Where Europe's identity apps actually reach | eID Adoption Race H1 2026 [Analysis]

Key takeaways

• Adoption can be measured in 23 of the 39 markets covered here, plus the Baltic group where Estonia, Latvia and Lithuania share a single app: 24 measurements in all, covering 93.5% of the adult population behind them. Where a figure is missing, the reason is stated. Finland has no consumer app to measure at all, only a scheme run through banks. Denmark's AltID was 2 months old at the measurement date and Bulgaria's Evrotrust fell 7 days short of the activity threshold, so neither has enough data to measure. Ireland's data contradicts itself, reporting more installs than active users. Serbia's app is a vendor product sold in several countries and cannot be split out. Measurement date July 2026.
• Getting an app onto phones and getting it used are different things, and the distance between them divides Europe. For every 100 installs counted since July 2023, Latvia has 4 monthly active users and Hungary 7, against 99 in the Netherlands and Sweden. In Estonia, Latvia and Lithuania it is slightly above 100, because Smart-ID has more people using it each month than it has installs in the window: most of its users installed it before July 2023.
• The identity apps Europeans actually use mostly log them in rather than hold their documents. The 5 highest figures all belong to apps that authenticate, and every new state wallet measured here is smaller than the app beside it: in Denmark AltID has 621,058 monthly active users against MitID at 3,102,138, in Belgium MyGov.be 289,910 against itsme at 6,214,739, in Switzerland swiyu 2.6% of the AGOV base.
• A government's user count and an app's user count are not the same number, and the difference is large. The United Kingdom reported 16,600,000 One Login account holders on 13 March 2026 against 1,000,872 monthly active users of the app; Austria about 6,000,000 ID Austria activations against 1,939,941. Where an official figure counts the same thing we do, the 2 land within roughly -17% to +19% of each other, in either direction.
• Of the 27 EU member states, 7 have said when their wallet arrives and 20 have said nothing, and in none of them could a wallet be established at any certification stage from a dated official publication. None of the 4 dates still ahead describes general availability before 24 December 2026. In 18 of the 27 the certification status is not established, which in this analysis never means not started. Status cut-off 31 August 2026.
• In 18 of the 27 member states the wallet is still an unnamed application. The other 9 name one in an official publication; of those 9, only 5 point at an application that already exists and is measured here.
• Measured against their own adult populations, the 2 furthest-reaching government-run identity applications in the EU are Dutch DigiD at 37.9% and Polish mObywatel at 24.8%, and both governments say that base does not carry over to the wallet. Sweden is higher, at 35.2%, but BankID is run by a bank consortium rather than by the state; counted in users rather than share, Italy's IO is larger than either. The Dutch State Secretary states that DigiD at assurance level high is currently used by a few hundred thousand citizens. Poland's Ministry of Digital Affairs states that mObywatel provides the substantial level while the wallet is to provide high, and that mobile documents in mObywatel are not electronic attestations of attributes. Poland goes further than any other state here and writes measured adoption into the decision itself: whether mObywatel and the wallet run as one solution is to be decided partly on the degree of adoption of the EUDI Wallet.
• What reliably sends people to download an identity app is an administrative deadline. Across the 40 applications with a daily download series, we flagged 440 days on which downloads jumped well above their own recent normal, grouped them into 204 episodes, and looked for a dated public cause published on or before the day. Only 17 episodes have one, and 6 of those 17 fall on tax or filing deadlines, more than any other kind of trigger.
• One state has already moved a launch date because of AI-driven fraud. On 25 February 2026 the Swiss Federal Office of Justice put the E-ID start at 1 December 2026; on 30 June 2026 it withdrew the date without naming a new one, citing the need to harden online issuance against injected malware and deepfakes. It is 1 market, outside the EU deadline, and no other market here records a comparable statement.

Download and active-user figures are estimates sourced via https://apps.similarweb.com. They provide one comparable metric across markets and may differ from adoption or usage statistics published by national authorities. Reporting period 1 January to 31 July 2026; growth metrics run against cumulative downloads to 31 December 2025; country statuses cut off 31 August 2026; store ranking snapshot 20 August 2026; both reach metrics are divided by the adult population 18+ at reference date 1 January 2025.

What is this analysis, and what is it not?

By 24 December 2026 every EU member state has to make at least one EUDI Wallet available. Whether anyone opens it depends on habits that already exist, or do not, in the identity applications people carry today. That is what this analysis measures, and every figure in it is given with its window, its denominator and its measurement date.

Statuses were checked up to 31 August 2026. Where something material was published between that date and publication, it appears as a dated note in the text and does not enter the tables or the data file, so the whole dataset describes one date rather than a mixture.

Four things this analysis does not do. It does not assess legal compliance with eIDAS 2.0. It does not score or rank countries by readiness. It does not forecast who will make the deadline. And it publishes no comparison with the February or March 2026 editions of this series, because four things changed at once between those editions and this one, and any difference would mix real change with method change. The full reasons are in the appendix at the foot of this article.

What is being counted, and what the labels mean

Three things have to be clear before any number in this analysis makes sense: what a line in a table describes, which application it describes, and what kind of product that application is.

What counts as one market here. Every line in every table is a market, with one exception. The analysis covers 39 markets, and one line is a group of 3 countries: Smart-ID serves Estonia, Latvia and Lithuania from a single user base that cannot be split by country, so it gets its own row and is called the Baltic Smart-ID bloc. That is why the tables show 40 lines for 39 markets.

Primary and Secondary. Every market gets one application marked Primary, and that is the one whose figures become the country's figures. It is chosen for what it can do, not for how many people use it: the choice runs off 7 recorded capabilities, such as whether the app stores a state ID document, holds other credentials, logs a person into third-party services, or produces a qualified electronic signature. A Secondary application is added only where leaving it out would misdescribe the market, typically a private or bank application that dominates a country whose state app is small. Secondary figures appear in the text where they matter; they never replace the country figure and are never added to it, because the same person would be counted twice and a country could pass 100%.

What kind of application it is. The recorded capabilities also decide the label, which is derived by formula rather than assigned by opinion. An authenticator logs a person into services and signs documents, but holds nothing: AusweisApp, BankID, DigiD. A wallet stores a state identity document and other credentials on the phone: mObywatel, France Identité, eAusweise. An ID-only wallet stores the identity document alone, as Czech eDoklady does. A credential wallet holds a state-issued credential that is not an identity document, as Italian IO does. A presenter shows a credential it does not store, reading the physical card and displaying the result, as Spanish MiDNI and Estonian Eesti äpp do. Across the 54 measured applications, 32 are authenticators and 16 hold something.

That distinction matters for reading the rest: the wallet obligation is about holding and showing credentials, and most of what Europe uses today authenticates instead.

What do these metrics actually measure?

