• USD/RUB USD/RUB 82.92
  • EUR/RUB EUR/RUB 96.86
  • CNY/RUB CNY/RUB 12.33
  • Bitcoin Bitcoin BTC $78065
  • Ethereum Ethereum ETH $2479
  • Ripple Ripple XRP $1.39
  • Solana Solana SOL $92.75
  • Cardano Cardano ADA $0.22
  • USD/RUB USD/RUB 82.92
  • EUR/RUB EUR/RUB 96.86
  • CNY/RUB CNY/RUB 12.33
  • Bitcoin Bitcoin BTC $78065
  • Ethereum Ethereum ETH $2479
  • Ripple Ripple XRP $1.39
  • Solana Solana SOL $92.75
  • Cardano Cardano ADA $0.22

Traffic from AI has increased 4-fold at Ozon in a year

Traffic from AI has increased 4-fold at Ozon in a year
Most Popular
19.08
Wildberries accepts FBW supplies through PVZ for 5 rubles
19.08
Wildberries: from August 31, the FBS commission depends on the assembly hours
19.08
Since August 17, the list of goods for temporary import into the EAEU has been expanded
19.08
Vapes and alcohol have been a strategically important commodity since August 20.
18.08
The Federal Tax Service named errors in the DPP, which cause money to hang up
18.08
Marking of building materials was shifted to June 1, 2027
Traffic from chatbots and neural networks to the websites of Russian online stores increased 1.7 times over the year. For Ozon— it is four times. The servers are not coping. A side effect for the market: small brands have a real chance to bypass large marketplaces in issuing neural networks without advertising budgets.

A year ago, traffic from AI search was perceived as an experimental channel that could be ignored. Now it is 1.7 times per year for the market as a whole and four times for Ozon. This is already a channel that cannot be ignored.

What's going on

Neural network search engines and chatbots - ChatGPT, Yandex Neuro, GigaChat, and others — have begun to include direct links to product pages and online stores when responding to user requests. The user asks "which running shoes are better to choose" — specific models with links appear in the answer. Some of this traffic goes directly to the store's website, bypassing the marketplace.

For Ozon, a four-fold increase means that neural networks are actively indexing product cards from the platform and giving them out in responses. But the same thing happens with independent stores, and an interesting situation arises here.

A small business opportunity

Neural networks do not sell advertising positions in the same way as Google. The link selection algorithm is based on the quality of the content, relevance, and structure of the information on the page. This means that a small specialty store with high-quality descriptions and good reviews can get into the neural network response on a par with a major player.

For importers who sell through their own website in parallel with marketplaces, this is an argument to invest in the quality of the content: detailed descriptions, honest technical specifications, and answers to typical customer questions.

The infrastructure problem

The increase in traffic has created an unexpected problem: the servers are not coping. Some sites have recorded slowdowns and unavailability during peak periods from AI bots that scan pages for training and indexing. For online stores with their own infrastructure, it is a signal to check the current load and reserve capacity.

What sellers should do

For those who trade only through marketplaces: make sure that the product cards are filled out in as much detail as possible — neural networks pull out well-structured content. For those who have their own website: check whether the site is indexed by neural network search engines, add schema.org markup for products — this increases the chances of getting into the answers.

The channel is still free. But the competition for a place in the responses of neural networks will only grow.