ChatGPT Took 81.4% of Traffic Across Six AI Chatbots in Graphite Study

Your buyer is already used to asking ChatGPT questions, and you are offering them a new AI assistant. That means they will most likely compare your product with a familiar service: how quickly the answer arrives, whether a file can be attached, whether context has to be repeated. The Graphite study puts a number on this habit: according to data cited by Business Insider, ChatGPT received 81.4% of global traffic across six major AI chatbots.

For a founder, this is a reason to check the product's positioning. For a product manager, it is a reason to understand what expectations users bring with them. But for a developer, choosing a model provider based on this chart would be premature: the traffic of a consumer service answers a different question.

What Exactly Was Measured

In the retelling of the study for July–September 2026, the shares were distributed as follows: ChatGPT — 81.4%, Claude — 6.8%, Gemini — 5%, DeepSeek — 4%, Grok — 1.9%, Perplexity — 0.9%. This refers to visits to six selected services. The wording "81.4% of the entire AI market" would stretch the conclusion far beyond this sample. The data and period of the study are described in the Business Insider article.

The original digest also cites 61.35 billion visits to ChatGPT in September and growth of almost 20%. These absolute figures should be used with caution: the available retelling does not allow verification of the counting methodology, the coverage of apps, or the exact definition of a visit. A visit cannot automatically be treated as a separate person, a paying customer, or a completed work task.

Even a correctly measured traffic share will not tell you how much companies earn from their APIs, whose models more often run inside corporate applications, or who has higher retention of paying teams. Separate data is needed for those conclusions. A narrower signal is useful here: within the group in question, user attention is heavily concentrated around a single service.

How to Apply This to Your Product

Suppose you are building an assistant for handling customer requests. At a presentation, a prospective customer asks: "Why can't employees just open ChatGPT?" The large audience of a general-purpose chatbot makes such a question entirely expected. The answer should explain concrete value: the assistant itself retrieves the order history, respects access rights, drafts a response according to company rules, and saves the result in the working system.

This is an editorial conclusion drawn from the audience structure, not a result of Graphite's experiment. But it suggests a clear action: check whether you can demonstrate the product's value on a single real request. If the demo boils down to the same input field and a similar answer, it will be hard for the user to understand why to pay separately.

A product manager should also check the first launch. Familiar scenarios — a question, an attachment, a clarification — can make onboarding easier. This should be verified by observing users: do they complete the first task, do they understand the result, do they come back. A market chart on its own does not prove that a particular interface will improve your conversion.

A developer needs their own set of typical requests. Compare models on identical examples: how many answers pass validation, how much time they take, and how much a successfully completed task costs. ChatGPT's popularity may help explain the product to a customer, but it will not replace such a comparison.

Where the Usefulness of the Number Ends

Be especially careful about transferring the results to a specific country, industry or corporate segment. The global structure of visits may differ noticeably from the habits of your buyers. A team of lawyers, a support service and a small development studio will have different requirements for documents, confidentiality and integrations.

The practical takeaway is simple: treat ChatGPT as a familiar point of comparison for the customer. Then test your advantage within their workflow. A traffic study helps frame the question of promotion and positioning; decisions about architecture, model and product price should rest on your own measurements.