Which Azerbaijan product your passports allow: the 2026 source-market matrix

Birtour is a Baku-based DMC, and this is the table we keep open when an operator asks what they can actually build. Twenty-nine source markets, six columns: entry regime, expiry date, whether that passport can be sold a Baku to Tbilisi rail leg, direct air access to Baku, the guiding languages we staff, and 2025 arrival volume. The rail column is the one nobody else publishes, and it is the one that decides whether a Caucasus circuit carries a surface sector or has to fly it.

Why this table exists

Because the two facts that decide Azerbaijan product design live in different places and nobody has ever put them side by side.

The first is the entry regime, which is well documented if you know to read the Ministry of Foreign Affairs register rather than a travel blog. The second is that since 25 May 2026 the Baku to Tbilisi sleeper sells tickets only to Azerbaijani citizens and to nationals who enter Azerbaijan without a visa. General travel media reported the line's return as though anyone could board it. They cannot.

Put those together and the world splits in two on the rail question. One set of passports can be sold a Baku to Tbilisi sleeper. The other set cannot, in either direction, and finds out at the border rather than at the ticket window. Which side your client sits on is not a preference or a budget question. It is their passport.

The road exit is the part people then get wrong by inference. It stays open to everyone who flew in, e-visa included, so an e-visa circuit is not forced onto an air sector. Both facts are in the table below and they do not move together.

The 2026 source-market matrix

Read the rail column first. Everything else on the page follows from it.

Dashes are honest. Where the State Statistical Committee publishes only a percentage share for a market, we do not back-calculate a five-digit number and print it as though it were counted.

How to read a row

Four columns change what you can sell. The other two tell you whether the market is worth the effort.

• Entry regime plus expiry together. A regime with a date in bold is a pilot, not a treaty. It caps entries at three in the window and requires a visa on the fourth, and the pilots are tourism-only, so a corporate or MICE group on a Japanese, Korean, Bosnian or Singaporean passport still needs a visa. That last point catches people.
• Rail leg sellable. Yes means you can put a Baku to Tbilisi sleeper in the itinerary. No means the Baku sector is an air leg, and you should be pricing an extra airport day into the circuit.
• Direct to Baku. A market with a nonstop supports a shorter product. A market routing through Istanbul or Dubai needs a day at each end that the brochure has to absorb somewhere.
• Guiding we staff. "On request" means confirmed per booking against actual dates, not a standing bench. We would rather write that than let a language column imply a permanent team.

The rail rule, stated once and precisely

From 00:00 on 25 May 2026, rail entry to and exit from Azerbaijan across the Georgian border is permitted only to Azerbaijani citizens and to nationals of countries entitled to visa-free entry. ADY sells tickets on that basis and the border checkpoint applies it.

Three things follow that most write-ups get wrong:

• It is symmetric. The decision covers entry and exit. An e-visa client cannot board train 38 out of Baku any more than train 37 into it. The widely repeated "arrive by air, leave overland" workaround does not apply to the train.
• The road rule runs the other way, and merging the two is the costliest error in this subject. A client who flew in may leave overland by road without permission: the foreign ministry states that foreigners are allowed to exit through land borders without the authorisation land movement otherwise requires. So an e-visa passenger who cannot ride the train can still be driven out at Balakan/Lagodekhi or Red Bridge. Reasoning from the train to the road produces a wrong answer and a missed connection.
• It is enforced. A Spanish passport holder travelling on a valid e-visa was removed from the train at the border in late May 2026.

The service runs daily both ways, train 38 Baku 23:10 to Tbilisi 08:41 and train 37 Tbilisi 21:00 to Baku 06:24, on Stadler stock of roughly 200 berths, from about 81 AZN one-way. ADY prices international fares off the Swiss franc, so budget 85 to 190 AZN by class and date rather than quoting a fixed figure. Confirm with the carrier before you sell it; ADY's own portal would not serve us a page to verify against.

What each column changes about the product you can build

A visa-free, rail-eligible market can be sold the widest product in this market: a genuine multi-country Caucasus circuit with a surface sector, no visa lead time, and no per-passenger visa line in the quote. Russia, Türkiye, Kazakhstan, Uzbekistan, China, Japan and Korea all sit here. Most of them are also markets where nobody has built the trade content, which is not a coincidence.

