ChatGPT ≠ Google-scale.Google: ~14 B searches/day; ChatGPT: ~37 M (~1 : 400). Only ~15 % of ChatGPT prompts look like classic “search”; most are writing/code tasks. Google’s own search volume grew 22 % in 2024 and still holds >90 % share. An LLM citation is nice for credibility, but it won’t move traffic or revenue anytime soon.
back
6 comments
The 37M/day is an estimate from Rand Fishkin that often gets quoted as gospel. It is based on limited external data. OpenAI mentioned 1B/day - and had significant overall growth in usage since then.
Also, 1 search on ChatGPT easily replaces 5-10 searches on Google.
Many B2B SaaS companies already get the same amount of leads from ChatGPT that they get from Google. Because clicks from ChatGPT are better informed and have a significantly higher conversion rate. I am talking up to +700% CVR vs traffic from Google for some companies.
We looked at the same data that Rand Fishkin used and definitely came to a different conclusion.
But users are not clicking on search results in google. They get satisfying response from Gemini & end there. It is a good thing for users but bad for inbound traffic to websites.
Also bad for users who get completely bullshitted responses.
1. (Anecdotal) I'm barely using Google anymore. I'm using ChatGPT for a ton of queries and getting far better results.
2. Antitrust actions might (should) strip Google of their "panes of glass" with which they force Google Search as the default. Most Google Search queries are simply the result of defaults. Once those defaults are gone, those queries will go elsewhere.
To add on top of this, how many Google searches also contain Gemini answers in the search results? I've been seeing more and more, especially for code and general factoid searches.
Turtles all the way down.
It is definitely interesting to see how much public opinion has been shifting on Google in AI as of late. I wonder what the main force for that is…
1. From 2010 - 2023(ish), Google slept on DeepMind and allowed OpenAI to steal the narrative. That led to a boom in AI development outside of the Google labs. This is akin to Microsoft's loss of internet/web. A half dozen trillion dollar companies emerged from Microsoft's stumbling, and we're likely to see the same with Google's missteps.
2. After the rise of so many AI startups in the 2020 - 2023 period, the end of Google was being forecast by many. Most of Google's revenue comes through search, and everyone (Nadella, Altman, investors, et al.) were talking about the incremental value of search - Google had to retain 100%, and other players just had to grow.
3. The Google founders, sensing a major blow to their cash cow, took a break from their zeppelin startups and came back. They gave Deep Mind more autonomy, took a knife to product culture, and empowered and encouraged everyone to innovate. The whole company has been re-focused and told they must win AI or face extinction.
4. As of 2025, Google has been killing it on their releases. Gemini, Veo, ... you name it, and they've got industry-leading developments that out-perform and undercut the competition. It's beginning to look like not only will Google not die, but that Google could wipe the field with their AI superiority. It looks like they'll be able to dance circles around OpenAI.
The looming threats are (1) DOJ antitrust eroding Google search ingress and (2) other players stealing Search TAM without the new AI markets being able to replace the search / ads revenue.
Any non-Google players would be wise to put antitrust pressure on Google. Even after the current case ends, they should try to strip away defaults on web and mobile. Stop Google from being able to deploy AI through Google Search, Android, and Chrome. Make Google use the same word of mouth marketing that the rest have to.
It'll be an exciting series of battles ahead.
Yet.
And even that ignores Google runs an LLM on their search too.
splitting these are not a good measure. people who know how to search (a skill the latter generations seem to have lost) also searched for the lowest common denominator coding recipes and produced naive code just like people do with current llm models. only sellers of the llm models make the distinction. it's all search.