Four metrics describe four different things and they do not substitute for each other. Two further columns describe the direction of the active base, and one more describes visibility rather than adoption.

The 4 reach and usage metrics, the 2 growth columns and the visibility snapshot. Coverage is counted across the 39 markets and the Baltic Smart-ID group, 40 in all. An empty cell in any of them is never a zero: the market's own line gives the reason. Underlying data: the file at the foot of this page, sheet T0_metrics.

Two of these read more easily with an example. App Usage is the share of a month's users who open the app on an average day, so 31% is a habit and 3% is a document people fetch when something forces them to. Velocity of 16% means the install base grew by a further sixth over the 7 months of the period.

Active Reach is the adoption metric in this edition, and Install Reach read against it is the more useful pair: the distance between distribution and retained use is where the markets separate. MAU change and MAU trend answer different questions and neither corrects the other, so they are published together or not at all, and in 7 of the 21 markets that have both, they point in opposite directions.

Three rules govern how these figures may be read, and they matter more than the definitions.

Active Reach rests on a single month, and that month is the last one we have. The figure is July 2026, and July turned out to be the strongest month of the period in far more markets than chance would put there. Two explanations fit it equally well. Either something really did move in July across Europe, or the most recent month comes in high because the estimate comes from a panel of devices whose most recent readings are the least settled. The 2 cannot be separated until August data exists, and it did not exist at the cut-off.

Rather than argue about which it is, we can show what is at stake. Recalculated on a May to July average instead of July alone, the typical country figure moves by about a tenth of its value, with Norway moving most and Hungary not at all. No published figure is corrected on that basis, because switching anchors on a suspicion would break the alignment with Install Reach, which runs across the whole window. The test behind this, and the market-by-market movement, are in the appendix.

One application carries each country, and in fragmented markets that understates the country. Applications inside a country are never summed, because the same person would be counted twice. The application is chosen on what it can verifiably do rather than on how many people use it, so in 7 markets the application carrying the figure is not the busiest one.

An empty cell is never a zero. Every market without a figure gives its reason: no consumer application exists, the scheme has no application, there is too little activity data to measure use, the users are spread across several countries, or what we can measure is a vendor's product rather than a national app. The reason is given in the table itself, and the appendix explains each gate.

One limit belongs here rather than in the appendix, because it explains most of the questions we get about these figures: a scheme is not an application. Official counts of registered users run 3 to 16 times the monthly active users of the corresponding app in the United Kingdom, Austria and Greece, because they count people who registered in a browser, or credentials rather than users. The FAQ gives the figures.

Where an official figure counts the same quantity we do, the two can be compared directly. That happens in 3 markets, the gap runs from roughly minus 17 to plus 19 percent and has no stable sign in either direction, and the comparison is set out in the appendix at the foot of this article.

Which markets have real adoption, and which only have downloads?

The table below covers every market, with the Baltic group on its own line. Every figure is divided by the adult population of that market, and wherever a cell is empty the data-status column gives the reason rather than a zero.

Table 1. Reach of the principal eID application in 39 markets, plus the Baltic group. Both metrics are divided by the adult population 18+ (Eurostat, reference date 1 January 2025; UN DESA WPP 2024 for Bosnia and Herzegovina, Kosovo, Ukraine and the United Kingdom). Active Reach stands on July 2026 active users, with the caveat set out above. An empty cell is never a zero. App Usage, the 2 growth columns, Velocity and what exactly was measured in each market are in the searchable table at the foot of this page. The last column gives the reason in short; the full wording, with the thresholds each market missed, is in the data file. Underlying data: the file at the foot of this page, sheet T1_reach.

Three things in that table are worth reading before the detail.

• An adoption figure exists for 24 of the 40 lines, covering 93.5% of the adult population behind them.
• The 16 without one are missing for stated measurement reasons, not because nobody uses an app there.
• Install Reach exists for 32 of the 40, so distribution is measurable in more places than use is, and the 2 columns rarely move together. That gap is what the rest of this section is about.

What kind of application carries each country figure

The country figure is not the same kind of product everywhere. The 2 maps below show which kind it is: an authenticator, a wallet holding a state credential, or a presenter that shows a credential it does not store.

Tap or hover over a country on any map in this analysis to see which application carries its figure and what the number is.

Classification per market, status August 2026. Categories are derived from what each application can verifiably do, not from an opinion about maturity. Source: Authologic eID Adoption Race, H1 2026. Underlying data: the file at the foot of this page, sheet Map_classification.

These maps answer something the adoption figures cannot.

• The first shows whether a country's number describes a product that holds credentials or one that only logs people in. Of the 54 measured applications, 32 authenticate and 16 hold something, and the wallet obligation lands on the second group.
• The second shows who runs the identity means people actually use: 15 state-led markets, 17 dual and 2 private-led, with 6 carrying no model because no application is measured. Where the model is dual or private-led, a state building a wallet is not starting from a base it controls.

Getting an app installed is not the same as getting it used

The two reach metrics answer different questions, and reading them against each other is more informative than either alone.

Read as active users for every 100 installs counted in the window, the range runs from 4 in Latvia, where active users equivalent to 1.17% of adults sit behind installs equivalent to 27.77%, and 7 in Hungary, at 2.98% against 39.92%, through 13 in the United Kingdom and Greece, 15 in France and 16 in Czechia, up to 99 in the Netherlands and Sweden and slightly more than 100 in the Baltic Smart-ID bloc.

How much of the distribution turned into use: July 2026 active users against installs from July 2023 to July 2026. Countries without a value are missing one of the 2 metrics, and the reason for each is in Table 1. Source: Authologic eID Adoption Race, H1 2026. Underlying data: the file at the foot of this page, sheet Map_retention.

A low ratio is not evidence of abandonment. Install Reach counts reinstalls and device changes, never falls, and covers a longer window than the July 2026 active reading, so the ratio describes what the use looks like rather than how many people left.

What it describes is a market where distribution has been achieved and routine use has not, and routine use is exactly what the wallet has to build on.

The highest reach in the analysis sits outside the notification framework

Smart-ID is an authentication app run by a private company and used across Estonia, Latvia and Lithuania to log into banks and public services.

The bloc has active users equivalent to 41.1% of adults and installs equivalent to 40.1%, on 2,061,999 monthly active users in July 2026 against a summed adult population of 5,017,956, and it is not a notified eID means in any of the 3 countries. Notification is the procedure by which a state puts a means forward for recognition in the other member states, so this is a statement about cross-border standing rather than about whether the app is lawful or supervised at home. Read from the Commission overview of pre-notified and notified eID schemes, page version of 2 February 2026, accessed 30 August 2026: Smart-ID appears in none of the 3 national schemes, and no pre-notified or peer-reviewed entry exists for it.

