Do public sellers show the energy label when they announce property sales in Spain's Official Gazette? A lot-by-lot audit of BOE notices, January 2025 – September 2026

Abstract

Spanish law requires the energy-efficiency label in «toda oferta, promoción y publicidad dirigida a la venta o arrendamiento» of a building (Royal Decree 390/2021, art. 15.2). Public bodies that sell property must announce the sale in the Boletín Oficial del Estado (BOE).

We read all 361 candidate notices that the BOE's open-data summaries list for 1 January 2025 to 30 September 2026. We kept the 254 notices that offer property to the public and split them into 2,833 lots. For each lot we decided whether the Royal Decree applies, using its scope (art. 3), its exclusions (art. 3.2), its definition of «edificio» (art. 2.h) and the ministry's FAQ. We then checked whether the notice states the energy rating. A rule-based classifier made the first pass. Its rules were developed on 2024 notices, but with knowledge of a pilot that had already read the study period. An AI agent then reviewed the uncertain and non-compliant lots, and an independent AI agent re-read a sample.

144 of 306 covered lots announced in the BOE in the period (47.1%, 95% CI 41.5–52.7%) did not include the energy rating. With the lot parser as frozen, before a fix that corrected how lots were split, the figure is 92 of 255 (36.1%). The fix added 52 lots without the rating, all of them INVIED premises that the frozen parser had merged. The rate differs sharply by seller:

• Tesorería General de la Seguridad Social (TGSS): 14 of 160 (8.8%);
• Delegaciones de Economía y Hacienda, which sell the State's property: 17 of 33 (51.5%);
• INVIED (Defence): 86 of 86 (100.0%);
• GIESE (Interior): 21 of 21 (100.0%);
• FOGASA: 5 of 5;
• one port authority: 1 of 1.

These figures describe what the notices say. Whether a mandatory gazette notice is «oferta, promoción y publicidad» has not been settled by any authority, so they are not findings that anyone broke the law.

Under a rule that requires both letters (consumption and emissions), as the recognised document to which art. 17.3 refers does for press advertisements, 149 of 306 lots lack the rating. Under the strictest reading of the sellers' declared exemptions, 157 of 319 do.

1. Introduction

Since 2013, a Spanish seller or landlord who advertises a building has had to show its energy rating. The rule now sits in Royal Decree 390/2021: the label «se incluirá en toda oferta, promoción y publicidad dirigida a la venta o arrendamiento del edificio o de parte del mismo» (art. 15.2). Studies and inspection campaigns have measured how often private advertisements comply, on property portals and in estate agencies (§3). We found none that looks at the State's own sale notices.

Public bodies sell a steady flow of property:

• the Social Security treasury (TGSS) sells offices and flats it no longer uses;
• the Delegaciones de Economía y Hacienda sell the State's patrimony, much of it inherited from people who died without heirs;
• the Defence housing institute (INVIED) sells premises and military land;
• the Interior ministry's GIESE sells former Guardia Civil and police buildings.

The law on public patrimony requires these sales to be announced in the BOE. The notices are open data: the BOE publishes a daily summary through a documented API, and every notice is a public web page.

We asked a narrow question. When a public body announces in the BOE the sale of a building that falls under RD 390/2021, does the notice state the energy rating? We answer it lot by lot, because one notice can sell one flat or 226 rustic plots. We separate lots the Royal Decree covers from lots it excludes, and we set out the reading of «publicidad» on which the question depends.

2. Legal frame (quoted as in force in the period)

The consolidated texts below were read through the BOE's consolidated-legislation API on 4 October 2026. They are consolidated texts of a merely informative character, not the official texts. RD 390/2021's consolidated text was last updated on 23 July 2026; its arts. 15 and 17 have a single version, in force since 3 June 2021.

2.1 The obligation

RD 390/2021 (BOE-A-2021-9176) was amended by RD 659/2025 with effect from 23 July 2026. Among other changes, RD 659/2025 added arts. 4 bis to 4 quáter and 7 bis and annexes I-IV on competent technicians. It also rewrote arts. 6.6, 9.1 and 10. In art. 3.2.e it corrected a cross-reference («apartado f)» → «apartado e)») and added Ceuta and Melilla. Arts. 3.1, 3.2.a–d, 15 and 17 were not changed. Two provisions set the duty:

• art. 15.2: «La etiqueta de eficiencia energética se incluirá en toda oferta, promoción y publicidad dirigida a la venta o arrendamiento del edificio o de parte del mismo»;
• art. 17.3: «Toda persona física o jurídica que publique o permita la publicación de información sobre la venta o alquiler de un edificio o de parte del mismo, ya sea en agencias inmobiliarias, vallas publicitarias, páginas web, portales inmobiliarios, catálogos, prensa o similares, estará obligada a incluir la información relativa a su calificación de eficiencia energética, de acuerdo con lo dispuesto en el correspondiente documento reconocido».

