OpenAI just told Google it now owns 20% of search intent, then immediately kneecapped its own power users by forcing them to split tokens between chatting and coding.
The Summary
- ChatGPT now handles 20% of Google's search query volume, measured by actual web searches, not simple definitions or how-tos
- OpenAI is merging ChatGPT's Chat and Work tabs, forcing shared usage limits between ChatGPT and Codex starting December
- Power users who relied on separate pools for conversation and code generation now face allocation constraints across Plus, Pro, and Business tiers
The Signal
OpenAI's consumer product lead Adam Fry dropped the 20% number at a Monday press event, and it matters because of how they're counting. They define a search as any query that triggers a web search, not the easier stuff like "define photosynthesis" or "how to hard boil an egg." That's pure search intent. People choosing ChatGPT over Google for the hard questions, the ones that require synthesis and context.
One-fifth of Google's volume puts ChatGPT somewhere north of 200 million searches per day if Google is still processing around a billion. That's not a rounding error. That's "Ask Chat" becoming the default verb for a meaningful slice of the internet.
"In many circles, 'Ask Chat' has replaced 'Google it.'"
But here's where it gets weird. The same week OpenAI is crowing about search dominance, they're merging ChatGPT and Codex into a shared usage pool. Starting in December, Plus, Pro, and Business users won't have separate buckets for conversation and code generation anymore. You get one allocation. Burn tokens asking ChatGPT to summarize meeting notes, and you have fewer left for Codex to write your API endpoints.
The official help docs confirm it: usage limits are merging. The discussion on Hacker News is predictably salty. Developers who structured their workflow around coding in one tab and research in another now have to ration. This isn't just UX consolidation. It's capacity management disguised as simplification.
Key trade-offs of the merge:
- Flexibility: one interface for all tasks vs. cognitive overhead of tracking one pool
- Power users: developers who leaned hard on Codex now share tokens with casual chat
- Business tier: teams that scaled Codex usage separately now compete internally for allocation
The timing tells you something. OpenAI is growing search share fast enough to brag about it publicly, yet they're tightening usage constraints on paid tiers. Either compute costs are worse than they're letting on, or they're preemptively managing demand before the next model drop strains infrastructure even harder. Both can be true.
The Implication
If you're a developer using ChatGPT for work, the December merge means auditing your workflow now. Figure out which tasks actually need the premium models and which can run on lighter alternatives or local tools. The people who adapt fastest will be the ones who already treat tokens like a budget, not a bottomless well.
For Google, 20% should be terrifying. Not because ChatGPT is better at search across the board, but because it's better at the high-value queries where users want answers, not ten blue links. If that wedge keeps growing, the ads that subsidize everything Google does start losing inventory. Watch for Google to push Gemini integration harder into Search and probably start talking about "agentic search" in every earnings call.
Sources
The Verge AI | Mashable Tech | TechCrunch AI | Hacker News Best