Two readings have to be ruled out here. The figure does not rank the Baltic states as Europe's adoption leaders, because it describes one application and not a national scheme. And the absence of notification says nothing about Smart-ID's standing at home: its operator holds the status of qualified trust service provider, which is supervised under EU law. The 3 countries are measured together because one application serves all of them and its users cannot be split by country.

In 7 markets the application carrying the country figure is not the one most people use

In 7 markets a second measured application has more monthly active users than the one carrying the country figure.

Table 2. The 7 markets where a second measured application has more monthly active users than the one carrying the country figure, July 2026. The country figure stays with the application chosen on what it can verifiably do; the 2 are never added together, because the same person would be counted twice. An empty Active Reach means the application did not pass a quality gate, not that it has no users. Underlying data: the file at the foot of this page, sheet T2_second_app_ahead. User counts are rounded for reading; the exact figures are in the data file.

Belgian itsme alone has active users equivalent to 65.1% of adults against 3.0% for the Belgian Primary, and Danish MitID 63.9%. On installs the same reversal appears in Denmark, 20.8 times, and in Spain, 4.6 times.

The selection is not wrong. The Primary is chosen on what it can verifiably do rather than on how many people use it, and the two applications in one country are never summed, because the same person would be counted twice. But it means the country figure in those markets describes the more capable application, not the busiest one.

The new state wallet is smaller than the incumbent everywhere it exists

The same pattern appears in every market where both exist: the more capable application has the smaller base.

Table 3. Every market where a new state wallet stands beside an older application, July 2026 monthly active users. The 2 are never added together. Underlying data: the file at the foot of this page, sheet T3_new_wallet_vs_incumbent. User counts are rounded for reading; the exact figures are in the data file.

Sweden and Norway are heading the same way without a product yet. Sverige-id opens for applications on 1 December 2026, and the Norwegian state wallet is targeted for 2030.

Whether the incumbents hold is not something 5 markets on a single measurement date can answer, and in 3 of those markets the newer application is less than a year old.

The identity app people open most often is a bank authenticator

Seven markets pass 10%, and 5 of them are running a bank authenticator rather than a state app.

How often the application carrying each country figure is opened: daily active users divided by monthly active users, July 2026, 24 markets. The figures are panel-based estimates, not official statistics. Source: Authologic eID Adoption Race, H1 2026. Underlying data: the file at the foot of this page, sheet Map_app_usage.

Table 4. App Usage, daily active users divided by monthly active users in July 2026, for the 25 markets where both figures pass the quality gates. It says how often an application is opened, not how good it is. Underlying data: the file at the foot of this page, sheet T4_app_usage.

App Usage measures how often an application is opened, not how good it is. An authenticator embedded in payments is opened for reasons a document wallet is not, and a wallet holding a driving licence has no reason to be opened weekly.

Which bases grew, and which direction they were moving

Four of the 21 markets with this metric ran a smaller active base in the reporting period than in the previous one: Luxembourg -30.5%, Czechia -20.5%, Greece -11.6% and Sweden -6.0%. Seventeen are higher, led by Switzerland +185.0%, Spain +129.6% and the United Kingdom +54.8%.

The direction inside the period is a different story. Two of the 4 falling units were rising while the period ran, Luxembourg at +23.5% and Sweden at +12.4%. Five units run the other way, holding a larger base than in the previous period while falling through this one: Hungary -17.3%, the Baltic bloc -7.6%, the Netherlands -5.1%, France -4.2% and Croatia -0.7%.

None of this says a scheme lost or gained users. MAU describes one application, no cause is established for any of these movements, and the two columns answer two different questions rather than checking each other. Four further markets, Iceland, Slovakia, Belgium and Latvia, have no growth figure at all because their series are too noisy for any definition to describe.

The largest movement anywhere in the set needs its caveat attached to it. Swiss AGOV access shows +185.0% period over period against a trend of only +4.7% through the reporting period, in other words the growth happened before the period rather than during it. The Federal Chancellery states the increase is driven by the switch from the outgoing CH-LOGIN, with roughly 2.7 million CH-LOGIN accounts still to move and completion expected at the end of 2027, and the April 2026 login split was 59.6% CH-LOGIN against 40.4% AGOV. Switzerland did gain users; the headline growth figure cannot be read as new adoption.

One market obliges everyone to accept its identity app and publishes nothing about its use

Royal Decree 255/2025, third transitional provision, in force 2 April 2025, gave Spanish public and private entities 12 months to be ready to accept the digital DNI, so the obligation applies from 2 April 2026. Searches of the Directorate-General of the Police, the Interior Ministry, the DNIe portal and the state gazette on 25 August 2026 found no count of registrations, downloads or active users. What we measure for Spain:

• Active users equivalent to 5.1% of adults, installs equivalent to 8.3%.
• MAU change +129.6% against the previous period, the second largest of any market here.
• A trend of +47.0% through the reporting period.

The check reached named sources on one date. A documented absence of a reachable publication is not proof that no publication exists.

Velocity, and why it does not get a map

Table 5. Velocity, 22 markets. Installs in the reporting period divided by the cumulative install base to 31 December 2025, with the monthly average alongside. Velocity is blocked where an application has fewer than 24 months in the base window or where base coverage falls below 80%, which is why the United Kingdom, Spain, Belgium and Hungary are absent. No claim about European growth rates rests on this metric alone. Source: Authologic eID Adoption Race, H1 2026; SimilarWeb estimates. Underlying data: the file at the foot of this page, sheet T5_velocity.

Velocity covers 22 markets of the 40 lines and is the weakest metric in the set. The United Kingdom, Spain, Belgium and Hungary are all missing, because an application needs 24 months in the base window and 80% base coverage to qualify. A map of Velocity would look like a map of Europe while describing a minority of it, so the metric gets a table and no map. No sentence about European growth rates rests on Velocity alone.

Two maps, because one map would say something the data does not support

Active users in July 2026 as a share of the adult population 18+, split by what the measured application does. Map A covers markets whose principal application authenticates. Map B covers markets whose principal application holds a state credential. Countries with no value show the reason instead of a grey fill, and that reason is the one in the data-status column of Table 1. Spain appears on neither map: MiDNI presents a credential it does not store, and its figure is in Table 1. The Baltic Smart-ID bloc is not mapped, because one value painted across three countries would suggest three measurements; Estonia and Latvia also appear separately, with very different figures. Source: Authologic eID Adoption Race, H1 2026; SimilarWeb estimates for July 2026 against Eurostat and UN DESA population figures. Underlying data: the file at the foot of this page, sheets Map_authenticators, Map_credential_apps and Map_labels_no_value.

Read the two maps together and a contrast is visible: across the 11 mapped authenticator markets the median is 6.6%, against 4.9% across the 11 mapped markets whose application holds a credential. It is worth describing, but it does not support a claim. The 2 groups overlap almost completely, from about 1% at the bottom of each to the high twenties and thirties at the top, so a gap between medians is a contrast rather than a rule.

