Georgia Tourism Overview — Q3 2026: 2.6 Million Visitor Trips, Travel Revenue Up 11.9%

Key takeaways

• International visitor trips reached 2,598,257 in Q3 2026, +3.3% y/y; tourist (overnight) trips 2,131,908, +4.7%, which GNTA calls a third-quarter record (GNTA, Ministry of Internal Affairs data).
• Over nine months visitor trips were flat at 5,332,761 (+0.1% y/y): tourist trips +2.3%, same-day trips −9.0% (GNTA).
• International travel receipts were USD 1,864.4 mn in Q3 2026, +11.9% y/y; nine-month receipts USD 3,796.1 mn, +4.4% (derived sum) (NBG, via GNTA).
• EU + UK visitor trips rose 16.6% y/y to 224,419, a derived 8.6% share of visitor trips (7.7% in Q3 2025, derived); Middle East region trips fell 22.0% and Saudi Arabia 32.8% (GNTA).
• Air-border visitor trips rose 8.5% y/y to 1,078,807, with Batumi airport +24.3%; land-border trips fell 1.8% (GNTA).
• Survey spending by visitors aged 15+ fell 8.2% y/y to a monthly average of GEL 1,900.5 mn, while balance-of-payments receipts in USD rose 11.9%; the two are different measures (Geostat; NBG, via GNTA).
• Branded-hotel occupancy reported by STR Global was 77.4% in Q3 2026 against 74.1% a year earlier, ADR GEL 318 (STR Global, via GNTA); this is not the bank panel’s measure.

1. Visitor trips

Source for all figures in this section: GNTA, International Travel workbooks (Q3 2026 and 9M 2026 files, last modified 20 Oct 2026; Q3 2025 file, 11 Nov 2025), underlying data from the Information Centre, Ministry of Internal Affairs of Georgia (“GNTA/MIA”). The Q3 2025 figures in the 2026 file equal the original 2025 file; no revisions.

Q3 (July–September)

Nine months (January–September)

In Q3 2026 growth came from overnight stays: tourist trips added 95,177 (derived) and made up 82.1% of visitor trips (GNTA), while same-day trips fell 2.8% y/y. Over nine months the same split holds more sharply: tourist trips +2.3% against same-day trips −9.0%, leaving total visitor trips at +0.1% y/y (GNTA/MIA). Geostat’s inbound survey count (monthly average 866.085 thousand visits, ×3 = 2,598,256, derived) matches GNTA’s Q3 visitor trips (Geostat, Inbound Tourism Survey, Q3 2026).

2. Source markets

Top 15 countries, international visitor trips, Q3 (GNTA/MIA)

Russia accounted for 26.8% of Q3 2026 visitor trips (derived). Saudi Arabia moved from 8th place in Q3 2025 to 11th (derived ranking from the same table).

EU + UK and selected markets, Q3 (GNTA/MIA, “EU” sheet and country sheets)

EU + UK held an 8.6% share of Q3 2026 visitor trips, against 7.7% in Q3 2025 (both derived). Geostat’s “EU” row excludes the UK: 206,186 in Q3 2026 (derived, 224,419 − 18,233) (Geostat, Inbound Tourism Survey).

Fastest risers among the top 15: Azerbaijan +36.2%, China +30.2%, Ukraine +21.3%, Israel +19.1%. Outside the top 15, Lithuania +46.8% and Latvia +25.6%. Fallers: Saudi Arabia −32.8%, Türkiye −7.4%, Armenia −4.0%; India −25.9%, Kyrgyzstan −19.6%, Kuwait −38.5% (GNTA/MIA). GNTA attributes the Gulf and India declines to the regional conflict and flight cancellations (GNTA, Statistical Data Overview 2026 III Quarter, p. 5–6). The pack gives no stated reason for the other movements.

3. How visitors arrive

International visitor trips by border, Q3 (GNTA/MIA)

Land remained the largest channel in Q3 2026 at 1,474,909 trips, but air accounted for the growth (+84,247 trips, derived). Railway Gardabani supplied 23,162 of the 34,521 railway trips. Both main land crossings, Sarpi (−2.8%) and Kazbegi (−7.7%), fell y/y.

Passengers carried at Georgian airports for Q3 2026 are not yet published (Geostat, source Georgian Civil Aviation Agency; latest quarter Q2 2026: 1,992.0 thousand, against 2,112.1 thousand in Q2 2025, −5.7% derived; Q3 2025: 2,832.4 thousand). GNTA’s air-border visitor trips (1,078,807, +8.5%) are used here as the proxy.

4. Travel revenue

International travel receipts, USD (NBG, via GNTA, “Economic Indicators” workbook, sheet “International Tourism Receipts”)

The workbook’s 2026 total (USD 3,796,085.6739 thousand) equals the sum of the three quarters. GNTA calls Q3 2026 a record quarter. Q1 and Q2 2026 were +0.5% and −3.8% y/y, so the nine-month increase of 4.4% (derived) comes from Q3.

By country, Q3 2026 (NBG, via GNTA, Statistical Data Overview 2026 III Quarter, p. 7, text)

Only these four markets are confirmed by GNTA’s text. GNTA’s accompanying chart shows shares for further markets and for Q3 2025, but its labels sum to 92.7% (2026) and 90.1% (2025), not 100%; those chart values and any Q3 2025 USD values or y/y rates derived from them are therefore not printed here. NBG’s full by-country table was not read (see section 6).