The recognised document «Calificación de la eficiencia energética de los edificios» sets the format. We read its version recognised from 30 September 2026. The 2015 version, in force for the rest of the period, could not be read (§7), but the ministry's FAQ (version 6.0, 3 August 2021) restates it. Both texts give the minimum for printed advertisements in the same words. FAQ no. 42: «En el caso de anuncios de prensa se permitirá mencionar solo la calificación energética en Consumo y en Emisiones (letra asociada a las mismas)». So the minimum that art. 17.3 points to is both letters. The FAQ states that it «carece de valor jurídico». The 2013 FAQ (no. 6.1) names who must show the label: «es el promotor o propietario del edificio el responsable de mostrar la etiqueta en toda oferta, promoción y publicidad».

The sanctions sit in the land law (TRLSRU, RDLeg 7/2015, additional provision 12), unchanged since 2015. Three minor infractions are relevant:

• 5.a, «Publicitar la venta o alquiler de edificios o unidades de edificios que deban disponer de certificado de eficiencia energética sin hacer mención a su calificación de eficiencia energética»;
• 5.b, «No exhibir la etiqueta de eficiencia energética en los supuestos en que resulte obligatorio»;
• 5.h, «Cualesquiera acciones u omisiones que vulneren lo establecido en materia de certificación de eficiencia energética cuando no estén tipificadas como infracciones graves o muy graves».

Fines run from 300 to 600 euros (additional provision 13). The regional governments investigate and sanction, and the text exempts no type of seller.

2.2 Scope and exclusions

The Royal Decree applies to «Edificios o partes de edificios existentes que se vendan o alquilen a un nuevo arrendatario» (art. 3.1.b). A building is a «construcción techada con paredes en la que se emplea energía para acondicionar el ambiente interior» (art. 2.h). Art. 3.2 excludes:

• a) protected buildings, «siempre que cualquier actuación de mejora de la eficiencia energética alterase de manera inaceptable su carácter o aspecto»;
• b) provisional constructions of two years or less;
• c) «Edificios industriales, de la defensa y agrícolas no residenciales, o partes de los mismos, de baja demanda energética»;
• d) «Edificios independientes … con una superficie útil total inferior a 50 m2»;
• e) «Edificios que se compren para su demolición o para la realización de las reformas» defined in art. 3.1.d. «Para hacer efectiva la exclusión …, el propietario del edificio o de parte del edificio, según corresponda, realizará una declaración responsable ante el órgano competente de la comunidad autónoma».

The FAQ adds answers we used:

• garages and storage rooms do not need a certificate (no. 14);
• neither do non-habitable or shell premises sold as such (no. 19);
• an empty building is not exempt for being unused: «La clave no es que esté consumiendo o no energía actualmente … sino que sea un espacio que necesita ser acondicionado» (no. 13);
• offices of 50 m² or more in industrial buildings must be certified (no. 17);
• flats and premises of less than 50 m² inside a building are covered (no. 15).

On art. 3.2.e the regulation and the FAQ differ. The regulation names the owner «del edificio o de parte del edificio». FAQ no. 7 says the exclusion «es sólo aplicable a un edificio completo, y no a partes del mismo». We follow the regulation and report the FAQ's reading as a sensitivity (§5.4). The FAQ says nothing about ruins, undivided shares or auctions.

2.3 Is a BOE sale notice «oferta, promoción y publicidad»?

We found no official statement, guidance or judgment on this. Our search covered:

• the MITECO FAQ and the recognised document on the label;
• the consumer authorities' campaign sheets;
• the press;
• two searches for case law. CENDOJ could not be queried with our tools.

The arguments, each anchored in a text:

For.

1. The words are broad: «toda oferta, promoción y publicidad», «Toda persona física o jurídica que publique … información sobre la venta», with a list that ends «prensa o similares».
2. The patrimony law itself treats the gazette as one channel of publicity. The convocatoria «se publicará gratuitamente en el "Boletín Oficial del Estado" … sin perjuicio de la posibilidad de utilizar, además, otros medios de publicidad» (Ley 33/2003, art. 138.3).
3. The notice describes the building and its price. The regulation requires «La descripción del bien … con indicación de sus datos catastrales y registrales» and «El tipo de la subasta» (RD 1373/2009, art. 103.2), and nothing in it forbids adding the rating.
4. The definition of advertising covers public persons and immovables: «Toda forma de comunicación realizada por una persona física o jurídica, pública o privada, … con el fin de promover … la contratación de bienes muebles o inmuebles» (Ley 34/1988, art. 2). The 2013 FAQ makes «el promotor o propietario» responsible, and here the owner is the public seller.
5. EU law no longer limits the duty to commercial media. The 2010 directive's English text spoke of «advertisements in commercial media», but its Spanish text said «los anuncios publicitarios que aparezcan en los medios de comunicación». The 2024 recast requires the class «in online and offline advertisements» (Directive (EU) 2024/1275, art. 20(4)); its recital 67 says «all advertisements».
6. Practice. Most TGSS notices, and some Hacienda notices, already print the rating in the gazette (§5).