The appendix sets out both ranges. The Baltic Smart-ID bloc is on neither map and has the highest figure in the whole analysis, which is a further reason not to turn the comparison into one. The same discipline applies to every regional pattern here: either the data separates the groups or the pattern is described as an observation and nothing is built on it.

Visibility is not adoption, and the two point in opposite directions often enough to be a rule

The table below is a single day of store rankings for the applications that hold a credential, reported separately for each store.

Table 6. Chart Presence on 20 August 2026, applications that hold a credential only (app type Wallet, ID-only wallet or Credential wallet). Rank in Top Free, all categories, reported separately for each store. A ranking position is not a measure of adoption. Underlying data: the file at the foot of this page, sheet T6_visibility.

On 20 August 2026:

• Romania. A card reader that stores nothing, RO CEI Reader, ranked 1 in the App Store, while the notified ROeID ranked 26 there and 110 on Google Play.
• Austria. ID Austria ranked 3 against eAusweise at 5, while doing half as much.
• Italy. CieID ranked 1 and 2, with 22,571,324 installs in the window against 9,274,176 for IO, while July 2026 active users run the other way: IO 7,592,607 against CieID 5,585,231.
• Poland. The one market whose identity application ranked 1 in both stores at once.

The rankings are not wrong, and a low rank does not mean low adoption. Chart Presence is a one-day snapshot of recent download velocity, reported separately for each store, and it enters no adoption calculation.

Two things about the table above. It covers only applications that hold a credential, which is why the Romanian card reader appears in this paragraph and not in the table: a table mixing readers, authenticators and wallets would not answer one question. And that filter removes Sweden, Norway, the Netherlands and the Baltic bloc, which are the highest-reach markets in the analysis, so this section does not overlap with the adoption section and should not be read as a second view of it.

How many of these apps can show one attribute instead of all of them

Showing a shop that you are over 18 without showing your date of birth is the capability the wallet regulation is built around. The map shows where the application carrying the country figure can already do it.

The flag for the application carrying each country figure, August 2026. Not established means the national sources checked do not answer the question; it is not a no. Source: Authologic eID Adoption Race, H1 2026. Underlying data: the file at the foot of this page, sheet Map_selective_disclosure.

At least 14 of the 54 measured applications perform selective disclosure or age proof, and for 13 the flag is unestablished. The 13 are unknowns rather than negatives, and the flag describes what an application does today rather than what has been announced. Documented live retail use exists in Greece, where a seller sees only a green or red answer to an age band; in Hungary; in Slovakia, with discrete 18+ and 65+ verification; in Norway, in NorgesGruppen stores; and in Sweden, where Freja is accepted for age proof at more than 5,000 physical locations.

The reason so many flags stay open is itself a finding, and it is in the appendix: national documents usually describe the wallet a state is going to build rather than the application its citizens already use.

Where do the 27 member states stand against 24 December 2026?

The obligation binds 27 of the 39 markets here, and it falls on the state as wallet provider rather than on any user: who does what in the ecosystem decides who has to deliver what by that date. For those 27 the questions are availability and certification, and they are two axes rather than one. A reader who collapses them misreads both ends: Denmark has a wallet in the stores and no established certification stage, Finland has no application, an announced year of 2027 and certification named by its own agency as the bottleneck.

Table 7. The 27 EU member states against 24 December 2026. Availability and certification are two separate axes and are not combined. Status cut-off 31 August 2026. Cells are the opening statement of the workbook entry; the full text ships with the data file. Underlying data: the file at the foot of this page, sheet T7_EU_status.

Three things stand out in that table and each has a section below: 7 states with a launch date and 20 without, no certification stage established anywhere, and 1 wallet actually downloadable.

Seven dates, and what they actually say

The first axis is availability, and it starts with whether a state has committed to a date at all.

Whether a dated official publication gives a launch date, 27 member states, status cut-off 31 August 2026. The map records what has been published, not whether a state will meet the deadline. Romania published its own 4-stage project calendar on the cut-off day itself; it is described in the note under this section and is not in the map data. Source: Authologic eID Adoption Race, H1 2026. Underlying data: the file at the foot of this page, sheet Map_EU_launch_dates.

Seven member states have a launch date from a dated official publication.

• Czechia. The turn of 2026 and 2027, with the supplier of the client part still being chosen.
• Denmark. 3 June 2026, already past.
• Finland. 2027.
• Latvia. End of 2026, no month given, stated 4 September 2025.
• Poland. December 2026, in a pilot version, stated 4 March 2026.
• Romania. 2026, phase 1, first version of the application, stated 26 June 2026. Overtaken by the project's own calendar, published on the cut-off day; see the note below.
• Sweden. December 2026, first version in a test environment, with the broader launch dated to 2027.

Note added after the status cut-off: on 31 August 2026 the Romanian Ministry of Internal Affairs published the RO Wallet project website at rowallet.gov.ro, together with public documentation in the GitHub repositories rowallet-documentation and rowallet-pid. The application is not in the stores; the site states it will be published there at the official launch.

Certification: nothing established anywhere, and the market with the most published work explains why

Certification stages established from a dated official publication across all 27 member states as of 31 August 2026: none. In 18 of the 27 the status is recorded as not established, which is a statement about what could be read rather than about what a state has done; Finland is the single case where the agency itself says the process has not started. Sweden's Digg published a first draft certification system on 30 June 2026, opening 20 technical documents for comment, and states in the same document that the dialogue with the accreditation body Swedac has only started, that conditions for accrediting certification bodies still have to be created, and that the functional-requirements scheme depends on coming EU implementing-act revisions.

Three markets have a designated framework and no product: Luxembourg in a bill that is not yet law, Romania on a scheme taken over from Germany, Sweden on its own draft. Romania has since published a dated certification target, on the cut-off day itself; it is a target and not a stage reached, and the note under the seven dates sets it out. Where a state names a reason for arriving later, the reason it names is the EU certification and specification track rather than national capacity. That is the agencies' statement, not our explanation.

One live national wallet, and a legal layer that moves independently of it

Denmark's AltID has been downloadable since 3 June 2026 under Act No. 301 of 24 February 2026, with 281,390 accounts created at 4 August 2026 and no established certification stage; the executive orders under that act went to consultation on 30 June 2026 with responses due 7 September 2026. Ireland runs a closed pilot by invitation with a public sign-up form. Italy states a legal condition instead of a date: the opening to users arrives with completion of the national regulatory framework. The Netherlands has a dated statement that implementation will take longer than the Regulation provides, with the national implementing act going to consultation at the end of 2026.

