After the Purge: What Cancelling 6,765 Lisbon AL Licences Did to the Containment Map

How a missing insurance certificate became a cancellation

The legal ground is narrow. Decree-Law 128/2014, as rewritten by Decree-Law 76/2024 from 1 November 2024, requires every AL operator to hold and keep valid civil-liability insurance (article 13.º-A) and lets the council ask for proof at any time, to be produced within three days. Article 9.º(1)(d) allows the mayor to cancel a registration for "inexistência do seguro obrigatório válido ou a falta de envio da informação", the absence of valid insurance or the failure to send the information. The insurance duty dates from 2018. What 2024 added was the obligation to send proof through the state platform, and the national register (RNAL) now shows each policy's expiry date.

The machinery followed. The Administrative Modernisation Agency opened a gov.pt service for uploading proof in mid-March 2025, and owners were notified in phases until mid-April. Lisbon and Cascais began sending notices of intention to cancel on 23 June 2025, and Sintra on 14 July, each with ten working days to comply and a prior hearing for each file. ALEP estimated in June 2025 that about 70,000 registrations nationally had not submitted proof. The Ministry of Economy counted 49,887, or 39.5%, in December 2025.

Cancelling thousands of files one at a time was beyond most councils, so ARTE, the state's technology reform agency, built a bulk-cancellation function into the register. By April 2026, 159 councils had asked to use it, 35 were using it and 108 had not yet named the technician needed to switch it on, according to the government office that oversees ARTE. About 35,000 cancellation processes had been opened between July and December 2025. Lisbon finished first. Its reference date is 1 February 2026, and the result became public on 19 and 20 February.

The council's description of what it removed is broader than its legal ground. It spoke of properties "that stopped receiving guests or whose registration was made preventively without the activity going ahead". The legal test was the insurance upload. No list of cancelled registrations, no check against platform listings and no other evidence of inactivity has been published.

What 6,765 is, and what it is not

The title figure holds. Every outlet that reported it, from Renascença and RTP to ECO, idealista and Observador, traces it to the mayor's office or the council, and no revised total has appeared since. The office's own sentence was "Encontram-se ativos 11.779 registos de AL e foram cancelados 6.765 registos": 11,779 registrations active and 6,765 cancelled, with reference to 1 February 2026.

The percentage is where the reporting slipped. Headlines said Lisbon closed 40% of its AL, dividing by a rounded base of "about 19,000" or "about 20,000". The council never published the base. Adding its two figures gives 18,544, which matches ALEP's November 2025 estimate of 18,600 and the Ministry of Economy's 18,537 in December 2025. On that base the purge removed 36.5%. On 19,000 it would be 35.6%. Neither is 40%.

The number left behind is slightly unstable too. The mayor said 11,779. Público, relayed by ECO in April, gave 11,774. A DN fact-check in September used 11,860, and ALEP's rounded September figure is 11,700. The spread is under 2% and probably reflects different snapshot dates, but no source explains it.

The most revealing figure is older. When the 2023 Mais Habitação package required every operator to prove activity, Lisbon had 19,917 registrations and 11,447 complied, according to figures Left Bloc (BE) councillors cited in February 2026. The difference, 8,470, matches exactly the number of Lisbon registrations the Ministry of Economy listed as at risk in December 2023, 42.5% of the city's register. Two years earlier ALEP had estimated that more than 3,500 of 19,651 Lisbon registrations were inactive. A core of about 11,500 registrations has shown signs of life in every test since 2023. The February purge moved the register onto that core. It did not cut into it.

Two of those corrections concern our own reporting. Our containment zones explainer, published before this research, said the mayor's "zero" new licences covered February to April and that BE was warning of rising registrations. The mayor's statement dates from 24 February and covers about three weeks. BE's warning was a prediction without a count. Both points are corrected above.