Against.

1. The notice is the mandatory publication of an administrative act that opens a competitive procedure (Ley 39/2015, art. 45.1.b), and the regulation fixes its content (RD 1373/2009, art. 103.2), without the rating.
2. The definition of advertising requires the communication to be made «en el ejercicio de una actividad comercial, industrial, artesanal o profesional». A public body disposing of surplus patrimony under public-law procedures may not be exercising one.
3. No guidance mentions official gazettes or auctions.

The verb of infraction 5.a, «Publicitar», is narrower than the Royal Decree's «oferta» or «publique … información». That is not an argument against coverage. A breach of art. 15.2 or 17.3 that is not «publicitar» still falls under the catch-all infraction 5.h.

The reading this study takes is that the BOE notice is at least a «publicación de información sobre la venta» under art. 17.3, and an offer under art. 15.2. Our numbers measure what the notices contain. They are not a finding that any body has broken the law: that would need the reading above, and the facts of each lot, to be settled by the competent authority.

2.4 EU law and transposition

Directive (EU) 2024/1275 had to be transposed by 29 May 2026 (art. 35(1)), and the 2010 directive was repealed from 30 May 2026. MITECO held a prior public consultation on a royal decree to transpose it (7 January – 6 February 2026). Up to 30 September 2026, no norm published in the BOE had amended arts. 15 or 17 of RD 390/2021.

3. Prior work

What we searched. On 4 October 2026 we ran 215 searches, each recorded in data/prior_work_search.csv:

• Crossref (31) and OpenAlex (29 searches plus 3 look-ups);
• GitHub repository (24) and code (4) searches;
• 4 searches of the Tribunal de Cuentas site;
• 32 web discovery searches, used only to find URLs;
• the arXiv API: 12 queries sent, and 29 earlier queries that our fetcher refused;
• a press search: 25 Bing News queries, 5 site searches (Civio, eldiario.es, Newtral) and 5 GDELT queries, 4 of which were rate-limited (HTTP 429). DuckDuckGo answered with a challenge page, which we did not bypass.

The queries were in English, Spanish, French, Dutch, Portuguese, German and Italian, for example:

• "energy performance certificate advertisements compliance";
• "certificado de eficiencia energética anuncios inmobiliarios";
• "BOE subastas inmuebles";
• "INVIED subasta locales";
• "Observatorio del Alquiler Casa 47 etiqueta energética".

We read about 110 documents. arXiv returned 12 entries; none measures energy information in advertisements. Pages we could not read are listed in the working notes: the 2013 Commission study on certificates and prices (HTTP 403) and the MITECO monitoring report (robots.txt).

Private advertising in Spain.

• CECU. The consumer federation's survey checked 450 portal advertisements (Idealista and Fotocasa) in six cities. Its 2026 report recalls that in 2025 «el 33,8% de los anuncios no disponía de etiqueta energética». In the 2026 edition, collected in July 2025, «el 20% carecen de ella», and of those, «el 61,1% afirma tener el etiquetado energético en estado de tramitación».
• Portal studies. Marmolejo-Duarte (2016) found the rating in «sólo 4.248» of 35,116 flats for sale on a Barcelona portal in November 2014 (12%). Céspedes-López et al. (2020) found it in 9,194 of 52,939 Alicante listings in 2017–2018 (17.4%).
• Inspections. Campaigns by the regional consumer authorities report shares of compliant agencies, not of ads:
  • the Basque Country: «33 establecimientos correctos (34,38%)» of 96;
  • Madrid (2021): 5 of 50;
  • the national campaign of 2021, carried out by six regions (Asturias, Castilla-La Mancha, Extremadura, Madrid, Murcia and La Rioja): «111 59,36 %» of 187 services with irregularities, 55 of them with no label in their advertising.
• OCU (January 2026) surveyed 60 agencies in five cities and wrote that «es habitual que falte el certificado energético», with no rate.
• Reporting to the EU. Spain reported «9,084» advertising controls up to 2017. Its report adds: «No additional actions have been undertaken in order to explicitly assess compliance».

A public seller's portal. The Observatorio del Alquiler (Fundación Alquiler Seguro, with researchers of Universidad Rey Juan Carlos and Universidad Complutense) reviewed the rental ads that Casa 47, the State's public housing company, published on 8 September 2026. It reported that «Los anuncios de viviendas publicados en el portal Casa47 no muestran la etiqueta de eficiencia energética correspondiente». It also reported that the ads give a rating next to the words «calificación energética superior», which it considers misleading for dwellings rated E or F. The finding concerns the label image and its wording, not a missing letter. It covers rentals on a portal, not sale notices in the gazette, and gives no count. It is the closest precedent on a public seller we found.

Other EU countries.

