What shoppers really ask before they buy from your brand?

Among conversations the shopper typed, 35.4% start with no product in mind. They want the store to choose with them.

Most of them still give a constraint, like a budget, a skin type or a car model. That covers 24.8% of typed pre-purchase conversations, so the store has something to work with.

Across all conversations, 81% happen before an order exists. Brands that staff chat as after-sales support answer the wrong conversation.

Help me find a product is the top topic at 33.4% of pre-purchase questions. Price and discounts sit at 11.0%, a third of that.

Returns and warranty and Trust and authenticity, the staples of every FAQ page, land at 1.4% and 1.0%. Will it work for me follows advice at 15.2%.

Shipping and delivery follows at 6.3%. In smaller stores, shipping and price questions run about two points higher.

We converted every timestamp to the shopper's local time. 59% of pre-purchase conversations land outside office hours, meaning before 9 am, after 6 pm or on weekends: 30.4% after 6 pm and 27.2% on weekends.

These are buyers, not browsers. 57.6% of attributed orders came from conversations outside office hours, so a team that only works office hours misses the larger half of the sale.

Peaks differ by market. French speakers peak at 10 am and Italians at 3 pm. German speakers peak at 8 pm and Slovak shoppers at 9 pm.

Materials questions fall from 6.0% to 1.2% as price rises. Specs questions double, from 9.4% to 18.4%.

Above $500, conversations reach 3.2 messages, about 40% more than the $25 to $50 band. Handoffs to a person climb from 2.1% to 5.4%.

Patience grows with price. Median time to order goes from 8.4 minutes at $25 to $50 to 45.0 minutes above $1,000.

68.4% of typed AI search queries carry no product. The median query has 2 words, so shoppers type it like a search bar and expect a shortlist.

Product page questions do the opposite. Only 2.8% are discovery, because the shopper is already looking at a product.

Channels also change what a brand gets back. We divided each brand's attributed orders by what it paid for its AI assistant. The average brand collects $47.9 for every dollar, and 97.8% of brands collect more than they spend.

The return grows with each channel. Brands on one channel collect $29.1 per dollar, on two channels $51.0, and on three or more $83.7. The gap holds at every budget level.

The return also grows over time. For the same 70 brands, the average return rose from $56.2 per dollar between November and January to $67.8 between August and October. The earlier period includes Black Friday and Christmas.

Between January and October 2026, the attributed orders of the same 82 brands almost tripled. Over the same months the assistant learned to act, not only answer: more on that in chapter 10.

We compared each topic against the same store's conversion rate. Five product questions sit just under the average, at 0.90 to 0.93 times: Will it work for me, Dimensions and specs, How to use or install, Help me find a product and Materials and ingredients.

These are the questions a complete product sheet answers. When the assistant knows the specs, the use cases and who each product suits, it answers on the spot and moves the shopper to the next step.

That next step converts. Comparison questions convert at 1.31 times the store average, price questions at 1.53 times and payment questions at 1.64 times.

Trust questions convert at 0.79 times and contact requests at 0.68 times. They flag doubt, so answer them with proof before any offer.

One-message conversations convert at 6.6% with a median order of $69.3. Five or more shopper messages convert at 12.6%, with a median order of $107.5.

Conversations handed to a person convert at 11.4% against 8.2% for the rest. Their median order is $139.4, close to double.

Speed still matters. The median order lands 13.8 minutes after the first question, and 52.5% within 15 minutes.

Custom skills let the assistant act inside a brand's own tools. It checks an order in the shop system, reads a compatibility table or opens a helpdesk ticket. We tagged every skill action from May to October 2026.

Order and shipment lookups make up 42.8% of skill actions, product data lookups 20.9% and helpdesk tickets 14.4%. Two in three actions happen after the purchase, which takes routine work off support teams.

Before the purchase, skills sell. In the same store, pre-purchase conversations with a skill action convert at 7.97% against a 5.73% average. 71.2% of brands see the gap.

Product data lookups run 94.9% of the time before the purchase. They answer what a static FAQ cannot, like whether a part fits a given model or a size is still in stock.