Chatonbo Chat Index, September 2026: after-hours chats rise to 48.8%, shipping questions double ahead of Q4

TL;DR: In September 2026, 48.8% of chats on live store and business websites started outside business hours, 56.1% happened on a phone, 12.0% ended with the visitor leaving contact details, and shipping questions doubled their share as stores entered the run-up to Q4. Same definitions as every month; the series lives at [chatonbo.com/research/chat-index](/research/chat-index).

What the Chat Index is

The Chat Index is a monthly set of shares computed from the AI assistants running on real Chatonbo customer websites: when visitors chat, on what device, how often a chat becomes a lead, and what they ask. It uses the same definitions as our [5,470-conversation study](/research/what-shoppers-ask-ai-chat-2026), published every month for completed months only, and is free to quote with attribution (CC BY 4.0). This is the September 2026 edition.

September 2026 at a glance

After-hours share climbed to 48.8%

Nearly half of September's conversations began when nobody was at the desk, up seven points on August. Two things moved it: the mix of live assistants shifted toward stores in evening-heavy categories (home, hobby and gifting sites chat late), and daylight hours shrank in the northern hemisphere, which pushes browsing later. Either way, the practical reading is the same as the baseline, only stronger: an assistant that cannot answer on its own misses about half the questions.

What visitors asked

Product questions stayed the largest topic, as they are every month. The mover was shipping and delivery at 16.3%, roughly double its share in the June to September study. That is the Q4 pattern starting early: "how long does delivery take", "do you ship to Canada", "will it arrive before the weekend". A store whose assistant cannot answer shipping times by destination is about to lose a sixth of its chats to a dead end.

Phones and leads held steady

56.1% of chats came from a phone, flat against August and well above the summer baseline of 47.7%. The lead rate, 12.0% of chats ending with a name, email or phone number, was unchanged. Only 0.4% of conversations asked for a human.

What to do with it before November

1. Put delivery times by destination, cut-off times and the returns window on a page the assistant can read. Shipping is the fastest-growing question and the easiest to answer in advance.
2. Treat the evening as the main shift. Half of your questions arrive then.
3. Check the assistant on a phone screen: short answers, product cards, one tap to the cart.
4. If you run a Black Friday offer, give the assistant the code and the rules. "Do you have a discount code?" is the most typed sentence on store sites.

Method, in short

Every conversation opened on the live website of a Chatonbo customer in the month, storefront sessions only. Internal workspaces, test chats and app-store review installs excluded. After hours: conversation start shifted by one fixed offset per visitor country falls outside 09:00 to 17:59. Phone: browser user agent. Lead: the visitor left a name, email or phone number. Topics: the assistant's own classification. Completed months only, when the month's sample is large enough to be meaningful. Full method and the baseline study: [What shoppers actually ask an AI chat assistant](/research/what-shoppers-ask-ai-chat-2026). Cite as "Chatonbo Chat Index, September 2026, chatonbo.com/research/chat-index".