• Commission study (ICF 2015). Rates reported by Member States ranged «from 100% (Austria, Slovakia) down to 13% (Estonia)», with Spain absent.
• Flanders reported that in 2019 «2,040 controls were executed … Ninety percent (90%) of the controlled advertisements complied».
• Brussels. The compliance rate «reached 84% (92% for real estate agencies and 64% for private sellers/owners)». The gap by type of seller is the closest parallel to our result.
• Brussels rental ads. Cornago and Dressler (working paper 2018, published 2020) matched rental ads to the register: «only 8% disclose the rating».

The public sector. The Tribunal de Cuentas audited the Social Security's property in 2022 and again in 2026:

• the inventory held energy data for «505 (el 13,26 %)» of 3,806 TGSS properties in 2019;
• by the follow-up, 792 properties had a certificate according to the report's text, against 510 in the previous audit (its annex table totals 712; a footnote says some have since been sold), and 286 were recorded with an expired certificate;
• the TGSS held 89 to 149 auctions a year in 2022–2025.

Neither report looks at what the sale notices say. The Tribunal's reports on Defence property sales do not mention energy certificates. The press search found many articles announcing TGSS, Hacienda, INVIED and GIESE auctions, none of which discusses the energy label in them. On GitHub, the collectors of the Portal de Subastas define no energy field, and we found no repository that parses section V-B of the BOE.

What is new. As far as this search reaches, nobody has measured whether public bodies' official sale notices carry the energy rating, lot by lot, or used BOE sale notices as data for this question. The private-ad rates above are context, not a comparison: the population, the channel and the seller all differ.

4. Data and method

4.1 Data

The BOE's summary API returned 549 daily summaries for the period (89 dates had no issue). We read the 59,211 items of section V and kept 361 candidate notices whose title combines a sale word with a property word or a known seller. The filter is deliberately wide. On a separate development period (July–December 2024) we listed every section-V title that mentioned property or a known seller without a sale word; none was a sale offer of property.

For each candidate we downloaded the notice's HTML rendering (txt.php), which robots.txt allows. We did not use the XML rendering, which robots.txt disallows, or the Portal de Subastas, which disallows everything. Requests sent our User-Agent, at most one per second. For INVIED we also read the sale pages its notices point to, on www.defensa.gob.es (§5.3).

4.2 Rules, and how blind they were

The classification rules (METHOD.md §1) were written on the development period. They were hashed and copied at 12:41 UTC on 4 October 2026, before this study downloaded any notice of the study period. That freeze was not blind. The lab's scouting pilot of 3 October 2026, which proposed the study, had already run a keyword pass over the same period. It reported per-seller rates: TGSS with the rating in 93 of 108 notices, INVIED in 0 of 9, FOGASA and GIESE in 0 of 11. It also knew the TGSS's fixed sentence. The development/study split protects the code from tuning on study-period lots, not the rules from what the authors knew. The freeze is also self-attested: the hash and the frozen copies sit in a private folder, with local timestamps only.

The classifier:

• assigns a notice kind (offer to the public, correction, annulment, result, direct sale to a named party, not real estate, lease);
• splits offers into lots (lot headers, numbered items, table rows, or one lot);
• types each lot and decides its scope as in §2.2;
• reads the energy statement.

Ratings. A rating is a capital letter A–G in a paragraph about energy, after blanking door letters («letra D», «puerta B»). Statements without a letter are classified as exempt, being processed, no certificate, or a mere reference to a certificate.

Decided in advance. A statement that a certificate exists, is being obtained, or is available in the file does not count as the label. Art. 15.2 asks for the label in the offer, and art. 17.3 for «la información relativa a su calificación».

Lenient on one point. The primary rule accepts one letter. Art. 17.3 refers to the recognised document, whose minimum for press ads is both letters (§2.1). We therefore also report the two-letter rule, which is better anchored in the text.

4.3 Deviations after the freeze

Every change is listed in METHOD.md §9. Two matter for the headline.

The parser fix (D1). After the freeze, a check for lots split wrongly (more cadastral references than lots, or a first lot label other than 1) found two faults:

• the parser read land-registry numbers («Finca número 19.732, al tomo …») as lot headers;
• it did not recognise rows numbered «1; …».

The fix corrects lot extraction, not a classification rule: the classifier's code is unchanged. Its effect falls almost entirely on one seller. The frozen parser had merged INVIED's 25 to 32 numbered premises per notice into one or two lots, or left them in the notice's preamble.

• Before the review, the fix raised the covered lots without the rating from 126 to 185, and INVIED's from 38 to 97.
• With the frozen parser and the same review outcomes, the headline would be 92 of 255 (36.1%, 95% CI 30.4–42.1%) rather than 144 of 306, and INVIED would have 34 lots without the rating rather than 86.

The frozen parser is published (scripts/lots_frozen.py) and the comparison is in data/d1_comparison.csv.