What the purge did to the containment map

Lisbon's rule, in force since 6 December 2025, measures registered AL units against permanent-residence dwellings from the latest census at three levels: the city, the 24 parishes and 274 monitored neighbourhoods. A ratio of 10% or more puts an area in absolute containment, where new registrations are barred. From 5% to under 10% is relative containment, where only rooms in an owner's home and rehabilitation projects can be authorised. The denominator did not move in February. The numerator lost more than a third.

The last published ratios date from 1 November 2025, three months before the purge. Santa Maria Maior stood at 66.9%, Misericórdia 43.8%, Santo António 25.1%, São Vicente 16.1%, Arroios 13.5%, Estrela 10.8%, Avenidas Novas 6.6% and the city as a whole 7.2%. On 24 February the councillor for urbanism, Vasco Moreira Rato, told the Municipal Assembly that Santa Maria Maior and Misericórdia had each lost more than 1,000 registrations and Arroios more than 800. No other parish figure has been given.

The recalculation came in early April, first reported by Público on 5 April. Santa Maria Maior, Misericórdia and Santo António stayed in absolute containment. São Vicente, Arroios and Estrela fell to relative containment. Avenidas Novas left containment. At neighbourhood level ECO counted 22 neighbourhoods in absolute containment and 11 in relative, with five moving from absolute to relative and seven leaving relative containment altogether. The names of those twelve neighbourhoods have not been published, and neither has the instrument that made the change. A property-management firm's guide, updated in July, said the new list was still awaiting publication in the Boletim Municipal.

Without new ratios, the size of each move can only be bounded. The table below shows what share of each parish's pre-purge registrations had to disappear for the reported change to happen, assuming the denominator was unchanged. Estrela needed to lose only 7.4% of its registrations to fall below 10%. Arroios needed to lose more than 25.9%, São Vicente more than 37.9%, and Avenidas Novas more than 24.2% to drop under 5%. Santa Maria Maior would have needed to lose 85.1% to leave absolute containment, which is why no clean-up of dormant files could move it.

Those thresholds sit close to the city-wide share removed. A parish that lost the city average of 36.5% would clear Estrela and Arroios easily and São Vicente narrowly, which is what the reclassification shows. If the dormant share had been equal everywhere, the city-wide ratio would have fallen from 7.2% to about 4.6% (derived), below the 5% line. The city has confirmed none of this. The reclassification tells us which side of each line each parish landed on, and nothing about how far.

The political reading split along the same line. Rato said in February that the cancellations "do not necessarily open vacancies" in containment parishes. Six weeks later three parishes had moved to a tier that allows room and rehabilitation authorisations, and one had moved to a tier that allows ordinary registration. BE councillor Carolina Serrão argued in April that the rule approved with Chega's votes, combined with the purge, "opens the door to new licences" in four parishes. Both statements hold in part. Removing dormant files does not create new rentals, and in the three parishes that matter most it changed nothing. In four others it lowered the bar.

What happened next is not measured. Mayor Carlos Moedas told the Municipal Assembly on 24 February that the council had issued "zero" AL licences since the cancellation. Seven months on, no figure for new registrations in Avenidas Novas, São Vicente, Arroios or Estrela has been published, and neither has a city total after February. The monthly recalculation the regulation requires is, in practice, visible only when the press reports a change of status.

Twelve years of registrations

Seen over a longer run, the purge looks less like a policy and more like an audit. Lisbon's register rose from 14,461 in August 2018 to 17,937 in January 2019 and then barely moved for six years through three different regimes: the 2019 containment rule, the 2023 national suspension and the 2024–25 Lisbon freeze. It peaked at 19,917 in 2023 and stood at 18,537 in December 2025. The national register kept climbing to an all-time high of 126,320 in December 2025, before falling for the first time in its history.

The 2023 episode explains why the dead weight survived so long. The Mais Habitação test found 45,747 registrations nationally without a valid activity declaration, 37.9% of 120,719 (derived), but nobody executed the cancellations. ALEP's president said at the time that cancellation "has not been immediate, nor can it be", and the Lisbon mayor said the council had two staff for the job. The 2026 round worked because ARTE's bulk tool turned thousands of individual procedures into one.

