OpenAI has announced that the volume of web searches carried out through ChatGPT has reached 20% of Google's search volume. The person who shared the figure was the company's head of ChatGPT consumer products. At first glance it's a striking ratio; but what that 20% actually means depends on how it's counted.

What happened?

For the first time, OpenAI publicly put a concrete number on the usage volume of ChatGPT's web search function. The point of comparison is Google's global search volume. The figure floated is 20%. The statement reads less like a product announcement than a positioning move.

Why does it matter?

The real question here is methodological. "Web search" and "retrieving information inside a chat" are not the same thing. On Google, a query means an intent; the user searches, sees a list of results, and clicks a link. On ChatGPT, the user asks a single question, and when needed the model runs multiple web queries in the background. In other words, a single user intent can technically generate several "searches." In that case, the 20% may represent not equivalent user behavior, but two differently defined metrics placed side by side.

This distinction is not trivial. Because what OpenAI is measuring probably doesn't map one-to-one onto what Google measures. For decades the advertising world has priced the "query" unit around intent and conversion. The automatic queries ChatGPT generates in the background, however, carry no commercial intent; the user sees no ads and doesn't browse a results page. So approaching Google in volume doesn't mean approaching it in revenue or market power.

The timing of the announcement is itself a data point. AI companies tend to make growth metrics more visible during periods when the funding narrative is under strain. Figures like growth rate, user counts, and usage intensity function as leverage in valuation debates. That doesn't mean the number is wrong; but it's a reminder that which number is shared, and when, is also a message.

Is the definition of search changing?

The more accurate reading is this: rather than dethroning Google, ChatGPT is pulling part of "search" behavior toward itself. Open-ended queries — those asking for a definition or a comparison — are especially migrating to the chat interface. For transactional and local searches like "nearest pharmacy," "flight time," or "today's match score," Google still dominates.

Query typeBetter-suited interfaceWhy
Open-ended, explanatoryChat (ChatGPT)Synthesized answer, follow-up questions
Transactional / local (maps, stores)Classic search (Google)Location, live data, links
Commercial intent (purchasing)Classic searchAds, price comparison
Research / in-depthBoth togetherSynthesis + source verification

Google sees this shift too, and is responding by embedding AI summaries into its own interface. So the battle isn't "search engine vs. chatbot"; both sides are converging on an interface where the synthesized answer and the list of blue links blend together. The difference lies in who holds on to their existing distribution power. Google Search is accessed by default from billions of devices; for ChatGPT to reach that scale, it first has to become the default.

How would this play out in Turkey's search market?

In Turkey, search is a field overwhelmingly dominated by Google. But the spread of AI summaries carries a direct risk of lost traffic for publishers; a user may get the answer on the results page and never click through to the site. Competition authorities' scrutiny of AI summaries feeds on this very concern. If the synthesized answer becomes the norm on both Google and ChatGPT, content creators' visitor- and ad-revenue model comes under pressure on two fronts at once.

What's next?

There are a few things to watch. First, whether OpenAI will over time clarify how it defines this metric; nailing down the "search" unit is what will make the number meaningful. Second, the commercialization of ChatGPT search — once ads or shopping integrations arrive, volume turns into a genuine revenue comparison. Third, how regulators will intervene in the balance between synthesized answers and publisher traffic.

In the end, the 20% ratio shows that OpenAI's growth is real; but reading it as the end of Google is premature and most likely wrong. The sounder interpretation is this: search is dispersing from a single box into multiple interfaces, and the real competition will be fought not over volume, but over who gets to hold the user's default reflex.