The ordering reverses across markets. Bulgaria's draft law went to consultation on 17 February 2026 and closed on 9 April, the report on submissions appeared on 20 August 2026, and the field for the final Council of Ministers act was still empty on 26 August, in a market with no state wallet application at all. Denmark reached the stores while its own secondary legislation was still in consultation. Hungary's law designates the supervisor and delegates the naming of the wallet provider to a government decree that was not established across the issues checked.

In most of the EU there is no application whose adoption could describe the future wallet

Nine of the 27 member states name an application in an official publication: Belgium MyGov.be, Denmark AltID, Finland Suomi.fi Wallet, France France Identité on an undated official page, Ireland Government Digital Wallet, Italy IT-Wallet inside app IO, the Netherlands the publieke NL-wallet, Poland the wallet as a third pillar of mObywatel, Romania RO Wallet. Eighteen name none.

Of the 9, 5 point at an application already published and measured here. Four point at an application that is not generally available, 3 of them not published at all and one, Ireland, in a closed invitation pilot. This does not say the other 18 are behind: a state can designate a provider, a supervisory body or an accreditation architecture without naming an application, and several have done exactly that.

Several had not contracted a supplier for the client application in mid-2026. Czechia ran a competitive dialogue with 4 bidders and expected a winner during summer 2026, stated 21 April. Estonian applications were due 29 June, Finnish tenders 22 June. Lithuania still listed the choice of a technological solution among outstanding tasks on 26 March 2026. Malta published tender CT2274/2026 on 11 July 2026, estimated at EUR 15,659,714, with tenders opened on 24 September 2026 and the award date blank when read on 27 August, in the only EU market with no consumer eID application of any kind. A procurement calendar is a fact about contracting. It is the reason no adoption data exists for these future wallets, not a judgement about them.

One member state is building its wallet on another member state's national solution. The Romanian Government stated on 26 June 2026 that Romania will adopt the German solution after evaluating solutions at different stages in other member states, taking over source code, certification schemes and technical components free of charge under the open-source licence, and citing budget savings and reduced duplication. STS is named as Access Certificate Authority and Attestation Scheme Provider, MAI as wallet provider and issuer of the person identification data that a wallet has to hold, at the high assurance level, MEDAT as supervisory authority. A scheme taken over from another state is still a framework and not a stage. On 31 August 2026, the last day this analysis covers, the same ministry published the project website and its technical documentation; what that does and does not change is set out in the note under the seven dates.

The bridge between the two halves of this analysis

Measured against their own adult populations, the 2 furthest-reaching government-run identity applications in the EU are Dutch DigiD at 37.9% and Polish mObywatel at 24.8%. Two qualifications belong in the same breath. Sweden is higher, at 35.2%, but BankID is run by a bank consortium rather than by the state. And counted in users rather than in share, Italy's IO is larger than either, at 7,592,607 monthly active users in July 2026 against 7,430,378 for mObywatel and 5,587,488 for DigiD. What follows is about the 2 states whose own publications address the transfer of that base.

eIDAS grades identity means by how strongly they prove who someone is, from low through substantial to high, and a wallet has to work at high. In the Netherlands, activation of the publieke NL-wallet is to require a one-off DigiD login at that high level, and the State Secretary states that use of DigiD-Hoog is currently limited to a few hundred thousand citizens, which affects possible adoption of the wallet. DigiD has active users equivalent to 37.9% of adults in July 2026, the highest figure of any single country here.

In Poland, the Ministry of Digital Affairs states that the wallet is to provide the high level while mObywatel provides substantial, and that mobile documents in mObywatel are not electronic attestations of attributes and require a different technical structure. mObywatel has active users equivalent to 24.8% of adults.

What that means for the organisations that will have to accept a wallet is a separate question from the one measured here; it is worked through in our EUDI Wallet readiness checklist and, for the acceptance duty on private relying parties, in our legal analysis of eIDAS 2.0 and SCA.

Neither state says the existing base is irrelevant. Both plan to use the existing application as the route into the wallet. What they say is that the assurance level and the credential format do not transfer automatically, which is precisely the thing a reach figure cannot show on its own.

Poland goes one step further and writes measured adoption into a decision rule. Answering consultation submissions on bill UC122 on 17 April 2026, and repeating the formula in at least 12 separate answers, the Ministry states that the minister responsible for informatisation will have the option to provide mObywatel and the wallet within one technical and organisational solution, and that the decision will be taken on the basis of user and market needs, technological possibilities and the degree of adoption of the EUDI wallet. This creates an option rather than an obligation, and it names no adoption threshold.

What are the markets outside the EU aligning to?

Twelve of the markets covered here, and 2 more we track without measuring, sit outside the 24 December 2026 mandate. The organising question for them is not readiness but which generation of EU law they are aligning to, and how far that alignment has travelled. The first generation, eIDAS 1 from 2014, covers electronic signatures and mutual recognition of national eID schemes. The second, Regulation 2024/1183, adds the wallet. A country that finished transposing the first and has nothing on the second sits in a different place from one drafting a law for the second. A completed previous generation and a draft for the current one look similar in a status table and are not on the same ladder.

Table 8. Fourteen markets outside the EU mandate, ordered by generation of instrument, not by readiness. Status cut-off 31 August 2026. Cells are the opening statement of the workbook entry; the full text ships with the data file. Underlying data: the file at the foot of this page, sheet T8_interoperability.

Which generation of EU law each market outside the mandate is aligning to, and how far that alignment has travelled. The categories describe the instrument, not readiness: a completed transposition of the older regulation and a draft law for the current one are different places, not better and worse ones. Source: Authologic eID Adoption Race, H1 2026. Underlying data: the file at the foot of this page, sheet Map_nonEU_tracks.

Three of these markets are not late; they are not yet obliged. EU law does not apply automatically in Iceland, Liechtenstein and Norway; it has to be taken into the EEA Agreement by a decision of the EEA Joint Committee first. EEA-Lex, the register that tracks this, records Regulation (EU) 2024/1183 as an adopted act still under scrutiny for those 3 countries, with no Joint Committee decision at any stage, read 27 August 2026. Norway's own EEA note records the EFTA standard form as received 8 May 2024, due 28 August 2024 and returned blank, and Digdir states that progress on the wallet depends on when eIDAS 2.0 becomes Norwegian law. Icelandic law still implements eIDAS 1 only. Placing those 3 on the same axis as EU member states misreads all 3.

The rest split into 4 groups.

