Wildberries has launched AI auto-responses to reviews: sellers set up once

Wildberries has launched AI auto-responses to reviews: sellers set up once
Most Popular
21.02
GosLog will be launched on March 1: a single window for carriers
20.02
Uzbekistan–Turkey launch container trains via Iran and Turkmenistan
20.02
India tripled purchases on the Belarusian stock exchange: almost $5 million in 2025
20.02
Putin created the BRICS National Business Committee: It will be headed by Oreshkin.
20.02
The Constitutional Court ordered online sellers to give remote returns of “high-quality” goods
19.02
Refunds on marketplaces will be prescribed in the law: disputed categories will be divided
Wildberries has launched an AI tool for sellers that can automatically generate and publish responses to reviews according to set rules (for example, only 4-5 stars). The model uses the product card and the seller's past responses, and the published texts can be edited. This reduces the manual load and increases the speed of the service.

Wildberries continues to "compact" the service layer for sellers and puts work with reviews into semi-automatic call center mode: RVB has launched an AI response tool that not only offers text options, but can generate and immediately publish responses to customer reviews according to specified rules.

The key change is in the distribution of responsibility over time. Previously, the seller manually selected the generated version he liked and published it himself. Now the seller's participation is reduced to the settings "at the start", and then the work goes on automatically. This is an important shift for the marketplace economy: the speed of reaction to reviews is becoming not an advantage of large teams, but a basic function of the tool.

How it works:

  • The seller chooses which ratings the system will respond to automatically (for example, only 4-5 stars), and the rest of the reviews remain in manual processing.;
  • The model generates a response using the product profile and the text of the review, and also takes into account the style of the seller's previous responses.;
  • responses can be edited after publication.;
  • if there is insufficient data (for example, the product card is "empty"), the review may remain unanswered in order not to produce incorrect formulations.

For sellers, this is a tool not only about saving time, but also about risk management. Auto-replies to high marks are a safe scenario: they help to maintain a "warm" communication without overloading the team. But in the negative (1-3 stars), automation is a zone of increased responsibility: any unsuccessful phrase can worsen conflict and conversion. Therefore, the logic of “leaving low marks to a person” looks pragmatic.: They need fact-checking, personal solutions, and sometimes compensation mechanics.

A separate layer is the AI recommendations in the responses: the system can substitute the articles of other goods of the seller, selecting them as "related" to the purchased one. For the marketplace, this turns reviews into a new advertising inventory within trusted content, and for the seller, into a soft cross—sell without direct traffic costs. But a measure is important here: excessive recommendations can be perceived as spam and reduce the credibility of the responses.

From a logistical point of view, there is also an effect, albeit an indirect one. Fast and high-quality responses reduce the proportion of "refunds due to misunderstandings" (size, configuration, usage features) and reduce the burden on return logistics. And if the AI is forced to “remain silent” because of a poor product card, this is a signal to the seller: a weak description is a direct operational risk (disputes, refunds, rating drawdowns). As a result, the new feature pushes sellers to discipline data: the more accurate the card, the better the automation of the service works.