Only 6.9% of ChatGPT Chats Show Buying Intent

The GEO Show
November 19, 2026
Watch this episode on YouTubeListen on Apple Podcasts, Spotify and more

ChatGPT purchase intent is rare: SparkToro and Verso found that only 6.9% of 41,961 real US ChatGPT conversations showed any intent to buy, and 87.9% of responses named no brand at all. Paris Childress explains why that points brands back to Google AI Overviews and AI Mode, then covers AI Mode temporary threads, LAUSD turning off AI Overviews, Perplexity's 1 billion dollar run rate, and GPT-6.2.

Key takeaways

  • SparkToro and Verso found that only 6.9% of 41,961 real US ChatGPT conversations showed purchase intent: 4.9% researching options and 2.1% asking about a specific brand.
  • 87.9% of ChatGPT's responses in the panel named no brand, product, or business, which undercuts pitches that call ChatGPT the new shopping funnel.
  • Rand Fishkin compares this with 15% plus of Google queries being purchase related and argues Google AI Overviews and AI Mode matter more for AI visibility than ChatGPT.
  • Temporary threads in Google AI Mode give one-tap answers detached from personalization, which could make AI Mode results less consistent for anyone measuring AI visibility or share of voice.
  • Los Angeles Unified turned off AI Overviews for students and says it could switch its default search engine overnight, the first big public sector buyer to openly consider leaving Google over AI answers.
  • Model upgrades such as GPT-6.2 Astra and Sol can change which sources ChatGPT cites, so annotate every release in share of voice reports even when the launch post says nothing about citations.
  • 🛒 Only 6.9% of ChatGPT chats show purchase intent: SparkToro and Verso analyzed 41,961 conversations from 12,321 US panel users (January 1 to October 29, 2026). 4.9% were researching options and 2.1% asked purchase questions about a specific brand, and 87.9% of responses named no brand, product, or business. Rand Fishkin compares this with 15% plus of Google queries being purchase related and argues AI Overviews and AI Mode matter more for AI visibility. Caveats: opt-in, phone-heavy panel that skews younger, female, and lower income, no API use, and the 131 heaviest users removed.
  • 🕶️ AI Mode gets temporary threads: private conversations detached from your account, not saved to history, and not used for personalization, rolling out globally on desktop and mobile browsers in every AI Mode language. Paris's take: one-tap non-personalized answers could make AI Mode results less consistent, so factor it into AI visibility and share of voice measurement.
  • 🏫 LAUSD turns off AI Overviews for students: Los Angeles Unified disabled AI Overviews and is reviewing whether to keep Google Search, with its CIO saying the default could switch overnight. New York City and Chicago are weighing other providers, with DuckDuckGo and Bing under consideration, while Google says it won't offer Search without AI features.
  • 📈 Perplexity crosses a 1 billion dollar run rate: CEO Aravind Srinivas says Perplexity has crossed a 1 billion dollar annual revenue run rate, up from under 250 million dollars in January. Still a small source of AI referrals, but a sizable commercial platform brands should track.
  • 🤖 OpenAI ships GPT-6.2 Astra and Sol: available to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex and in the API. Astra costs 8 dollars input and 40 dollars output per million tokens, 20% below GPT-6 Astra, and Sol stays at 2 and 10 dollars. The launch post says nothing about search, browsing, or citations, so annotate the change in share of voice reports.
  • ⚡ Lightning round: AI Overview citation cards shown in full instead of faded, Anthropic's IPO reported to price above its 70 to 78 dollar range, and Joy Hawkins flagging fake Search Console clicks with high CTRs on low-ranking keywords.

Full transcript

Paris: Hi, everybody. Welcome back to The GEO Show, brought to you by GEOforge, full self-driving for AI visibility. I'm your host, Paris Childress, recording on Thursday, November 19th. Today's big headline is everyone's racing to get their brand recommended in ChatGPT. But new data from forty-two thousand real conversations says that fewer than seven in one hundred are buying anything. So that leads us to our first story. SparkToro and Verso did a survey that discovered that six point nine percent of ChatGPT conversations show purchase intent.

Paris: SparkToro looked at nearly forty-two thousand ChatGPT conversations from about twelve thousand US users who shared their chat histories through a data panel. Only six point nine percent showed any purchase intent. Almost eighty-eight percent of ChatGPT's responses named no brand, product, or business at all.

Paris: So here's more about the data. Verso runs an opt-in panel of US ChatGPT users who linked their accounts. The sample is forty-one thousand nine hundred and sixty-one conversations and one hundred and eighteen thousand prompts from twelve thousand three hundred and twenty-one people, covering the 1st of January through October twenty-ninth.

Paris: Claude coded each conversation against thirty-four intents. A second labeler agreed on the categories ninety percent of the time. What they found was no purchase intent ninety-three percent of the time. Researching options, just four point nine percent. Purchase questions about a specific brand, only two point one percent.

Paris: Brands and categories. 12% of responses named at least one brand. Images and visual design was the largest category at 16.1%. Shopping brands and local services was 5.7%. So Rand Fishkin compares this with 15-plus percent of Google queries being purchase-related. He argues that Google AI Overviews and AI Mode matter more for AI visibility than ChatGPT does. All right, a few caveats here.