• A national instrument at eIDAS 2.0 level, at 5 different stages. Albania, law adopted 7 May 2026; Ukraine, government resolution in force since 11 June 2025; Moldova, a wallet rolling out on a published plan without a new alignment act; Montenegro, a government bill of 11 June 2026; North Macedonia, an amending act aligning to Regulation 2024/1183.
• The previous generation and nothing at 2.0. Bosnia and Herzegovina at state level, on the 2006 signature law and the 2014 electronic document law, and Kosovo, which completed its eIDAS 1 transposition with 10 implementing regulations all published on 30 December 2022.
• A live state wallet outside the framework. Switzerland and the United Kingdom.
• Neither. Liechtenstein runs an inbound recognition project, and Turkiye has nothing established. Bosnia and Herzegovina is described at state level only; the entity layer exists in press reporting alone and is outside this edition by decision.

Recognition: one arrangement in force, and it is about signatures

One mutual recognition arrangement is actually in force anywhere in this analysis. From 2026, Ukrainian qualified electronic signatures on the Commission's third-country Trusted List are recognised in Latvia as equivalent to handwritten signatures, following amendments to Latvian law, with Ukraine having recognised EU qualified signatures in 2023. This covers trust services and signatures, involves one member state rather than the Union, and does not mean Ukraine and the EU recognise each other's wallets.

Three other arrangements sit beside it, each at a different stage. Written into a statute and not concluded: Montenegro's bill of 11 June 2026 makes recognition of EU wallets and eID means conditional on an international mutual recognition agreement built on reciprocity, and no such agreement is established. The statute names the mechanism; it does not create the recognition. Begun and not closed: the United Kingdom reports bilateral engagement between DSIT and DG Connect, to be explored in 2026, while stating it does not intend to seek EU certification. Halted: Switzerland dropped the connection to international e-ID systems for budget reasons after a parliamentary cut of CHF 1.7 million for 2026.

The only dated year for starting EU certification comes from outside the EU

Moldova publishes a 3-stage plan in which 2027 brings integration of the electronic signature into the wallet and the start of the EU certification process for the instrument, stated 2 April 2026, while the review of all 27 member states returned no certification stage at all. This does not say Moldova is ahead of the EU. The start of a certification process is not a certification stage of a product, and Moldova is outside the mandate, so the two facts sit side by side rather than on one scale.

One launch date has already moved because of AI-driven fraud

On 25 February 2026 the Swiss Federal Office of Justice stated the E-ID would presumably start operating on 1 December 2026. On 30 June 2026 the same office withdrew the date without naming a new one, citing developments in artificial intelligence that require the online issuing process to be hardened against injected malware and deepfakes, and separated the trust infrastructure, expected to start in the first half of 2027, from the credential itself. This does not say Switzerland has abandoned the E-ID, and the reason is not generalised to any other market. No EU market in this analysis records a comparable statement.

The largest measured active base in the analysis is Turkish

Turkey's e-Devlet Kapisi has 16,612,389 monthly active users in July 2026, active users equivalent to 26.0% of adults, installs equivalent to 63.2% and App Usage of 11.0%, against a UN-sourced adult denominator. The operator publishes 69,358,782 registered users at 31 July 2026. Turkey is covered here for interoperability reasons, the registration count and the monthly active count measure different quantities, and neither puts Turkey in an EU adoption ranking.

FAQ

What is the EUDI Wallet deadline, and what does it actually require? Regulation (EU) 2024/1183 requires every EU member state to provide at least one European Digital Identity Wallet to its citizens by 24 December 2026. The obligation is on availability: a state has to make a wallet available, and nothing in the regulation sets a target for how many people use it. Certification of a wallet is a separate track, and as of 31 August 2026 no member state could be established at any certification stage from a dated official publication.

Which EU countries have said when their wallet arrives? Seven, from dated official publications: Czechia at the turn of 2026 and 2027, Denmark on 3 June 2026, Finland in 2027, Latvia at the end of 2026, Poland in December 2026 in a pilot version, Romania in 2026 in a first version, and Sweden in December 2026 in a test environment with the broader launch in 2027. The other 20 have published no date. This analysis records what states have published; it does not predict who will meet the deadline.

Does "Active Reach 24.8%" mean a quarter of Poles use mObywatel? No. It means monthly active users in July 2026 equivalent to 24.8% of the adult population. The measurement describes devices, not people, and one person with two devices counts twice. It also stands on a single month, and July 2026 is the last month in the series, which is why the July caveat applies to every Active Reach figure.

Why is the figure for my country lower than the number my government publishes? Because the two usually count different quantities. A scheme count includes people who registered in a browser and never installed the application, and a credential count includes documents rather than users. The United Kingdom, Austria and Greece each show gaps of 3 to 16 times on exactly this difference. Where an official figure counts the same quantity we do, the gap runs from about minus 17 to plus 19 percent and has no stable sign.

Why does a country have several identity apps and only one figure? Because summing them would count the same person twice and can exceed 100%. Metrics are computed for one application per market, selected on what it can verifiably do rather than on how many people use it. In fragmented markets the country figure understates total eID adoption, and those markets are flagged in Table 1.

Does a top-ranking app mean better adoption? No. A store rank is ordinal and reflects recent download velocity, so a saturated application with low current downloads ranks low without losing a single user. Rankings are reported separately per store and enter no adoption calculation.

How many of these apps do selective disclosure or age verification? At least 14 of the 54 measured applications, with the flag unestablished for 13. The 13 are unknowns rather than negatives, and the flag describes what an application does today rather than what has been announced.

Why is there no comparison with your February and March 2026 reports? Because 4 things changed at once between them: the population source, the download window, the store scope and the application selection. The main metric in this edition has no predecessor in those editions. Any difference would mix real change with method change, so we publish none.

Can I use these numbers? Yes, with attribution to "Authologic eID Adoption Race, H1 2026" and a link to this page. The data file linked at the foot of this page includes every table, the country data behind them and the full source list.

Appendix: how this was measured, in full

This appendix sets out the detail the body of the analysis leaves out: the population figures, every quality gate, the work behind the download spikes, what could not be read, and why no comparison with the earlier editions is published.

The denominator

Adult population 18+ from Eurostat, demo_pjan (TOTAL) minus yth_demo_010 (Y_LT18), reference date 1 January 2025. Eurostat published 1 January 2026 headline totals in July 2026, but every Eurostat dataset with an age breakdown still ends at 2025, and we do not build a 2026 population from 2025 children. Four markets fall back in full to UN DESA World Population Prospects 2024, medium variant, broad age groups 18+: Bosnia and Herzegovina, Kosovo, Ukraine and the United Kingdom. No country mixes 2 sources. The Ukrainian denominator comes with a wartime caveat that applies to every Diia figure: it is a projection from July 2024 assuming a partial return of refugees.

Two markets appear twice in the denominators and this is not double counting of a metric. Estonian and Latvian adult populations serve as their own national denominators and also sit inside the summed denominator of the Baltic Smart-ID bloc, because Smart-ID serves Estonia, Latvia and Lithuania and its user base cannot be split by country.