Industrial lots (D7). Art. 3.2.c excludes only parts «de baja demanda energética». Three TGSS lots of one industrial unit with a registered rating are therefore counted as covered, and a FOGASA complex with a front pavilion of unstated use as undeterminable. Workshops, warehouses, silos and agricultural sheds stay excluded.

4.4 Review

The review was done by the study agent, an AI agent, in two passes. First pass. It covered four sets:

• R1: every covered lot the classifier found without the rating (185);
• R2: a random sample of 60 of the 147 covered lots with a rating;
• R3: every undeterminable lot (137);
• R4: every excluded lot other than land (54), and a random sample of 60 land lots.

The agent read printed excerpts of each lot and the full text when the excerpt did not settle it. It recorded about 490 verdicts in about seven minutes, in nine batches. Confirmations used standard notes: all 60 R2 notes are identical, and so are 55 of the R4 notes. An independent AI agent then re-read 40 lots from the live BOE, a stratified random sample, together with every INVIED and GIESE notice. It agreed with 40 of 40 outcomes.

Second pass. Following that agent's report, the study agent:

• dropped 40 notices that the frozen rules exclude but the classifier had kept: 36 direct adjudications to a party who showed interest (Ley 33/2003, art. 137.4) and 4 expropriation notices, none with a covered lot;
• read the 10 multi-property records not yet reviewed, splitting two of them into 16 lots;
• split four tables of rustic plots into 278 lots;
• searched the roughly 2,000 unreviewed land lots for words that indicate a building, and reclassified 5 of the 31 hits;
• re-checked the industrial and art. 3.2.e cases.

4.5 Measures

• Headline. Among covered lots, the number and share whose notice does not state the rating (one letter suffices), with a Wilson 95% interval. The lots are a census of the period. The interval describes the uncertainty of a rate produced by the same process.
• Other rules.
  • two letters: consumption and emissions both required;
  • lenient: certificate references and «en trámite» count as stating the rating;
  • FAQ reading of art. 3.2.e: exemptions declared for parts of buildings count as covered lots without the rating;
  • strict: as the FAQ reading, plus every declared exemption with no valid ground;
  • distinct properties: a property announced several times counts once.

5. Results

The figures below describe what the BOE notices say. Whether a BOE notice is «oferta, promoción y publicidad» is unsettled (§2.3), so none of them is a finding that a seller broke the law.

5.1 Notices and lots

Of the 361 candidate notices, the classifier found 313 sale notices, all in section V-B. The review removed 59 that are not offers of property to the public:

• 38 direct adjudications or direct sales to a named party, 36 of them in Valencia and Castellón;
• 12 sales of harvest, timber, grazing or beekeeping rights by the National Parks agency;
• 7 expropriation notices and the approval of a road project;
• a withdrawal notice.

The remaining 254 notices hold 2,833 lots:

• 2,386 are land;
• 75 are garages or storage rooms;
• 15 were declared exempt by the seller;
• 11 are industrial, agricultural or defence buildings of low demand;
• 40 are undeterminable: ruins, undivided shares, units or buildings of unstated use, properties not described in the notice;
• 306 are covered by the Royal Decree.

Land dominates because the Delegaciones de Economía y Hacienda sell large batches of rustic plots from land consolidation. One Segovia notice alone has 226 lots.

5.2 Headline

144 of 306 covered lots (47.1%, 95% CI 41.5–52.7%) were announced without the energy rating. For 142 of the 144, the notice says nothing about energy efficiency. One TGSS notice says «No cuenta con certificado de eficiencia energética». One Hacienda notice prints the heading «Certificado de eficiencia energética» with nothing after it.

5.3 By seller

Notice content only; see the caveat at the head of §5.

TGSS: 14 of 160 (8.8%). The TGSS usually prints a fixed formula: «De acuerdo con el Certificado de Eficiencia Energética de Edificios Existentes registrado el …, el inmueble ha obtenido la calificación: consumo de energía … emisiones …». It does so in all covered lots of 89 of its 96 notices. The 14 lots without it come from seven notices:

• nine lots of two central-office notices, for Barcelona and Cabrils;
• five lots of five notices from four provincial offices: A Coruña (two notices), Huesca, Málaga and Lugo.

Patrimonio del Estado: 17 of 33 (51.5%). Practice varies by provincial office:

• with the rating: Asturias, Ciudad Real, Cáceres, Cádiz, Madrid, Pontevedra and A Coruña;
• without it: Gipuzkoa, Cantabria, Teruel, Bizkaia, Valladolid, Zamora, León, Ourense and La Rioja.

At notice level, 12 of its 24 notices with covered lots state the rating for all of them.

INVIED: 86 of 86 (100.0%).

