The Agentic Commerce Readiness audit is a FlexPricer product. I developed the methodology and run every audit myself. It measures how visible and legible the online storefront of a pharmacy chain, manufacturer or distributor is to AI assistants: whether the data and the site are machine-readable, what assistants say about you and your competitors, and whether the path to purchase complies with pharmaceutical law.
When a shopper asks Alice AI where to buy a medicine in their city, the assistant builds its answer from Yandex Products and search results. Yandex's help pages say so. A pharmacy chain whose range is missing from Yandex Products may simply not be named. A stale feed puts an old price into the answer; outdated map listings put in the address of a pharmacy that has closed. Both problems show up only in the answers themselves, so I collect and check them.
Assistants already pick offers and, where the law allows, take the shopper through to checkout. Purchases like these are called agentic commerce. On 27 February 2026 Yandex announced YCP, its "agentic commerce protocol": the shopper moves from the chat to Yandex's checkout, and the store handles the sale.
I work with pharmacy chains, manufacturers and distributors. For a pharmacy chain the key measure is how often the assistant names it as a place to buy. For a manufacturer or distributor, it is how often assistants name their products.
FlexPricer is an online price-list and pricing service for pharmacy chains. It selects products for the online channel, provides daily prices for them and shows competitors' prices on pharmacies' online storefronts. FlexPricer answers the question of which products to list online and at what price. The audit checks whether a shopper will see that offer in an AI assistant's answer and whether the price shown there is right. More about the service at flexpricer.ru.
Product pages, crawler access, page markup, feeds, presence in Yandex Products, on maps and on pharmacy aggregators, and the path from the assistant's answer to the cart or a reservation. All of this is under the pharmacy chain's own control. The output is a readiness score from 0 to 100 under my methodology. The methodology is my own: there is no industry standard for scores of this kind.
I put the same set of "where to buy" queries for your products and cities to the assistants: Alice AI and GigaChat, and, by agreement, ChatGPT and DeepSeek. I record whether you are named, whether you are recommended as a place to buy, whether there is a link, and whether the price, availability and address are correct. Every share comes with its margin of error and the number of answers, next to the same shares for your competitors.
I check product pages, feeds and the purchase path against pharmaceutical law and platform rules. Systematic mismatches in price and availability between the page, the feed and the cart belong here too. In the methodology this layer is called the compliance gate, and it works as a cap: a serious violation keeps the readiness score under a ceiling, however many points the other blocks earn. An online order button on a prescription drug counts as a violation even if everything works technically.
I do not merge readiness and visibility into one number. Readiness is what you control. Visibility is the output of other companies' models, and the share of mentions moves from run to run without a single change on your side. Merged into one number, a jump in visibility after a model update would look like the result of your work.
The platforms do not treat optimisation for AI search as a separate discipline. Google calls it the same SEO ("still SEO"). Yandex writes that Alice AI's answers rely on search results, so no fundamentally new requirements arise. Neither Google nor Yandex requires special markup for AI. I do not sell "GEO optimisation" as a separate service.
The term GEO was introduced by Aggarwal and co-authors in a paper for KDD 2024. In their dataset 80% of queries were informational, and visibility was measured as a share of the answer text rather than by whether the assistant named a pharmacy. I do not carry their figures over to "where to buy" queries.
Russia is not on the list of countries where ChatGPT is officially available, so I start with Alice AI and GigaChat. A direct lever is confirmed only at Yandex: it states that feeds and structured markup are needed for the assistant's answers. The second layer of sources is pharmacy aggregators. How often the assistant sends the shopper there rather than to the pharmacy chain's own site will show up in the first measurement.
In pharmacy, a long path to the "Buy" button is not always a bad thing. For a prescription drug, the right path is a product page with price and availability and a reservation without online checkout.
I test the effect of the work against a control group: some products go into the work, the rest are left as they are, and we compare the gains. Without a control group, a fix cannot be told apart from the next model update.
A place in the answer. Yandex states plainly that priority placement in the agent's selection is not offered. A position in Alice's answers cannot be bought, and I will not sell one.
An explanation of why the assistant chose a seller. Models and ranking are closed. I record what the assistant answered and label the reasons as hypotheses.
Agentic purchase of medicines. Since 1 March 2026 the law has not permitted remote sale of prescription drugs: Article 55.1 of Federal Law No. 61-FZ is no longer in force. Over-the-counter drugs may be sold remotely only with a permit from Roszdravnadzor, the federal healthcare regulator. Google has excluded medicines from checkout under its UCP protocol; OpenAI prohibits prescription drugs and does not accept seller applications from Russia.
Revenue growth in percent. Revenue from assistants is invisible: a person reads the answer and walks into a pharmacy, and analytics will not show it. I count click-throughs as a lower bound and measure the effect of the work on a control group.
Medical judgements. I do not assess treatment, suggest analogues or substitutes, or ask assistants about symptoms and doses. I compare the price and availability of the same product.
After the audit you keep:
A readiness score across five blocks, with the compliance-gate status at the top of the report.
Visibility and data accuracy for each assistant and each city, with margins of error and next to your competitors.
A list of findings, each with evidence (the assistant's answer, a page snapshot, a feed line), its reach in products, pharmacies and queries, its basis (law, platform requirement, standard or hypothesis), the responsible department and a priority.
A PDF report and an evidence spreadsheet that lets you verify every finding yourself.
Every result records the version of the methodology and of the query set, so a measurement a quarter later is comparable with the first one. Competitors are named in the report, but only you see them.
Mini Check. from RUB 50,000 Runs on public data; nothing is needed from you. Output: a preliminary readiness score without feeds, a small query set for one city and a list of obvious errors. The report states plainly that this is not a full audit.
Full Readiness Audit. from RUB 300,000 Requires a feed, a price and stock export, a list of pharmacies and access to your Yandex accounts. Output: a score across all five blocks, the compliance-gate status, and visibility and accuracy on the full query set in two cities.
Competitive Benchmark. from RUB 110,000 The same measurements for three competitors of your choice. You see readiness on the blocks measured for everyone, and the visibility gap in percentage points with its margin of error.
Implementation Roadmap. from RUB 60,000 I sort the findings into a priority matrix and add a work plan by department, a control group of products and baseline values for the metrics used later to measure the effect.
AI Visibility Monitoring. from RUB 90,000 a month A monthly report on visibility, price and availability accuracy, and new violations.
Continuous Optimization. from RUB 140,000 a month I measure the effect of the work as the difference in gains between products in the work and control products, recalculate readiness every quarter and update the list of findings.
Prices are for one pharmacy chain, two cities and two assistants. Each additional city, assistant or competitor adds work and cost. I name the exact sum after the first conversation and fix it in the contract together with the scope and timeline.
To discuss an agentic commerce readiness audit for your pharmacy chain, write to me on Telegram.
Message on Telegram