The download scope is worldwide. National eID applications are downloaded almost exclusively by people connected to that country, including those living abroad, and that is a stated assumption rather than a defect. Where it stops holding, the figures are withheld rather than published with a caveat.

The quality gates, and what each one removes

The gates are written into the formulas rather than applied by hand, so no case is settled by judgement and the raw inputs stay visible. The last of them, the noise gate, needs a sentence in plain terms: it measures how far a market's monthly readings jump around their own trend line, and where they jump too far, any growth figure would describe the jumping rather than the market, so no growth figure is published.

Table 9. The quality gates, written into the formulas rather than applied by hand. Where a market fails a gate, the data status column of Table 1 names the gate and gives that market's own value. Underlying data: the file at the foot of this page, sheet T9_quality_gates.

One inconsistency is left standing on purpose. The Icelandic series is too noisy for a growth figure and still has a reach figure, because Active Reach stands on a single month by design and the noise gate does not reach it. Extending the gate to a single-month metric would be a third method change made in one week, so the question goes to the January 2027 edition together with the July question. Both ask how much a single month can be made to carry.

The two growth columns in full

The period comparison sets the monthly average of January to July 2026 against July to December 2025: 6 and 7 observations with midpoints 6.5 months apart. The trend is a log-linear fit across the 7 months of the period, expressed as total change.

The pair replaced a reading that compared 2 single months, December 2025 against July 2026, dropped on 28 August 2026. That version rested on 2 observations, changed sign for 5 of 25 markets against a comparison of whole periods, and was less than half as stable: its signal-to-noise ratio was 0.62 against 1.47 for the replacement, computed on a median monthly residual of 0.204 log points. Two variants were tested and rejected: year on year, which would have dropped 3 markets including 2 of the largest movements, because their monthly active user data starts in 2025; and a symmetric 7-against-7 window, which drops the United Kingdom by 1 month.

The July anchor, measured rather than argued

The estimates behind these figures come from a panel, a sample of devices from which activity for a whole market is inferred, and the reading for the most recent month in a panel is the one most likely to move as the sample settles. July 2026 is the peak month of the reporting period in 12 of the 25 measured markets. If the peak fell at random, it would land on July in about 4 of them. The typical July sits 7.5% above that market's own median for the period. The control runs on the same January to July window in earlier years and returns 2 of 19 for 2024 and 6 of 22 for 2025, so the tilt is not seasonal in this dataset. It is present but weaker in downloads, which run through a different estimation path on the same panel: 8 markets of 25 rather than 12. The strongest case is Norway, where July stands 1.96 times its own period median.

Recomputing Active Reach on the May to July 2026 average instead of July alone, for the same 24 measurements and the same population figures, moves the median figure by 10.3% of its value. The extremes run from Norway, 25.1% to 15.8%, down to Hungary, which does not move at all. This is a diagnostic and no published figure is recalculated on it. Adopting a second anchor on a suspicion would break the alignment with Install Reach, which stands on the full window. August 2026 data did not exist at the cut-off, and this is the first check of the January 2027 edition.

The two maps, and why the contrast is not a rule

Split by what the measured application does, the 11 mapped authenticator markets have a median Active Reach of 6.6% and the 11 mapped markets whose application holds a credential 4.9%. The medians differ. The groups do not.

Authenticator markets run from 1.2% in Latvia to 37.9% in the Netherlands. Markets whose application holds a credential run from 1.1% in Slovakia to 24.8% in Poland. Almost the whole of one range sits inside the other, and both contain markets at the bottom and near the top. Adding the Baltic Smart-ID group, which is not mapped, lifts the authenticator median to 13.0% and changes nothing about that overlap.

With 11 markets on each side and ranges that sit on top of each other, a gap between two medians is not something this analysis can call systematic. The maps are published as a contrast worth seeing, and nothing in the analysis is built on them.

How close the estimate runs to official counts

Three markets publish an official figure that counts the same quantity we do, so the two can be read against each other directly. They are published together or not at all.

Table 10. The 3 markets where an official figure and our measurement count the same quantity. The Hungarian and Estonian official figures are thresholds ("more than 3 000 000", "approximately 270 000") and count people rather than downloads, so both gaps are lower bounds. The three markets are published together or not at all. Underlying data: the file at the foot of this page, sheet T10_official_vs_measured.

The gap runs from roughly minus 17 to plus 19 percent and has no stable sign, so no sentence in this analysis says the estimate runs high or runs low, and no correction factor is derived from it. Three markets are not a calibration.

Two things follow from that, and both cut in the same direction. The estimate is the only measurement in this project that is built the same way in all 39 markets, which is exactly why it is used: an official count exists in a minority of markets, counts a different quantity in most of them, and is published on each state's own schedule. And where the estimate behaves in a way we cannot explain, as with the July anchor below, the honest response is to publish the diagnostic with the figure rather than to quietly correct it or drop the market. These notes describe how a panel-based estimate behaves at the edge of its window. They are published so that the reading can be checked, and so that anyone who measures the same thing better can say so.

Where a figure is withheld, and why

Six markets have no adoption figure of their own. Albania and Kosovo, where the state application is a service portal doing none of the 7 things we test for. Finland, where the Finnish Trust Network is a scheme without a consumer application. Malta, which had no consumer eID application at all in August 2026. Lithuania, whose adoption sits in the Baltic group. Serbia, where what we can measure is a vendor's product sold in several countries at once. Four of those 6 have a notified national eID scheme; what is missing is a consumer application this method can measure.

Two markets have no reach figure, because a national application serves users who do not live in the country. Luxembourg's LuxTrust Mobile has 526,028 monthly active users in July 2026 against 553,628 adults, which would be an Active Reach of 95.0%, and 730,166 installs in the window, which would be 131.9%. Liechtenstein's eID.li would show an Install Reach of 133.1% against 33,845 adults, and its download series covers only 76.9% of the reporting period, below the 80% gate. The numerator is worldwide and the denominator is resident, so the 2 do not describe the same population.

Serbia is the one case where the reason first recorded turned out to be wrong. ConsentID has 768,868 monthly active users in July 2026 on a worldwide series published by AET Mobile B.V. Serbia's ITE reported more than 1 million ConsentID activation parameters issued on 9 December 2024 and about 2.8 million registered eGradjani accounts on 14 July 2026. ConsentID is the Serbian high-assurance login means, supplied white-label. The series is excluded because one entity serves clients in several countries and cannot be apportioned, not because the product is foreign to the market.

How we tried to explain the spikes, and how often it worked

Scale first, in one place. We went through 206 store-ranking entries across 36 markets and found 84 country-and-application pairs relevant to identity. Of the 101 applications on the list, 54 were measured, 35 of them as the application carrying a country and 19 as a second application beside it. The analysis measures 39 markets, and the Baltic group takes a line of its own, which makes 40 lines in the measurement tables. Montenegro and North Macedonia are tracked for status only and appear in the status table, not in the measurement tables. Each application has up to 37 months of data, 1,666 monthly readings in all, built from 108 daily data pulls. 440 spike days group into 204 episodes, of which 17 have a documented dated cause. Behind all of it: 445 sources, 37 official figures across 17 markets, 41 country status entries and 298 recorded findings.