• The notices. Its five notices sell mostly commercial premises, in a fixed format: name, registry data, cadastral reference, surface, prices. None of them contains an energy word. They send readers to INVIED's website for the technical specifications.
• The website. We read the pages of the 12 premises in INVIED's current auction (BOE-B-2026- 22413, in our data) on 4 October 2026; the listing page was «Actualizado a 1 octubre 2026». All 12 have a field «Certificado de Eficiencia Energética:», and all 12 read «En tramitación». The independent reviewer found the same field on four pages, and no rating in three technical specifications.
• What this means. For these premises, INVIED recorded the certificate as being processed. The rating was not elsewhere in the sale file for the BOE notice to repeat.
• Distinct properties. Many premises reappear in later auctions. Counted once, 78 of 78 distinct properties lack the rating.

GIESE: 21 of 21 (100.0%). Its notices list each former Guardia Civil barracks or police station in one line, with address and price, and send readers to GIESE's website for the specifications. That site could not be read (its robots.txt redirects to a page that answers 403), so the specifications may hold the rating. The same buildings return in successive auctions; counted once, they are 12 properties.

FOGASA (5 of 5) and one port authority (1 of 1) also announce covered lots without the rating.

Notice level. 135 notices with at least one covered lot were published. 101 state the rating for all their covered lots, one for some, and 33 of them for none.

5.4 Declared exemptions

The sellers declared 23 lots exempt:

• some cite no ground: «vivienda exenta», «eximido», «Certificado de exención»;
• some cite a ground that fits the lot, such as an industrial building or a garage;
• 9 cite art. 3.2.e, the exclusion for buildings bought for demolition or major reform. All 9 are Hacienda lots.

Of the 9, 7 of them are flats, premises or parts of houses (6 distinct properties). The regulation gives the declaration to «el propietario del edificio o de parte del edificio», and the seller is the owner, so the headline accepts them. The ministry's non-binding FAQ no. 7 reads the exclusion as applying only to whole buildings. Under that reading the 7 lots count as covered without the rating, and the result is 151 of 313. Counting also the declarations that give no valid ground (none stated, or a ruin that the rules treat as undeterminable) gives 157 of 319.

5.5 Other rules

• The lenient rule changes one lot (143 of 306): the notices without a letter almost never mention a certificate either.
• Requiring both letters moves 5 lots (149 of 306).
• Counting each property once gives 123 of 250.
• The largest change is the parser: 92 of 255 without D1 (§4.3).

5.6 What the review changed

Of the 188 rows in R1 (185 lots; two records were split into several properties):

• 136 of the 188 were confirmed as covered lots without the rating;
• 6 did state it. The Navarra TGSS office writes «– letra D», which the frozen rule blanks as a door letter;
• 29 were not covered: land, an industrial building, undivided shares, ruins, a castle;
• 17 were not lots of an offer, mostly numbered conditions of Andalusian regional notices.

The other sets:

• R2. All 60 lots stated a rating.
• R3. 10 undeterminable lots became covered: 8 without the rating, mostly former police stations and premises, and 2 with it. 23 rows were dropped.
• R4. 7 excluded lots turned out to be covered and rated: a flat «junto con su respectivo trastero» typed as a storage room, and the three industrial-unit lots of §4.3, among others. None of the 60 sampled land lots changed outcome.
• Second pass. The 40 dropped notices, the 16 lots split from multi-property records and the 278 plots split from tables changed no covered lot. The targeted land search made four lots undeterminable, none with a stated rating: a 60 m² building in Grañén, a former military works depot, a lighthouse plot and a ruinous house.

data/classifier_vs_review.csv and data/review_agreement.csv give the counts.

6. Discussion

Public sellers in the BOE fall into two groups:

• Sellers that state the rating as routine. The TGSS does so in nine lots out of ten, with a fixed sentence and the date of the registered certificate. This is consistent with the Tribunal de Cuentas' count of 792 TGSS properties with a certificate in 2026, up from 510. Some Hacienda offices do the same.
• Sellers whose BOE notices did not state it in any lot. INVIED, GIESE and FOGASA are in this group, and about half of the Hacienda offices.

For INVIED there was no rating to state: its own website records the certificate of the premises on sale as «En tramitación». The difference between the groups is not in the kind of property: both sell premises and flats in ordinary buildings.

The headline lies between the private-ad rates, but they are not comparable. At 47.1%, the share without the rating is higher than CECU's 20% for portal ads in 2025 and lower than the 83–88% implied by portal studies of 2014–2018. Those figures measure private ads on portals; ours measures official notices of public sellers, whose duty to publish comes from another law. The Brussels figures (92% compliance for agencies, 64% for private sellers) show that the type of seller matters elsewhere too.

Two facts temper any reading of the result as non-compliance:

1. Whether a BOE notice is the «publicidad» of art. 15.2 has not been settled by any authority (§2.3). If it is not, the label may still appear in the specifications or on the seller's website. For INVIED we checked, and it does not; for GIESE we could not check.
2. A notice that omits the rating may be selling a building with a registered certificate. The Tribunal de Cuentas counts 792 TGSS properties with one.

What the data do show is that the same information appears in some public sellers' gazette notices and not in others'.

