Fewer than seven in every 100 ChatGPT conversations have anything to do with buying. That is the headline finding of a study Rand Fishkin of the marketing research firm SparkToro published on 17 November 2026. In 93.1% of conversations there was no purchase intent at all. In 4.9% the user was comparing options, and in 2.1% they asked about a specific brand, product or service.
If you pay to get your business mentioned in AI answers, this is a number to know before deciding how much to spend and on which channel.
How the sample was built
The data came from an opt-in panel of American users who agreed to share their chats, supplied by a company called Verso. It covers 118,163 prompts across 41,961 conversations from 12,321 users, between 1 January and 29 October 2026. The researchers removed 131 very heavy users who had written 21% of all prompts on their own. Most of the figures come from a subset of 9,675 users and 32,980 conversations, weighted to match the US census.
A language model did the sorting. Anthropic's Claude labelled each conversation against a list of 34 intents grouped into 12 categories, and a second labeller agreed on the category 90% of the time.
Images top the list, shopping sits near the bottom
The most common use is not text at all. Creating and editing images comes first, followed by information and entertainment, then writing. The weighted figures:
| Category | Share of conversations |
|---|---|
| Images and visual design | 17.6% |
| Information and entertainment | 14.4% |
| Writing and communication | 12.7% |
| Everyday life and admin | 10.3% |
| Health and wellness | 8.5% |
| Schoolwork and learning | 8.0% |
| Money, work and business | 7.0% |
| Shopping, brands and local services | 5.3% |
| Tech support and coding | 4.7% |
The article text gives slightly different numbers from its chart, up to two percentage points apart per category. The ranking holds either way.
One more figure matters for businesses: 12.1% of ChatGPT's responses named at least one brand, product or company. Business names come up in answers even when the user is not shopping.
Google is still where people go to buy
For comparison, an earlier SparkToro analysis of 332 million Google searches found that more than 15% were part of researching a purchase. That is more than twice the ChatGPT rate, though the two studies sort queries differently and do not compare one to one.
Fishkin's reading is that ChatGPT behaves like a consumer app, not a search engine. In his view, the reason to care about visibility in AI answers is mostly Google's AI Overviews and AI Mode, and much less ChatGPT itself.
The study's limits
The panel is US only, and it skews younger, more female and lower-income than average. Most panelists used their phones, so desktop and workplace use is under-represented. Use through developer APIs, which Fishkin guesses is a large share of total usage, is missing entirely. And the study measures what people ask about, not how much they use the tool.
There is no Israeli data in it, and no reason to assume the numbers in Israel or anywhere else match.
Where a small business should put its effort
For a small business looking for customers, the practical takeaway is an order of priorities. Your Google Business Profile, your website and your ranking in Google search still reach people in the middle of a buying decision. ChatGPT reaches a lot of people, but most of them are busy with something else.
That is no reason to ignore it. The 2.1% who ask about a specific service are customers who have nearly decided, and when they ask for an accountant or a designer in their city, you want your name in the answer. The guide to getting ChatGPT to recommend your business walks through the steps, and since last week businesses in Israel can also claim a profile in ChatGPT and see how often they were mentioned.
The study follows another figure Fishkin published a week earlier, that the web's top million sites lost 12% of their visits since 2022. Read together, the two make the case for counting customer enquiries rather than site visits, and checking where each enquiry came from before moving budget between channels.