An e-visa market carries two cost lines a competitor may have left out: USD 29 per passenger, and a three-working-day lead time that pins your latest sale date. On routing it is more open than most people think. The train is closed to these passports, but the road exit is not, so a Madrid or Warsaw circuit can still run Baku first and cross to Georgia by road at Balakan/Lagodekhi. What it cannot do is the reverse: a "Caucasus in full" itinerary that draws a land line into Baku cannot be delivered, and we have re-quoted operator programmes this year that arrived with exactly that.

A pilot market carries a clock. The commercial consequence is not that the window closes; it is that a brochure printed for a 2027 season may outlive the regime it assumes. Print the expiry date next to the claim, or write the claim so it survives.

A market with no nonstop is not a market to skip. It is a market where trip length has to absorb two connection days, which usually argues for a longer product rather than a cheaper one.

What the demand data actually says

Arrivals fell in the first half of 2026: 1,092,600, down 10.3% on the same half of 2025, after a 2025 that was itself down 2.1% at 2,570,212. That is the honest headline and we would rather lead with it than with a growth story.

Underneath it the picture is not uniform, and the exceptions are where the opportunity is:

Two markets moved the other way hard enough to mention. India fell more than two thirds, with Q1 2026 at 17,373 against 54,840 a year earlier, and we would not build an India programme on this quarter's numbers. Germany and the UK each reached 1.1% of arrivals and entered the top fifteen for the first time, from below the reporting threshold a year ago.

One measurement note, because it changes how you read the table. The Statistical Committee splits arrivals by purpose, and the split is dramatic. Saudi Arabia is 97.7% tourism purpose. India is 93.9%. Türkiye is 47.7% and Iran is 8.7%. A headline arrival count badly overstates the addressable leisure volume in the two largest non-Russian markets, and any competitor ranking source markets on raw arrivals is ranking them wrong.

What we deliberately left out

• The visa detail itself. Entry caps, the fourth-entry rule, the tourism-only restriction and the Gulf resident visa-on-arrival route are held on our Azerbaijan visa rules for group travel page, which owns that layer and its review cadence. This page reproduces status at a glance and links out rather than keeping two copies that can drift apart.
• The urgent e-visa fee. The portal publishes the USD 20 state fee and the USD 9 service fee and does not differentiate the three-hour track. Reseller sites quote USD 108 to 148, which is their markup rather than the government rate. We are not printing a number we cannot source.
• Astana to Baku frequency. Aggregators disagree between twice weekly and near daily, and the near-daily listings look like a codeshare rather than capacity. Ask us at quote and we will read the live timetable.
• Jordan. Azerbaijan and Jordan signed an ordinary-passport visa exemption on 8 July 2026, but it has not entered into force. Jordan is an e-visa nationality today, and we will move the row when the agreement does.
• Any claim that a pilot will renew. None of the four has been extended yet. Treat the dates as dates.

What we re-verify, and when

A wrong row here poisons every page that links to it, so this one carries a review calendar rather than a vague promise.

• 1 December 2026: the Hong Kong, Macao and Gulf pilots, ahead of their February expiries.
• 1 March 2027: the whole matrix, including air access, which decays fastest.
• 1 June 2027: the Japan and Korea window ahead of 1 July, and the Bosnia and Singapore window ahead of 1 August.
• At each quarantine extension: the rail rule and the land-border position, currently set to 06:00 on 1 October 2026.

Air access is the column that goes stale fastest. Baku to Shymkent went from twice weekly to three times weekly within a fortnight of launch, and FlyArystan's Aktau season ends on 23 October 2026. Treat frequencies here as of August 2026 and ask at quote.

Twelve of these markets now have their own page, because the rail column is only one of the things that differs: Türkiye, Russia, Kazakhstan, Uzbekistan, South Korea, China, Japan, Israel, the United Kingdom, Germany, Poland, Italy, Spain and US luxury advisors.

For operators pricing a specific circuit against these constraints, the routing consequences are worked through on the Caucasus circuit operator guide. For an agency ready to contract the Azerbaijan leg, the commercial terms are on the partner programme page.