Was anything actually trading?

If the cancelled registrations had been live rentals, Lisbon would have lost more than a third of its short-term supply overnight and the data would show it. The best public series is Eurostat's count of nights booked through the four big platforms, which covers Grande Lisboa, the nine-municipality region around the capital. Those nights fell 4.5% year on year in January 2026 and 6.0% in February, then rose 0.3% in March. The first quarter came to 1,988,137 nights, down 3.1%, while Portugal as a whole rose 4.9%.

That is a real softening, and Lisbon underperformed the country. It is not a 36% supply shock. The fall also began in January, before the purge took effect, and the March comparison is distorted by Easter, which fell on 5 April in 2026 and on 20 April in 2025. Eurostat had published nothing after March when this article was written. For the full year 2025, platform nights in the region rose 2.0% to 11,752,334.

INE's accommodation survey points the same way at national level. In the second quarter of 2026, AL establishments with ten or more beds took 14.5% of nights in Portugal, down 1.2% year on year, while hotels grew 1.6%. The survey excludes the small flats that make up most of Lisbon's register, and no Lisbon-level AL figure for 2026 has been reproduced. Hotels dominate the city's accommodation anyway: as our Lisbon tourism statistics article shows, they took 81.4% of the municipality's 15.7 million overnight stays in 2024.

Commercial trackers cannot settle the question because they disagree with one another. AirDNA counts 15,792 active Lisbon listings in the twelve months to August 2026, down 29.4%. AirROI counts 13,896 over a similar period, up 15.7%. GuestFavorites counts 11,272, and Airbtics, before the purge, counted 13,014. The base implied by AirDNA's 29.4% fall is about 22,400 listings, more than the 18,544 registrations that existed, which suggests double counting across channels or a change of method.

One comparison is suggestive. Airbtics' pre-purge count of about 13,000 active listings sat roughly 5,500 below the 18,500 registrations on the books at the time. That gap is of the same order as the 6,765 later cancelled. It is indirect evidence from a commercial scraper, but it fits the account given by the council and by ALEP, whose president said owners "kept the registration but did not use the property for this purpose" and that several flats had been sold without the registration being cancelled.

Did any homes come back?

The housing case for cancelling licences is that a flat which stops being a holiday let can become a home. The purge offers almost no test of that, because the flats it removed were mostly not holiday lets to begin with. The data that exists points the other way. idealista's count of homes advertised for long-term rent in Lisbon city fell 18% year on year in the first quarter of 2026 and 27% in the second, against national falls of 13% and 22%.

Rents give a mixed signal. INE's first rent release in more than a year, built on a contract base widened with tax-authority data, put Lisbon's median rent on new leases at €17.42 per square metre in the first quarter of 2026, up 8.2% year on year. idealista's asking-rent index had Lisbon at €21.8 per square metre in June, down 1.7%. Neither series can isolate any effect of the AL register.

A DN fact-check published by ContraProva on 22 September 2026 rated as misleading the claim that Lisbon's AL fell 40% in three years while rents did not fall. Its reasoning matches the evidence above: most of the reduction did not correspond to rentals ceasing to operate. A clean-up of the register cannot lower rents if the entries it removed were not competing for tenants.

Lisbon in the national clean-up

Lisbon was first, and it is still most of the national total. The 10,324 cancellations reported by DN, ECO and Observador in June 2026 are the exact sum of four councils that had finished the process: Lisbon 6,765, Lagos 1,548, Porto 1,413 and Lagoa 598. Lisbon accounts for 65.5% of them (derived). Other councils had processes under way but had not reported totals, so the figure is a floor for the country rather than a final count.

Lisbon's cancellation rate also stands out. It removed 36.5% of its register. Lagos removed about a quarter of 6,071 registrations, Lagoa about 16% of 3,755 and Porto 11.5%, leaving 10,821 active. Porto's council attributed its lower share to stricter inspections in earlier years. The 2023 data shows the same gap: 42.5% of Lisbon registrations were at risk then, against 17.9% in Porto. Our Porto destination page covers the city whose register turns out to have been the cleaner of the two.