7. Limitations

• The freeze was not blind, and it is self-attested (§4.2). The rules were written after a pilot had read the same period, and the freeze evidence is local.
• The parser fix after the freeze (D1) changes the headline from 92 of 255 to 144 of 306, almost all through INVIED. We consider the fix correct, since the frozen parser merged dozens of premises into one lot, but it was made after the freeze.
• The review was quick. About 490 first-pass verdicts were written in about seven minutes, from excerpts, with standard notes for confirmations. The independent agent's 40-lot re-reading agreed with all 40.
• Land. Land lots were checked by a random sample of 60 and by a keyword search for buildings, not read one by one. The search found five built lots typed as land, none with a rating; it may have missed lots described without those words.
• Channel. We measured the BOE notice only. For INVIED we also read the website; we did not read the specifications of other sellers, the Portal de Subastas (closed by robots.txt) or regional bulletins. Regional and local sellers mostly publish elsewhere.
• Period. 1 January 2025 to 30 September 2026, 21 months. Rates at other times may differ.
• Scope decisions. A lot was covered when its description named a building or part of one outside the exclusions. When the description did not let us decide, the lot was undeterminable (40 lots). Ruins and undivided shares, on which the guidance is silent, are undeterminable. Declared exemptions were accepted in the headline.
• The letter test. We did not check the letters against the registered certificate, its validity date, or whether it is a project or existing-building certificate (art. 15.2's second sentence).
• Re-auctions. The headline counts each lot of each notice. The distinct-property count is approximate for lots without a cadastral reference.
• Search. The 2015 recognised document on the label could not be read (robots.txt). Case law was not searched in CENDOJ. The press search could not use DuckDuckGo (challenge page) and most GDELT queries (rate limit).
• Legal reading. The question in §2.3 is open, and the result depends on it.

8. Data, code and licences

• Per-lot table: data/lots.csv, with the BOE identifier, lot, date, seller, type, scope and reason, energy statement and letters, and the classifier's and the review's values. Its columns rating_stated_<rule> say whether the notice states the rating under each rule (yes, no, n/a). They describe notice content, not compliance: whether a BOE notice is «oferta, promoción y publicidad» is unsettled.
• The review: data/review_verdicts.csv. The parser comparison: data/d1_comparison.csv.
• Aggregates: the other files in data/.
• Code (Apache-2.0) and tests: scripts/ and tests/. scripts/run.sh regenerates data/ from the downloads, and scripts/check_headlines.py checks the headline numbers.
• Our data and text: CC BY 4.0.

BOE content is reused under the AEBOE's conditions of reuse. This work is «basado en datos de la Agencia Estatal Boletín Oficial del Estado» (https://www.boe.es); it does not reproduce notice texts, and the AEBOE does not take part in it or endorse it.

9. Automation and review

AI agents (Claude models) did the work on 4 October 2026:

• they read the BOE's reuse conditions and robots.txt;
• they downloaded the summaries and notices;
• they quoted the legal texts from downloaded files;
• they wrote and tested the parser and classifier, and froze the rules;
• they reviewed the lots and computed the tables;
• they searched for prior work and drafted the text.

Three agents worked in parallel. The study agent built the data, did the lot review and wrote the paper. One agent compiled the legal and official-guidance baseline, and one ran the prior-work search.

The lot review was quick, as §4.4 describes: about 490 verdicts in about seven minutes, written by the study agent from excerpts, with standard notes for confirmations. A second pass followed the independent review.

An independent AI agent then reviewed the study adversarially:

• it recomputed every figure;
• it re-read the legal texts;
• it re-read 40 lots from the live BOE, agreeing with all 40;
• it read every INVIED and GIESE notice and INVIED's website.

Its main findings changed the study:

• the disclosure that the freeze was not blind;
• the effect of the parser fix;
• the 40 notices the rules exclude;
• the art. 3.2.e and 3.2.c readings;
• the two-letter rule;
• the description of the review.

scripts/check_headlines.py checks the headline numbers in the README and in this paper against data/.

10. Competing interests

EasyByte Hub S. Coop. Mad., which runs EasyxLab, develops validation tools for Spanish regulations, such as verifactu-lint, a checker for the Verifactu invoicing format. A validator of property advertisements against RD 390/2021, built on this study's parser, was considered internally. It is not planned under the lab's rules for building tools. EasyByte has no relationship with the bodies named, and no product or service related to energy certification, property sales or public auctions. This work may be claimed internally as research and development.

11. Ethics

The notices are public acts of public bodies about public property. Some name people: the officials who sign them, deceased persons whose estates the State inherited, neighbours in the property boundaries, and the interested buyers in direct sales. We processed the notices locally, only to classify lots, as the BOE's reuse conditions require for personal data (Regulation (EU) 2016/679). We publish no text, address, cadastral reference or name, only our classification of each lot by BOE identifier.

