Surface cut / research corpus / Central Asia and the South Caucasus

The region is watched for two things. It publishes about neither.

Central Asia and the South Caucasus are followed abroad for sanctions circumvention and for labour migration. Across twelve months of the region's own Telegram space, those two subjects together account for under two percent of what is published. What occupies the space instead is the price of living, and it is growing. This page is a surface reading of that mismatch. It takes no position on any government and names no channel, organisation or person.

4,148,141
posts read across twelve months
3,522
channels in eight national spaces
8
publishing languages in the selection

September 2025 to August 2026. Russian, Uzbek, Kazakh, Azerbaijani, Armenian, Georgian, Kyrgyz and Tajik. Markers are lexical and overlapping; they measure what is written about, not verified conduct.

What a surface cut means here

No client, no commissioned collection, no hypothesis set in advance. One selection of channels, one year, one pass. The value of this particular cut is that it inverts the usual direction of analysis: instead of asking what the region is doing, it asks what the region talks about, which is a different question and is answered from the region's own publishing rather than from coverage of it.

Six things visible from the surface

01 / The external frame is almost absent from the internal one

Sanctions, re-export and compliance language occupies a little over half a percent of everything published, and it stayed there all year. Labour migration occupies under one percent and slipped further across the twelve months. These are the two subjects that carry most of the international analytical attention paid to this region, and inside the region’s own channels they sit close to background noise. An analyst working from translated coverage and an analyst working from the region’s own publishing are not looking at the same place.

One layer down: whether the subject is absent, displaced into other vocabulary, or discussed in closed rather than open channels - three very different answers with three very different implications.

02 / A sixth of everything is the cost of living

Price, currency, tariff and inflation language is by a wide margin the largest subject in this space: about sixteen percent of all published material, and that share held steady from the first month to the last. It runs at roughly seven times the volume of foreign-policy language and close to thirty times the volume of sanctions. Where the material conditions of daily life dominate a national information space at this scale, everything political is read through them.

Labour migration
55
42% anger
Cost of living
33
8%
Government and institutions
32
7%
Elections
31
10%
Sanctions and re-export
30
10%
Protest and civic action
30
13%
Energy
29
10%
Corridors and transit
29
9%
Reactions per post    Share of reactions expressing anger

03 / Institutional language recedes and nothing takes its place

References to government, ministries, legislation and reform fell from about ten percent of published material to about seven across the year. Electoral language did not move into the gap: it held flat below one percent throughout, in a period that includes a constitutional referendum. Institutional vocabulary is contracting without a political vocabulary replacing it, and the space it leaves is taken by everyday material rather than by contest.

One layer down: whether the contraction is uniform across countries, and which channels drive it.

04 / One national subject carries as much attention as all the others combined

Kyrgyzstan accounts for about fourteen percent of everything published in this selection - as much as Uzbekistan, Kazakhstan, Armenia, Azerbaijan, Georgia, Tajikistan and Turkmenistan taken together. The ratio barely moved across twelve months. National shares in this space are stable rather than fluid, which makes any genuine movement in them a strong signal. This is a property of the channels indexed here and should be read as such.

One layer down: which channels carry each national subject, and whether the stability holds inside individual languages as well as in the aggregate.

05 / The connectivity story is background, not news

Corridors, transit, railways, ports and logistics hold a steady share of about seven percent of published material, and energy about four, with no meaningful movement in either across twelve months. The infrastructure transformation that dominates external commentary about this region appears internally as constant, undramatic, everyday reporting. Constant is itself a finding: announcements are not landing as events in the domestic space.

One layer down: which specific projects register domestically at all, and which are discussed only outside the countries they run through.

06 / Volume tripled while the mix held

Monthly publication in this selection rose from about one hundred and ninety thousand posts to about six hundred and twenty thousand across the year. The composition barely moved with it: the share of material matching regional or topical vocabulary slipped only from roughly forty-five percent to forty-one. Growth of this size normally distorts any share-based reading, and here it does not - the subjects above hold their proportions through a more than threefold expansion, which is what makes the shifts that do appear worth reporting.

What the audience did with it

Publication is half the picture. The same twelve months were read again through the response layer: reactions on each post, the comment traffic each post generated, and the composition of the reactions themselves. The two halves do not agree with each other, and the disagreement is the finding.

07 / The largest subject dominates by volume, not by heat

Cost-of-living material is the largest subject in the selection by a wide margin: it carries more published posts and more total response than anything else measured here. Per post it is unremarkable - about 33 reactions each, inside the narrow band of 29 to 33 where every subject but one sits. It is also among the least hostile: approval and affection together account for about three fifths of the response and anger for under a tenth, with government and institutional material marginally calmer still. The subject that dominates this information space dominates by volume rather than by intensity, and it is not the subject that angers it.

Cost of living
48
13%
Government and institutions
29
29% anger
Energy
28
36% anger
Corridors and transit
25
29% anger
Sanctions and re-export
24
50% anger
Protest and civic action
21
76% anger
Elections
12
44% anger
Labour migration
11
46% anger
Reactions per post    Share of reactions expressing anger

08 / The anger is concentrated in one subject, not distributed

Across seven of the eight subjects measured, the share of reactions expressing anger sits inside a narrow band between seven and thirteen percent, and the reactions each post draws sit in an equally narrow band around thirty. Labour migration stands outside both at once: forty-two percent anger, five times the corpus norm, and about fifty-five reactions per post, well clear of any other subject - on barely one percent of published material. Volume of publication and intensity of feeling point in opposite directions, and they do so through a single marker rather than across small subjects generally, which means any reading based on publication counts alone will place the pressure in the wrong subject. One caution applies: part of the hostility carried by this marker attaches to criminal cases involving migrants rather than to migration policy, and separating the two requires a narrower marker.

