 245,000 visitors in 2024, is run by eight ticket-office staff and three rangers, none of them designated for tourism.

The paper fix and the moving target

The study’s toolkit is redistribution: time slots, caps, second gates. The data above says distribution is precisely what is broken. What is missing is anyone to operate the tools.

Read all ten site chapters and the prescriptions converge to the point of repetition: a time-slot reservation system with visitation caps; working online ticketing; geolocation-based flow monitoring; redistribution through second entrances, one-way routes and alternative trails; and, at every single site, a Tourism Management Committee meeting quarterly; most chapters add dedicated tourism staff inside the managing agency. The repetition is the finding. Ten different sites, six different managing institutions, one shared absence.

The international cases in Volume 2 make the same point from the other side. The world’s most visited heritage sites are the ones that meter the hour.

The Alhambra comparison is the sharpest, with its limits acknowledged. It is a complex that visitors spread across for hours, where Sigiriya is one climb to one summit. Even so, it admits two and a half times Sigiriya’s annual crowd under a hard daily ceiling, with its most fragile palace metered in half-hour increments sold months ahead. Metering did not shrink its tourism; it made the volume survivable. Sigiriya, with a bottleneck with a practical limit of about 600 people an hour, sells unlimited same-day tickets at the gate.

Institutions are where the study ends, because they are where every prior plan has stalled. Volume 3 states the baseline plainly. No formally established destination management organisations exist in Sri Lanka. The funding examples it reaches for describe machinery an order of magnitude smaller than the money already crossing Sri Lankan gates. Queenstown’s levy-funded DMO runs on a US$3 mn budget and 20 staff; San Diego’s is a self-assessed hotel fee. The Rs 14.3 bn the gates collected in 2025 is some sixteen times Queenstown’s entire budget; Sigiriya’s tickets alone, at Rs 3.73 bn or about US$12.5 mn, are four Queenstowns. That is gross gate revenue, not a management budget. But the revenue to fund visitor management plainly exists. It is collected at the exact sites that need it, yet none of it is ring-fenced to manage them.

Something did happen in September 2025 that had not before. At a Colombo workshop closing the study, representatives of all ten sites signed a Statement of Intent to form Tourism Management Committees, the modest quarterly-meeting kind. The study proposes twinning them in pairs (Yala with Hurulu, the Temple with Polonnaruwa, Sigiriya with Galle Fort) so each can borrow the other’s fixes, and growing them over time into funded DMOs. It is a signature, not a system. But it is the first time the ten most crowded places in the country agreed, on paper, that someone should be managing the crowd.

Between the study’s fieldwork and its publication, Yala grew 19.5% and Hurulu 21.2%. The target the committees were built for is already moving.