ChatGPT Shifts From Asking to Doing. Intent Just Moved.
OpenAI has, for the first time, published country-by-country data on how more than a billion people use ChatGPT, and the headline is stark: people are more than twice as likely to use it to do something at work than to simply ask a question. That single line, from the OpenAI report on how the world is putting ChatGPT to work, is the clearest primary-source evidence yet that a large slice of commercial intent has moved out of the search box and into the assistant.
If you run growth, search or content, treat this as a discovery-channel map, not a curiosity. The question is no longer whether AI absorbs your funnel. It is which stages it has already taken.
What actually happened
On 6 August 2026, OpenAI's Economic Research Team released the OpenAI Signals dataset, a public hub breaking down ChatGPT usage across 144 countries by behaviour and use case. The data covers Free, Go, Plus and Pro accounts, meaning it reflects individual users rather than managed enterprise seats.
The core finding is a behavioural split. Outside work, "asking" (seeking information and clarification) is still the largest category. At work, "doing" dominates: editing, coding, analysis, producing an output. Globally, people are more than twice as likely to be doing than asking once they are on the job.
When, and how fast it is spreading
The shift is not a rich-market phenomenon. Weekly active users have grown up to 25x in lower-adoption regions since mid-2023, and in Q2 2026 Peru, Uruguay and Costa Rica led the per-capita gains. Usage among people over 35 rose in nearly every country, up more than 10 percentage points in France and Czechia in a year.
| Signal | Figure | Why it matters |
|---|---|---|
| "Doing" vs "asking" at work | 2x+ more likely to do | Task completion, not lookup |
| Multimedia share of messages | 7.8% globally | Fastest-growing use case |
| Multimedia in Brazil and Colombia | 1 in 10 messages | Visual intent is regional |
| Over-35 message share (France, Czechia) | +10 points in a year | Mainstream, not early adopters |
How this changes your funnel
Here is the mechanism. When a user "asks", they may still click out to your site, so classic SEO applies. When a user "does", the assistant completes the task inside the chat: it drafts, compares, decides, sometimes buys. Your content is consumed as an input to that action, not visited as a destination.
In my opinion, that reframes the whole measurement problem. If commercial intent resolves inside ChatGPT, your analytics will show fewer sessions and more assisted, invisible conversions. That is exactly why making AI referral traffic countable in GA4 is now table stakes, not a nice-to-have.
Three moves to make this week
- Audit which funnel stages are being absorbed. Map your top tasks to asking or doing. Comparison, configuration and drafting are already doing.
- Build content an assistant can act on. Clear specs, pricing, steps and structured data, not just prose an LLM can cite.
- Add multimedia inputs. With multimedia at 7.8% and rising, product imagery and diagrams are becoming discovery assets.
From my observation, this is the same pivot that reshaped small-business operations when OpenAI made AI the default operating layer. The action is simple: stop optimising only to be found, and start structuring to be used.
Tags