Armenia measures for the first time which banks artificial intelligence actually “sees”
On 18 August 2026, the financial marketplace AFM.am is publishing the AFM AI Visibility Index. It is the first index in Armenia to measure the probability that an AI model will name a specific bank when answering a customer’s question about a banking product.
When a resident of Armenia chooses a bank today, they increasingly ask ChatGPT or Gemini rather than a search engine. And they do not get a list of ten links. They get a single answer naming one bank, three at most. There is no second page and no scrolling in that kind of output. A bank is either named, or for that customer it does not exist.
This is exactly what the AFM AI Visibility Index measures. The first edition is built on 600 measurements collected in August 2026: 50 real user questions, asked in Russian and Armenian, across three channels (ChatGPT, Gemini and AI Overview in Google Search), in two independent runs. Seventeen Armenian banks were included in the calculation.
The key result: the market is highly concentrated
According to the AFM AI Visibility Index (August 2026, 600 measurements, 17 Armenian banks), Ameriabank ranks first with a score of 82.7 out of 100. Inecobank is second (64.7) and Evocabank third (51.2).
| # | Bank | Index | How often named | Named first | Categories with visible presence |
| 1 | Ameriabank | 82.7 | 56.7% | 26.7% | 8 of 8 |
| 2 | Inecobank | 64.7 | 49.2% | 16.7% | 8 of 8 |
| 3 | Evocabank | 51.2 | 44.3% | 9.0% | 8 of 8 |
| 4 | ID Bank | 50.6 | 41.0% | 9.5% | 8 of 8 |
| 5 | Ardshinbank | 40.5 | 37.7% | 5.3% | 7 of 8 |
| 6 | Acba Bank | 38.3 | 31.5% | 5.5% | 7 of 8 |
| 7 | VTB Armenia | 32.5 | 28.2% | 5.8% | 5 of 8 |
| 8 | Unibank | 19.6 | 17.8% | 3.3% | 3 of 8 |
The top eight positions are shown. Full data on all 17 banks has been shared with market participants in individual reports.
The market is highly concentrated. In 26.7% of all AI answers about banking products in Armenia, Ameriabank is the first bank named. The gap to Inecobank in second place is 18 index points.
By level of visibility the market splits into four groups.
The dominant player. Ameriabank is present in all eight categories and is effectively a monopolist in two of them.
The chasers. Inecobank, Evocabank and ID Bank cover every category, but they become the first bank named three times less often than the leader.
The middle. Ardshinbank, Acba Bank and VTB Armenia appear in AI answers, but only in fragments.
The periphery. The remaining banks. Their appearance in an answer is closer to chance than to a position.
Every bank has its own territory
The overall ranking hides the most interesting part. The categories are divided between different players, and the market leader is far from winning everywhere.
| Question category | Named first most often | 2nd place | 3rd place |
| Loans and purchases | Inecobank, 34.8% | VTB Armenia, 11.9% | Ardshinbank, 10.4% |
| Debit cards | ID Bank, 29.3% | Ameriabank, 24.4% | Evocabank, 11.0% |
| Credit cards | Ameriabank, 33.3% | Inecobank, 25.0% | Acba Bank, 16.7% |
| Savings | Evocabank, 25.9% | Inecobank, 14.8% | Ameriabank, 14.8% |
| Mortgages and housing | Ameriabank, 22.2% | Ardshinbank, 17.5% | VTB Armenia, 12.7% |
| Services for non-residents | Ameriabank, 66.0% | ID Bank, 8.0% | VTB Armenia, 8.0% |
| Business banking | Ameriabank, 75.0% | Acba Bank, 10.4% | Evocabank, 6.2% |
| Service and trust | Ameriabank, 31.4% | Inecobank, 25.7% | ID Bank, 14.3% |
The most contested territory is consumer lending, and here the market leader does not even make the top three. According to the AFM AI Visibility Index, in questions about loans in Armenia the AI models name Inecobank first in 34.8% of answers, three times more often than VTB Armenia in second place (11.9%). ID Bank has taken the debit card category with 29.3% of first mentions, and Evocabank leads in savings with 25.9%.
The picture reverses in the corporate segment and in questions from non-residents. In questions about business banking in Armenia, AI models name Ameriabank first in 75.0% of answers, and in questions about opening an account as a non-resident in 66.0%. These are two categories where the models effectively have no alternative answer.
