Key findings

  • Commercial intent is rising both as a share of ChatGPT conversations (13.9% → 19.2%) and in absolute volume per user.
  • Commercial conversations are the deepest: the highest average turn count and the lowest single-turn share (53%).
  • Commercial share varies roughly 10x by industry, from 59.1% in Automotive to 6.2% in Education & Learning.

The share is climbing, and so is the volume behind it

Between June 2025 and June 2026, the share of ChatGPT conversations with commercial intent rose from 13.9% to 19.2%, with seasonal spikes around Black Friday (late November) and Christmas (late December). The growth came from people having more commercial conversations each (0.38 → 0.54 per active user per week, +41%), not from other intents shrinking. Informational intent held steady in the mid-teens while generative conversations declined.

Share of ChatGPT conversations by intent, weekly
June 2025 – June 2026 (en-US)
25%20%15%10% 13.9%19.2% Black FridayChristmas Jun ’25Jun ’26
CommercialInformationalGenerative

An estimated 28 billion commercial conversations a year

To translate the sample into scale, we combined observed per-user rates with OpenAI's reported weekly active users, which grew from 500M (March 2025) to roughly 1.0B (July 2026). Weekly commercial conversations rose from an estimated 243 million (week of June 23, 2025) to 533 million (week of June 22, 2026), a 119% increase. At current run-rates that is roughly 28 billion commercial conversations a year.

WeekCommercial shareCommercial convs / user / wkEst. WAUCommercial conversations / wk
Jun 23, 202513.9%0.38~640M243M
Jun 22, 202619.2%0.54~990M533M
+119%Weekly commercial conversations, YoY
+41%Commercial conversations per active user
~28BCommercial conversations per year

Commercial conversations are the stickiest intent

Conversations got shorter across every intent as models improved and accumulated context resolved questions faster, but commercial conversations kept the highest average turn count and the lowest share of single-turn exchanges.

IntentAvg. turns (Jun 2025)Avg. turns (Jun 2026)Single-turn share
Commercial3.202.4353%
Generative3.442.25—
Informational2.762.18—

Which industries drive commercial intent

Classifying 1.75 million conversations across 19 industries reveals a roughly 10x spread. High-consideration purchases dominate: Automotive (59.1%), Travel & Hospitality (56.1%), Telecom (47.9%), Hardware (47.6%) and Financial Services (47.1%) lead. Healthcare and Legal Services skew informational; Marketing, Non-profit and Consumer Goods skew generative. Education & Learning sits at the bottom with 6.2% commercial.

Commercial share of conversations by industry
  • Automotive59.1%
  • Travel & Hospitality56.1%
  • Telecom47.9%
  • Hardware47.6%
  • Financial Services47.1%
  • Healthcare~18%
  • Legal Services~15%
  • Education & Learning6.2%
Selected industries of 19. Healthcare and Legal shown approximately; both rank among the most informational.

What this means for brands

Nearly one in five ChatGPT conversations is now commercial, up from roughly one in seven a year ago, and the volume behind that share has more than doubled to approach 28 billion a year. The pattern differs sharply by industry, which means the purchase journey inside ChatGPT looks different for a carmaker than for a university. Brands that are visible in these conversations capture purchase decisions in a channel that barely existed two years ago.

Methodology

Sample: 7.5 million en-US ChatGPT conversations from June 23, 2025 to June 22, 2026, aggregated into 53 weekly cohorts and classified into three mutually exclusive intent buckets (commercial, informational, generative). Scale extrapolation applies observed per-user commercial rates to OpenAI's reported WAU milestones with linear interpolation; mid-2026 figures are estimates, and a ±10% sensitivity on the user base yields 25–30 billion annual commercial conversations. Industry analysis used keyword pre-filtering (3 million candidates), LLM classification against 19 categories, confidence calibration against hand-labelled data and validation against an independent random sample.

Davis McCainResearch, Flowgen