We used only open-data endpoints and pages allowed by robots.txt. The exception is the arXiv API, used within its terms of use (METHOD.md §2). A search engine that answered with a challenge page was not bypassed. Our log shows one pair of consecutive requests to www.boe.es 0.96 s apart. We contacted no one.

How to cite

EasyxLab (2026). Do public sellers show the energy label when they announce property sales in Spain's Official Gazette? A lot-by-lot audit of BOE notices, January 2025 – September 2026. Study S14. EasyByte Hub S. Coop. Mad. https://github.com/easybytehub/easyxlab

References

Law and official sources. Consolidated texts were read through the BOE API on 4 October 2026; they are consolidated texts of a merely informative character. RD 390/2021's consolidated text was last updated on 23 July 2026.

• Real Decreto 390/2021, de 1 de junio (BOE-A-2021-9176), as amended by Real Decreto 659/2025, de 22 de julio (BOE-A-2025-15230).
• Real Decreto Legislativo 7/2015 (TRLSRU, BOE-A-2015-11723), additional provisions 12 and 13.
• Ley 33/2003, del Patrimonio de las Administraciones Públicas (BOE-A-2003-20254), arts. 137–138.
• Real Decreto 1373/2009, Reglamento General de la Ley 33/2003 (BOE-A-2009-14788), arts. 97 and 103.
• Ley 34/1988, General de Publicidad (BOE-A-1988-26156), art. 2.
• Ley 39/2015 (BOE-A-2015-10565), art. 45.
• Directive 2010/31/EU, art. 12(4) (OJ L 153, 18.6.2010); Directive (EU) 2024/1275, art. 20(4) and recital 67 (OJ L, 2024/1275, 8.5.2024).
• MITECO, «Preguntas frecuentes RD 390/2021», version 6.0, 3 August 2021. Ministerio de Industria, Energía y Turismo, «Respuestas a preguntas frecuentes» on RD 235/2013, 20 September 2013.
• MITECO and MIVAU, documento reconocido «Calificación de la eficiencia energética de los edificios», recognised from 30 September 2026.
• Agencia Estatal Boletín Oficial del Estado, «Condiciones de reutilización» (licencia tipo of 27 June 2024), https://www.boe.es/informacion/aviso_legal/index.php.
• Tribunal de Cuentas, Informe nº 1.483 (2022) and Informe nº 1.679 (2026).
• Dirección General de Consumo, Campañas nacionales 2021, «Control de la información suministrada al consumidor en el alquiler y venta de pisos en relación con el certificado y etiquetado de eficiencia energética».
• INVIED, «Subastas en curso – Locales», https://www.defensa.gob.es/invied/02-ventas-inmuebles/subastas-en-curso/locales/, updated 1 October 2026, read 4 October 2026.

Studies and reports:

• CECU (2025, 2026). ¿Cómo de eficientes son las viviendas que queremos habitar?
• Céspedes-López, M.F., Mora-García, R.T., Pérez-Sánchez, V.R., Martí-Ciriquián, P. (2020). The influence of energy certification on housing sales prices in the province of Alicante (Spain). Applied Sciences 10(20):7129. https://doi.org/10.3390/app10207129
• Cornago, E., Dressler, L. (2018). Incentives to (not) disclose energy performance information in the housing market. ECARES working paper 2018-34; published in Resource and Energy Economics (2020), https://doi.org/10.1016/j.reseneeco.2020.101162
• ICF International for the European Commission (2015). Energy Performance of Buildings Directive (EPBD) Compliance Study. doi:10.2833/281509
• Marmolejo-Duarte, C. (2016). La incidencia de la calificación energética sobre los valores residenciales. Informes de la Construcción 68(543). https://doi.org/10.3989/ic.16.053
• Observatorio del Alquiler (2026). I y II Informe del Observatorio del Alquiler sobre Casa47. https://observatoriodelalquiler.org/estudios/i-y-ii-informe-del-observatorio-del-alquiler-sobre-casa47 (read 4 October 2026).
• OCU (2026). Survey of estate agencies, 28 January 2026.
• Concerted Action EPBD (2016, 2020). Implementation of the EPBD: country reports for Spain, Flanders and Brussels.

Cite this study

EasyxLab (2026). Do public sellers show the energy label when they announce property sales in Spain's Official Gazette? A lot-by-lot audit of BOE notices, January 2025 – September 2026. Study S14. EasyByte Hub S. Coop. Mad. https://github.com/easybytehub/easyxlab/tree/main/studies/s14-state-sales-energy-label

@techreport{easyxlab_s14,
  title       = {Do public sellers show the energy label when they announce property sales in Spain's Official Gazette? A lot-by-lot audit of BOE notices, January 2025 – September 2026},
  author      = {{EasyxLab}},
  institution = {EasyByte Hub S. Coop. Mad.},
  number      = {S14},
  year        = {2026},
  url         = {https://github.com/easybytehub/easyxlab/tree/main/studies/s14-state-sales-energy-label}
}