Strengths: what each bank is already doing well
| Bank | Its strongest result in the AFM AI Visibility Index |
| Ameriabank | The only bank the models treat as the default answer: 47.1% of all its mentions turn into first place, the best “presence conversion” on the market. It also has the most stable visibility: only 32.5% of its appearances fail to repeat in the second run, against 80–90% for the lower half of the ranking |
| Inecobank | The outright leader in the most competitive category, lending, with 34.8% of first mentions. The only bank noticeably stronger in Armenian-language queries, 54.7% against 43.7% in Russian, which looks like a deliberate Armenian-first content strategy. In ChatGPT it beats the market leader: 51.0% against 34.0% |
| Evocabank | First place in savings, a category where more than a third of answers contain no bank at all. Full coverage of all eight categories and high visibility in Gemini at 63.0%, level with Inecobank |
| ID Bank | Leader in debit cards, the most mass-market retail product: 29.3% of first mentions, ahead of Ameriabank. The best result among the chasers in Google AI Overview at 55.5% |
| Ardshinbank | Consistently second and third in the heavy, high-value categories: mortgages at 17.5% and lending at 10.4%. Present in 7 categories out of 8 |
| Acba Bank | Second place in business banking and third in credit cards. One of only two banks in the top group that is stronger in Armenian than in Russian |
| VTB Armenia | Second place in lending, third in mortgages and in non-resident questions. One of the best results for specificity: in 64.5% of its mentions the AI quotes a real figure, a rate, a limit or a fee |
| Unibank | When the bank does make it into an answer, the AI describes it more precisely than the market average: 57.9% of its mentions contain concrete terms, against a market average of 53.7% |
Three findings that were not obvious
The more visible the bank, the less the AI knows about its terms
Ameriabank is the most frequently mentioned bank in Armenia in AI answers, yet the models quote a specific rate, limit or fee in only 42.9% of its mentions. That is the lowest figure among all 17 banks, against a market average of 53.7%. For smaller banks it reaches 80–100%.
The mechanism is simple. A large bank gets named on reputation, as the safe answer. A small bank makes it into an answer only when the model has a concrete fact about it. For the leader this hides a real risk: it gets recommended, but the recommendation cannot be substantiated, so the customer leaves to double-check with whoever does have the number.
“AI visibility” is not one number but three different picture
| Bank | ChatGPT | Gemini | AI Overview |
| Ameriabank | 34.0% | 74.0% | 62.0% |
| Inecobank | 51.0% | 63.0% | 33.5% |
| Evocabank | 20.0% | 63.0% | 50.0% |
| ID Bank | 26.5% | 41.0% | 55.5% |
| Acba Bank | 37.0% | 39.0% | 18.5% |
The visibility of one and the same bank varies between channels by as much as 3.5 times. Ameriabank appears in 74.0% of Gemini answers and in only 34.0% of ChatGPT answers, and it is precisely in ChatGPT that Inecobank overtakes it with 51.0%. Managing “AI visibility” as a single number is impossible. These are three different markets.
There is an entire unclaimed category
In questions about savings and deposits in Armenia, 35.7% of AI answers contain no bank name at all and the model answers in the abstract (84 measurements, August 2026). For the question “how protected are deposits in Armenian banks and up to what amount”, not one of the 12 measurements produced a single bank name. The same holds for questions about protecting savings from inflation and about currency exchange. This is territory that anyone can still claim.
Comment
“You cannot manage what you do not measure, and until now nobody in Armenia had measured how visible banks are in AI answers. We are publishing the AFM AI Visibility Index not as a verdict and not as praise for any particular bank, but as a ruler. The picture is only beginning to take shape, and in two or three quarters it may look substantially different. That is exactly why it matters to start counting now, rather than once everything has already settled.
The main thing the data shows is that this layer can no longer be ignored. It is not a one-off phenomenon but a trend that will only grow stronger, and leadership in it will go to whoever pays attention earliest. In developed financial markets, in the United States and in Europe, working with visibility in AI answers has already become a discipline of its own, with its own tools and budgets. In Armenia this territory is still largely free.
At the same time, every bank in this data has a category where it already beats the market leader. Nobody had simply shown it before,” says Vladimir Lefteroglo, co-founder of AFM.am.
The index will be quarterly
AFM.am will calculate and publish the AI Visibility Index every quarter. The next edition is due in November 2026. Regularity matters here: a single measurement shows a snapshot, while a series shows the direction the market is moving in, and the results achieved by those who have already started working on their visibility in AI answers.
Methodology
The AFM AI Visibility Index is calculated on a dataset of 600 answers collected in August 2026: 50 user questions across eight product categories (loans and purchases, debit cards, credit cards, savings, mortgages, services for non-residents, business banking, service and trust) × 2 languages (Russian and Armenian) × 3 channels (ChatGPT, Gemini, AI Overview in Google) × 2 independent runs. Seventeen Armenian banks were included.
The final score is calculated using the formula:
Index = 0.40 × Reach + 0.40 × Primacy (normalised) + 0.20 × Breadth
| Component | What it measures | Weight |
| Reach | share of answers in which the bank is named | 40% |
| Primacy | share of answers in which the bank is named first, normalised to the leader = 100 | 40% |
| Breadth | in how many of the eight categories the bank reaches at least 20% of mentions | 20% |
Specificity of description (whether the answer contains a real rate, limit or fee), stability between runs and citation of the bank’s own website are deliberately excluded from the index score. They measure not visibility itself but its quality and mechanics, and they form a separate diagnostic layer.
The source domains the models relied on were recorded separately. The financial marketplace afm.am ranked among the three most cited domains in the entire dataset, more often than the websites of most banks, behind only the official sites of Ameriabank and Inecobank.
Individual reports for each bank, broken down by product category, language and channel, have been shared with market participants.