Nationally the register fell from 126,320 in December 2025 to 119,147 active in June 2026. More than 37,000 registrations were still without insurance proof in June, and 139 councils had not begun formal procedures. ALEP expects about 30,000 more cancellations as the rest catch up, which would leave between 85,000 and 93,000. That is a projection, not data. The share that matters most is in the Algarve, which held 40% of national registrations in February 2026; Lagos and Lagoa are the first Algarve councils to complete the process.

Porto shows that the purge can move a map elsewhere too. It uses a single containment threshold of 15% by parish rather than Lisbon's two tiers, and its 1,413 cancellations fell mostly on a register already judged cleaner. We found no recalculation of Porto's parish ratios after May 2026. Councils with more than 1,000 registrations have until 31 December 2026 to adopt or revise AL rules under Decree-Law 151/2026, so several maps may be drawn on newly cleaned registers in the next three months.

What the register still cannot tell us

The purge has made Lisbon's register a better measure of what exists, but it is still a list of permissions. The first official count of what actually trades should come from EU Regulation 2024/1028, which has applied since 20 May 2026. Platforms must display each host's RNAL number and send activity data to a national single digital entry point. We found no source naming Portugal's entry point or any data published through it.

Four further releases would show what the purge did. The first is any set of post-purge ratios, which would show how close São Vicente, Arroios and Estrela sit to the 10% line and whether the city-wide ratio has fallen below 5%. The second is a count of new registrations since February, above all in Avenidas Novas. The third is Eurostat's platform data for April to September 2026. The fourth is the two-year review of the regulation, due by December 2027.

What this means for travelers, businesses and analysts

Travelers. The purge changed almost nothing you can see. The registrations it removed were mostly not taking bookings, platform nights in Grande Lisboa fell only 3.1% in the first quarter, and every surviving registration in Alfama, the Baixa and Bairro Alto remains valid. Since May 2026 platforms must display a host's RNAL number; a Lisbon listing without one deserves caution. Expect central supply to hold rather than grow, because the three historic-core parishes remain in absolute containment. Our Lisbon travel guide covers where to stay, and the Lisbon destination page explains which neighbourhoods suit which kind of trip.

Tourism businesses and investors. Underwrite Lisbon on the 11,800 registrations that survived, not on the 18,500 that existed on paper. Nothing in the purge opens Santa Maria Maior, Misericórdia or Santo António. In São Vicente, Arroios and Estrela relative containment now allows rooms in an owner's home and rehabilitation projects, not new whole flats. Avenidas Novas is open to ordinary registration, but it sat at 6.6% before the purge and a run of new registrations can push it back over 5% at the next monthly recalculation. The purge also carries a lesson for operators: a missing insurance upload was enough to lose a registration, and in a containment parish a lost registration cannot be replaced. The threshold to watch is the city-wide 5% line; if the illustrative 4.6% is close to the real figure, the city sits just below it.

Journalists and analysts. Safe to cite: 6,765 registrations cancelled and 11,779 active at 1 February 2026 (CML); 36.5% of an implied 18,544 (derived); legal ground of missing insurance proof under DL 128/2014 art. 9.º(1)(d) as amended; São Vicente, Arroios and Estrela moved to relative containment and Avenidas Novas out of containment in April 2026; 10,324 cancellations in four councils by June 2026, 65.5% in Lisbon; Grande Lisboa platform nights −3.1% in Q1 2026 (Eurostat). Do not repeat: that Lisbon closed 40% of its short-term rentals, that the purge returned homes to residents, that Lisbon's listings fell 30%, that Portugal has cancelled 36,000 registrations, or that the cancellations opened no space for new licences. The finding this article adds to the series: for at least three years Portugal's AL register overstated Lisbon's live supply by about a third, and every ratio, map and per-resident comparison built on it carried that error.