That last pair is the point rather than a footnote. 17 of 204 is what survives a rule that a cause must be a dated source published before or on the day of the episode. Of the 17, 6 sit on tax or filing calendars, 4 on the withdrawal or repricing of a competing identity route, 3 on an identity capability being taken up, 2 on national emergency-alert events, 1 on an election being called and 1 on an operator feature release. Administrative calendars move these applications more reliably than identity events do.

The method cannot resolve markets running below roughly 1,000 downloads a day, and on 21 dates 5 or more countries spike at once and are excluded as a panel effect. An unresolved episode is not an absent event.

The strongest documented case and the clearest failure of the method look alike. In Spain, MiDNI spikes from 29 March to 2 April 2026 at up to 3.73x and again from 6 to 16 April at up to 7.45x, against an acceptance obligation dated a year earlier. In Austria, the ID Austria relaunch of 19 June 2026 made the eAusweise platform reachable from the ID Austria start screen and eAusweise spikes 9 days later, but that episode falls on 28 and 29 June 2026, the 2 heaviest panel dates in the dataset, so the change cannot be separated from the panel. A dated source preceding an episode is the strongest attribution this method allows, and it remains a coincidence with a dated event. Where a timeline matched but no source was dated, the mechanism was recorded and not published.

One question stays open in the daily data. Across 440 spike days, 138 fall at a weekend. That is 31.4% against the 28.3% a calendar would give, a difference too small to build on. Measured against each application's own median for the same weekday, the weekly pattern spans 0.75 to 1.40, and 2 applications run the opposite way. This does not say people download government applications more at weekends, and it does not say they do not. Separating a modelled weekly profile from a real one needs a fresh daily pull that was not run for this edition.

What could not be read

Nine official routes were still closed to automated reading at the end of the status review, out of 11 opened across 36 reviewed markets. Recorded barriers: bot detection, robots exclusion, script and cookie walls, and one HTTP 418. Each blocked route was logged with its address and the exact fact it would answer, and queued to be opened by hand in an ordinary browser; 2 were opened that way before the cut-off and 9 were not. This count measures our reach and not the silence of states. A blocked route, a silent state and a page nobody read look identical in a status table, so every affected cell says "not established", never "not issued", and states the scope of the check in the same sentence. One of the 3 earliest routes, in France, turned out to hold the single most load-bearing fact in its market once it was opened by hand.

What states publish about selective disclosure, and what they run

Of the 17 applications whose selective-disclosure flag was worked through national sources, 13 were left unanswered with a written reason, and the recurring reason is that the national document describes the future wallet rather than the application people use. Romania describes selective disclosure precisely, for RO Wallet. Slovenia does the same for eDenarnica, and Spain publishes a full age-credential specification, for Cartera Digital Beta. Where a state publishes a current feature list for a product that exists, the answer is available and negative, as for GOV.UK One Login and AGOV access. The gap between what is specified and what is running is the finding here, and it is not a gap in this research.

Why no comparison with the February and March 2026 editions

Four things changed at once between those editions and this one: the population source, the download window, the store scope and the application selection. The population change alone moved the denominator by more than 3% in 17 of 38 markets, and by up to -29.8% for Moldova. Store scope alone would move Norway by 32 points on the same application. The main metric used here, Active Reach, has no predecessor at all, because the earlier editions did not publish it. Eight candidate definitions were tested against the Czech series and none reproduces the -14.6% published in March; what the current data shows for Czechia is a fall concentrated in the first quarter of 2026, from 27,690 downloads in December 2025 to 16,065 in February 2026, recovering to 41,714 in July 2026.

None of this says the earlier figures were wrong. It says their window, store scope and denominator are not documented well enough for any difference to mean anything, so no market-level or aggregate comparison is published.

What this edition leaves out by decision

The citizen awareness thread from the two earlier editions is not in this one. It rests on two legs instead of three: 37 months of downloads and monthly active users for 54 applications across 39 markets, and country statuses sourced to dated official publications. The thread was dropped because it would have rested entirely on third-party studies whose reliability this analysis does not verify. It is not true that no awareness data existed.

Sources and data

Every figure here points at a dated source. The full source list ships with the data file, together with the tables above, the country data behind them and the application chosen for every country.

Ready-to-cite figures

Cite as: Authologic eID Adoption Race, H1 2026, measurement date July 2026, country statuses cut off 31 August 2026.

• Adoption, measured as monthly active users against the adult population, exists for 23 of the 39 markets plus the Baltic group, covering 93.5% of the adult population behind them. July 2026.
• Highest single-country Active Reach: the Netherlands, DigiD, active users equivalent to 37.9% of adults. Highest of all: the Baltic Smart-ID group at 41.1%.
• 7 of 27 EU member states have a wallet launch date from a dated official publication; 20 have none. Status cut-off 31 August 2026.
• Wallets at any stage of certification, established from a dated official publication across the 27 member states: none. In 18 of the 27 the stage could not be established either way.
• 204 episodes of unusually high downloads across the 40 measured applications, of which 17 have a cause in a dated public source published on or before the day. 6 of those 17 fall on tax or filing deadlines.
• At least 14 of the 54 measured applications can show a single attribute, such as being over 18, instead of a full identity. For 13 the national sources checked do not answer the question.

Explore the data

The table in the body carries 5 columns. Every metric for every market, including App Usage, both growth columns, Velocity and what exactly was measured in each market, is in one interactive table with search and sorting: open the full table (Datawrapper, built from sheet DW_T1_adoption).

Data file: every table and map on this page, plus the country data, the application chosen for each country and the full source list, in one spreadsheet:

About Authologic. Founded in 2020, Authologic is the trust infrastructure provider for regulated businesses verifying and authenticating users. It orchestrates three generations of identity through one integration: legacy identity verification, today's national eIDs and tomorrow's digital wallets.

Independence. None of the applications measured here is an Authologic product, and Authologic had no role in the publication decisions of any authority cited. Authologic sells the infrastructure that regulated businesses use to verify and authenticate their users, so it has a commercial interest in this subject, which is why the method is set out in full above and the underlying data is published with the analysis: every figure can be checked without taking our word for it.

Changelog

• 3 September 2026 - Initial publication. Carries one note added after the status cut-off, on the Romanian RO Wallet website and documentation published on 31 August 2026. Reporting period 1 January to 31 July 2026, country statuses cut off 31 August 2026, store ranking snapshot 20 August 2026.