Visitor spending, Inbound Tourism Survey (Geostat; visitors aged 15+; GEL; monthly averages for the quarter as Geostat prints them)

Spending per visit: GEL 2,467.5 in Q3 2025, GEL 2,194.4 in Q3 2026, −11.1% (derived). Average nights per visit: 6.31 → 6.44, +2.2% (Geostat). Main purpose, monthly average visits, thousand, Q3 2025 → Q3 2026: holiday/leisure 492.5 → 509.0 (+3.3%); visiting friends and relatives 174.0 → 186.8 (+7.3%); transit 102.4 → 96.4 (−5.8%); business 35.0 → 39.3 (+12.3%); shopping 26.2 → 26.0 (−0.9%); health suppressed by Geostat (Geostat).

The balance-of-payments receipts (NBG, via GNTA; USD; +11.9% y/y in Q3) and the survey spending (Geostat; GEL; visitors aged 15+; −8.2% y/y) are different measures. They are not reconciled here; the GEL/USD rate for the period was not obtained.

5. Hotels

The occupancy and average room rate of large branded hotels (national average, USD, excluding VAT) in the bank’s Tourism Sector Overview is the bank’s own panel data. It is not public and is not reconstructable.

Branded hotels, STR Global (STR Global, via GNTA, Statistical Data Overview 2026 III Quarter, p. 8; licensed data, GEL)

Q3 2026 occupancy by area (STR Global, via GNTA, text): Batumi 86.6%, Tbilisi 77.4%, rest of the country 80.4%. Q3 2026 occupancy was +3.3 percentage points (+4.5%) y/y and ADR +6.7% y/y (both derived from the table).

Inconsistencies, flagged and not corrected:

• GNTA’s text says occupancy rose “3.9%”; its own table gives 74.1% → 77.4%, i.e. +3.3 pp or +4.5% (derived).
• A national figure of 77.4% equal to Tbilisi’s and below both Batumi (86.6%) and the rest of the country (80.4%) cannot be a weighted average of the three; one figure is likely misprinted.
• STR’s ADR is in GEL and its tax treatment is not stated; it is not comparable with a USD, excluding-VAT room rate.

Short-term rentals (AirDNA, via GNTA): 21,073 listings and 40,000+ beds in September 2026; average occupancy 66.2% in Q3 2026.

Hotel survey, annual (Geostat, Main indicators of hotels, last update 26 Aug 2026)

The survey publishes no occupancy rate and no quarterly series.

6. Not yet published, blocked or proprietary

• Air passengers, Q3 2026 (Geostat/GCAA): not yet published; latest quarter Q2 2026; GCAA by-airport data annual only (latest 2025).
• Foreign card spending (NBG, via GNTA, sheet “Foreign Card Operations”): latest month April 2026; July–September 2026 not yet in the file.
• Tourism value added (Geostat, via GNTA): latest H1 2026 (GEL 3,460.2 mn, 7.4% of GDP at basic prices, preliminary); Q3 not yet published.
• Hotel survey (Geostat): annual, latest year 2025.
• Monthly visitor trips for July, August, September 2026: no public monthly table found; GNTA publishes by quarter only.
• NBG full travel-revenue table by country: nbg.gov.ge returned a JavaScript bot challenge on the whole domain; not worked around.
• GEL/USD exchange rate (NBG): blocked by the same challenge.
• Tbilisi airport operator statistics (tbilisiairport.com): firewall rejected the request; not worked around.
• Large-branded-hotel occupancy and room rate (USD, excl. VAT): the bank’s proprietary panel.
• STR Global and AirDNA: licensed; only the figures GNTA publishes are quoted.

7. Sources

All accessed 6 November 2026.

How it was made

Gather. On the morning of 6 November the official files were downloaded and parsed: GNTA’s three International Travel workbooks, its Economic Indicators workbook and its quarterly overview PDF; six Geostat workbooks; and the bank’s 1Q26 issue, read for its structure only. The figures went into the input pack, one table per source, with each file’s address and update date. The pack was committed to the repository at 10:47 Tbilisi, before the bank’s nine-month issue was opened, so the time stamp shows what we had when.

Write. Part 2 of prompt 05 was then given to a separate AI agent with the input pack and nothing else: no web access, no other files. Its output is the overview, reproduced above without changes except for links in the sources table.

Score gate. Every number in the overview was matched against the input pack by script. Score gate: 517 numbers checked, 0 mismatches. Four were not literal matches and were recomputed by hand: the nine-month receipts rounded to USD 3,796.1 mn, the two differences 95,177 and 84,247, and a date. All four agree with the pack.

Run facts. 6 November 2026; Claude Opus 5.5 for both the gathering and the writing; about 110 network requests during gathering; about 15 minutes of fetching and 1.5 minutes of writing; run inside a working session, so no per-run API bill.

Known limits

• The reference report was out first. Every other release beat its official counterpart; this one is twelve days behind it. The test still holds, because the bank’s numbers were not read before ours, but it no longer shows that AI can be earlier.
• NBG was not readable. Its whole site answered with a bot challenge, so the revenue figures are NBG’s own numbers as republished by GNTA, and the full by-country revenue table and the exchange rate are missing.
• The branded-hotel block is not reconstructable. The bank’s occupancy and room rate come from its own panel. The STR Global figures shown instead are a different panel, in lari, and carry an internal inconsistency we flag rather than fix.
• No monthly split. GNTA publishes visitor trips by quarter; no public monthly table for July to September was found.
• Air passengers for Q3 are not yet out. Air-border visitor trips are used as the stand-in.

What this release is

The fifth and last report type in GeoResearch, run once. The question is the series’ question: given one prompt written in advance and only public sources, how much of a professional quarterly tourism report can an AI agent rebuild, and where exactly does it fail? Here three of the report’s four blocks were rebuilt from official data (revenue by country only in part), the fourth was not reconstructable, and the public sources gave four extra blocks the report does not carry.

A numbers-only comparison with the bank’s nine-month issue will be added to this page within a few days: their figures beside ours, series by series, with no commentary on their analysis. That comparison is the experiment’s result.