ChatGPT Search Hits 20% of Google's Scale, Publishers Pay the Price

OpenAI's consumer product lead stood up at a dinner in New York this month and put a number on something publishers have been arguing about for two years: ChatGPT's search-like query volume, he said, is now roughly a fifth of Google's. Weekly users have gone from about 1 billion to 1.2 billion in just two months.

A Number Without a Yardstick

Adam Fry made the claim on November 2, and it spread fast because it's the first time OpenAI has put a specific ratio next to Google rather than just talking about growth in the abstract. But look closer and the comparison gets wobbly. OpenAI hasn't said what time period the Google figure covers, where that baseline number came from, or how it's counting a "search" in the first place. By its own definition, a search only counts when a ChatGPT query actually triggers a live web lookup — every other question, including plenty that a search engine would have handled, doesn't make the tally.

That matters because this isn't really a market-share claim, even though it's being read as one. It's a volume comparison with an undisclosed baseline, published by the company that benefits most from people believing the gap is closing. None of that means ChatGPT's search feature isn't growing fast — the user numbers back that up — it just means the "one-fifth of Google" line deserves more skepticism than the headlines it generated.

The Traffic Bill Lands on Publishers

Whatever the real ratio is, the people who actually write the news and reviews that AI answers are built from are the ones absorbing the cost. A widely cited TollBit analysis found that AI-driven search tools send roughly 96% less referral traffic — the visitors a search result actually sends back to the original site — than a traditional Google search does for the same query. When a chatbot reads five articles and hands you a three-paragraph summary, you rarely click through to any of them. Publishers call this the "zero-click" problem: the answer gets served, the page never gets visited, and the ad revenue that used to fund the reporting never shows up.

Being cited by an AI summary and being paid for it are not the same thing — and right now, almost nobody is pretending they are.

OpenAI has struck some licensing deals with large outlets for training data and real-time retrieval, but those arrangements are the exception, negotiated one publisher at a time by companies big enough to have the leverage. Everyone else just gets crawled.

India's Publishers Are Already Bleeding

This isn't an abstract worry for the Indian news industry — it's already showing up in the numbers. Industry tracking has pointed to referral traffic to Indian news sites falling sharply over the past year, with some outlets reporting declines as steep as 60-plus percent year-on-year as AI Overviews and chatbot answers intercept queries that used to land on their homepages. The Digital News Publishers Association (DNPA), which represents a large chunk of India's digital newsrooms, has already complained to the Competition Commission of India once before over Google's revenue-sharing terms, and it's now pushing the government to take a firmer position on how AI platforms use Indian journalism.

The regulatory gap makes this harder to fix than it looks. France has a "neighboring rights" law that forces Google to the negotiating table with publishers over snippets and summaries — a legal lever Indian publishers simply don't have. India's Digital Personal Data Protection (DPDP) Act governs personal data, not the copyrighted text of a news article or who gets paid when an AI model trains on it or answers from it. Until that gap gets addressed, infrastructure-level moves matter more than regulation: Code24 covered how Cloudflare started blocking unlabeled AI crawlers by default in September, which is the kind of technical fix small Indian publishers can actually use today, licensing deal or not.

For a newsroom or blog operator reading this, the practical checklist looks something like:

• Check your server logs or CDN dashboard for how much of your crawl traffic now comes from AI bots versus traditional search indexers.
• Don't assume "more AI traffic" means "more revenue" — audit whether chatbot referrals are actually converting to pageviews, not just citations.
• Watch what the DNPA and CCI do next; any precedent set against Google's revenue-sharing model will likely shape how AI licensing deals get structured in India too.

What Comes Next

The uncomfortable truth is that OpenAI's 20% number and the publisher traffic crisis are two sides of the same shift, and neither side has an incentive to slow it down. AI companies want bigger search numbers to justify valuations; readers want a fast answer over five tabs of blue links. The only people with a direct financial stake in the old model — the publishers who report, write, and fact-check the material these answers are built from — are the ones with the least leverage in the room. Until that changes, expect more numbers like Fry's, and more quiet traffic graphs trending downward on the other side of the ledger.