One layer down: the same split per country and per language, and the specific weeks in which tone moved ahead of volume.

09 / One subject splits the audience in half

Labour migration is the only marker where the response divides almost evenly: about forty-two percent of reactions express anger and about forty-three percent express approval. Every other subject leans clearly one way. A near-symmetrical split is not indifference; it is two audiences reading the same material in opposite directions inside one shared space, and it is measurable week by week.

10 / What the replies are about, and what they revise

The highest-response replies were read back against the findings above. One is confirmed, one is revised, one is refined, and one has to be qualified.

Revised - elections. The heaviest replies under electoral material are not disputes about a result. The vocabulary measured above is driven by a constitutional referendum, and the replies divide three ways: a procedural and symbolic argument over what was actually amended and which date should mark it, a rights argument about the language of instruction citing the constitution directly, and a visible layer of approving turnout messaging. The subject is contested, but the contest is over the meaning of the change rather than over the count.

Confirmed - migration. Replies divide along a single fault line: create work at home instead, against send them back where they came from. Both directions of hostility are present in the same threads. The anger measured on this marker is real and it is aimed in two directions at once.

Refined - cost of living. The affectionate reaction composition on this subject is not approval of prices. The replies that draw it are about households: wages that do not move while utilities do, pensions weighed against food, direct disputes with the official subsistence figure, and demands that the officials who publish that figure live on it. What the reaction layer records here is solidarity with the people in the material, not satisfaction with its subject. Any reading of this marker as a contentment signal would be wrong.

Qualified - sanctions. This marker carries no distinctive response signal on the full corpus, and the replies explain why. The vocabulary is lexically broad: it captures domestic restriction and prohibition debates - bans, penalties, calls for harsher punishment - alongside international sanctions and re-export, and the replies driving the response are overwhelmingly the former. The marker measures something other than its name suggests, which is why no finding is reported on it. Separating the two requires a narrower marker and is a commissioned rather than a surface-level task.

This is what reading the response layer is for. Publication counts and reaction counts establish where the pressure is; the replies establish what it is about, and occasionally that a marker is measuring something other than its name suggests. Comment findings are drawn from the highest-response replies to posts in this selection. No channel, account or individual is named, and no comment is reproduced.

Response figures are computed from the same archive as the publication figures, across all 3,522 channels in the selection and the full twelve months. Reactions are deduplicated per post and emoji before aggregation; comment traffic resolves only across the channels that have discussion enabled. Reaction composition is reported as shares within each subject, not as absolute counts.

What this page does not attempt

It reads eight publishing languages of the region and not every channel in them. It reads publication only; audience reaction is a separate layer and frequently points the other way. It reads twelve months, while the history behind it runs back years. It reads channels in isolation, while the repost graph shows who originates and who repeats. Markers are lexical: they measure the presence of language, not the truth of a claim or the occurrence of an act. Nothing here is attributed to any state, organisation or person, and no channel or account is named, by design.

Questions this environment answers when it is asked properly

Where this sits in your work

What an answer actually looks like

One subject goes in - a country, a policy, a project, a claim, a date. This is the shape of what comes back, with identifiers withheld here because this is a public page.

SUBJECT      [supplied by client]
FIRST SEEN   2026-01-11 · 4 channels · national language
DOMESTIC     0.4% of national space, week 1 → 3.1%, week 6
EXTERNAL     covered abroad from week 1; domestic pickup lagged 19 days
FRAME        infrastructure -> tariffs -> local grievance, between week 3 and 5
ORIGINATORS  [withheld] · 5 accounts · 611 subsequent carriers
NEIGHBOURS   registered in 2 adjacent national spaces, not in the other 5
Channel, account and person identifiers are delivered under agreement, never on this page.

First results in days. The same question can be asked of any of the other languages in the index, and answered backwards across years of history.

Scale

This is eight languages, one year, read from the surface, produced without asking the data a single hard question. The index behind it spans the great majority of the public Telegram ecosystem: well over a hundred languages, every populated region, more than a decade of continuous collection, growing daily, with a large discovery queue behind it. The constraint on depth is not the data. It is the question, and the question is yours to set.

What is available

Method, limitations and reproducible queries are provided to institutional partners under agreement. Work under NDA is the norm, not the exception. Nothing is published about a client engagement. NarraPrism holds no political affiliation and works for no side.

Most of what is worth knowing about this region is one question away from the surface you just read. If a line above sits closer to your remit than to your curiosity, that is the point.

Commission a cut on your question Put a subject under watch

Not ready to scope a project? Send one subject, one country or one claim and we will come back with a short read on what the environment holds on it. No commitment, no cost.

Cut reference CA-2026-09. Channel selection, queries, marker dictionaries and archive locations for every figure on this page are retained unchanged and supplied to institutional partners under agreement.

Back to overview