Paris: The panel is opt-in, and it skews towards younger female users and lower-income users. It is mostly phone usage. It excludes API use, and it drops the one hundred and thirty-one heaviest users who wrote twenty-one percent of the prompts. The source says prompts in one place and conversations in another for the six point nine figure.

Paris: That's probably no big deal. There's no separate agreement score for the purchase intent labels. Verso licenses de-identified chat data commercially, and the study has not been replicated. OpenAI has not commented. All that being said, this is the first large look at what consumers actually ask ChatGPT based on real chats rather than surveys.

Paris: It undercuts pitches that call ChatGPT the new shopping funnel. It does not measure how much a rare but high-value buying conversation is worth. I think in large part, this goes to show just how many varied uses there are for ChatGPT that go well beyond searching with commercial intent. So I wouldn't necessarily take this as a fact that people are not using ChatGPT for commercial intent search-related activity the way that they use Google.

Paris: I also think that Google's commercial intent queries also are probably a small percentage. Next story, Google rolls out temporary threads in AI Mode globally. Google has added temporary threads to AI Mode. These are private conversations detached from your account. They aren't saved to your history, and they don't shape future personalization. The feature is rolling out globally on desktop and mobile in every AI Mode language.

Paris: Here's how it works. You tap an icon in the top right of AI Mode. Answers aren't informed by your past preferences, don't affect future personalization, and aren't saved to history. Google says the thread is gone once you close it, so it sounds kind of like an incognito mode for AI Mode. Availability is on desktop and mobile browsers in all languages that AI Mode supports.

Paris: Google doesn't mention any account tier limits. And TechCrunch compares it to temporary chats in the Gemini app. So AI Mode is becoming more personalized through personal intelligence and connected apps. A one-tap way to get a non-personalized answer could make what different people see less consistent. It also changes how practitioners can check what AI Mode says about their brand. So if you are measuring AI visibility and share of voice in AI Mode, keep in mind that this new feature of temporary threads is going to be rolling out and probably affecting those answers.

Paris: Next story. LAUSD, which is Los Angeles Unified School District, turns off Google's AI Overviews for students. New York City and Chicago weigh other search engines. The LAUSD says that it has turned off Google's AI Overviews for students and is still reviewing whether to keep Google Search at all.

Paris: According to The New York Times, New York City and Chicago schools are also weighing other search providers. On November eighteenth, an LAUSD spokesperson said AI Overviews are disabled for students while the district keeps evaluating Google Search. LAUSD's CIO says the district could switch its default search engine overnight.

Paris: DuckDuckGo and Bing are under consideration. New York City and Chicago say they will look at other options if they can't reach an agreement with Google. Google's position here is that Google won't offer Search without AI features. They will not decouple Search from AI, saying that the AI is too deeply integrated.

Paris: It says it is giving schools more choices, not fewer. Google did not respond to NBC LA's request for comment, and this does follow Google's K-12 policy, which we covered back on November fourth Well, this standoff between public schools and Google seems to be escalating here. This is the first big public sector buyer to openly consider leaving Google over AI answers. Next up, Perplexity CEO says the company crossed a one billion dollar annual revenue run rate. Perplexity CEO Aravind Srinivas says the company has crossed a one billion dollar annual revenue run rate.

Paris: Reports earlier this year put it under $250 million in January. So Perplexity, still alive and kicking, still growing. Perplexity is still a small source of AI referrals and AI usage overall. I think that their market share is under 1%, but with these figures, it is still a sizable commercial platform that brands should track. All right, moving along. OpenAI ships GPT-6.2 Astra and Sol.

Paris: OpenAI released GPT-6.2 Astra, its frontier model, and GPT-6.2 Sol, its efficient everyday model to paid ChatGPT Work and Codex users and to the API. The launch post doesn't mention search, browsing, or citations. The availability is for everyone across Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, plus the API as GPT-6.2 Astra and GPT-6.2 Sol.

Paris: This was reported on November 17th. API pricing, Astra is $8 and $40 per million input and output tokens, which is 20% below GPT-6 Astra, and Sol is unchanged at $2 and $10 per million input and output tokens, respectively. So just as with every new model release, we do have to annotate these changes and these model upgrades in our share of voice reports.

Paris: These model upgrades can change which sources ChatGPT cites, but OpenAI announced nothing about that, of course. All right, let's wrap up with three lightning round stories. First, AI Overview citation cards are shown in full. Search Engine Roundtable says Google appears to have rolled out AI Overview citation cards that are fully visible instead of faded. Search Engine Watch separately reports a sticky citations test on desktop.

Paris: Next, Anthropic's IPO. Several outlets report that Anthropic will price above its seventy to seventy-eight dollar range. PitchBook calculates that seventy-eight dollars a share implies a valuation of about one point four trillion dollars. And last, Search Console CTR oddities. Joy Hawkins flags insane CTRs on keywords far from the top positions, but Google hasn't commented.

Paris: Okay, that'll do it for today's episode. Thank you all for listening, and we'll see you on the next one.

See how AI answers describe your brand

GEOforge measures how often ChatGPT, Google AI Overviews and AI Mode cite you, then builds the content and citations that change it.