Since the release of ChaGPT, OpenAI has become the embodiment of generative artificial intelligence for the general public. Praised for certain things, including its technology, and criticized for others, including its hesitations regarding governance and a lack of clarity on ethics, it raises a recurring question: Is it becoming the new Google? And, even if that is not its intention, should it be? And if so, could it?
Behind this comparison lies a deeper issue, which is the economic sustainability of the sector (AI heading for an economic dead end?).
To get a clear picture, we need to ask three questions: Does OpenAI want to, should it, and can it?
In short:
- OpenAI is not seeking to copy Google but to offer a cognitive agent focused on task execution rather than link searching.
- It is constrained by its economic dependence on ChatGPT, due to a lack of ecosystem or diversification.
- It faces major obstacles: lack of everyday use, agile competition from Google, high costs, and disintermediation.
- Advertising is not a viable option, hence the exploration of freemium, transactional, or B2B models.
- Regulatory and transparency challenges will need to be overcome to gain trust and legitimacy.
Does OpenAI want to become Google?
It’s tempting to say yes without thinking twice, given how similar their strategies are: burn cash to destroy the competition and then become a hegemonic (and highly profitable) leader. What’s more, since late 2023, OpenAI has integrated a web search feature into ChatGPT. This conversational approach to search is designed to contextualize responses and reduce friction.
But this development does not necessarily mean that OpenAI is seeking to become a Google clone. Its strategy is not to offer a search engine based on sponsored links, but rather a cognitive assistant that captures an intention or need, reformulates it, and provides a directly actionable response.
The goal is to shift the value of search toward task resolution. While Google organizes the web and provides paths to information, ChatGPT aims to deliver a comprehensive and definitive answer to a question. This is not a head-on collision, but rather a paradigm shift from hyperlinks to proactive agents.
Sam Altman, CEO of OpenAI, says he no longer uses Google on a daily basis, preferring ChatGPT. But could he say anything else? However, he did acknowledge that ChatGPT is not ready to replace Google for all use cases.
It would therefore be less about creating a clone of Google and more about making its use obsolete in the long term.
Should OpenAI become the new Google?
Here, the answer is clearer and more straightforward: it doesn’t really have a choice.
OpenAI is now a virtually single-product company. Most of its revenue comes from ChatGPT Plus ($20 per month), the sale of its API, and its partnership with Microsoft. Unlike Google, which can balance its revenue streams with multiple products, OpenAI has no leverage for diversification or an integrated ecosystem to lock in and retain its users.
OpenAI is therefore completely dependent on the massive success of ChatGPT. This dependence is exacerbated by its business model: the more ChatGPT is used, the higher the costs, because LLMs are not economically viable at scalewithout a real breakthrough that would reduce the inference costs associated with each generated response.
In 2024, OpenAI posted $4 billion in revenue and $9 billion in expenses, which means that the company cannot grow without putting itself at financial risk and must switch to massive, recurring, integrated use in order to hopefully reverse the trend. This is exactly what Google represents in our daily lives.
OpenAI is therefore forced to become a platform and not just an API provider because, as long as it is not the primary interface between us and the digital world, it remains disintermediable, replaceable, and economically fragile.
Can it succeed?
This is where things get complicated, because there are many obstacles.
The first is that OpenAI has no ecosystem. Google is in our phones (Android), our emails (Gmail), our documents (Workspace), our browsing (Chrome), our videos (YouTube). For businesses, it is an essential advertising platform. This omnipresence allows it to collect data, enrich its models, personalize its services, and virtually lock its users into its ecosystem.
OpenAI remains a niche product. ChatGPT is used occasionally, without any daily use, and is in no way comparable to an email inbox or a default search engine. There is no “ChatGPT reflex” outside of a population that is curious or aware of AI.
What’s more, Google isn’t standing still and has already started to integrate generative AI into its search engine with AI Overviews, Gemini modules integrated into Workspace, and experiments with synthetic responses in Chrome. We’re not talking about a complete overhaul, but rather a gradual infiltration: AI is trying to slip into existing uses without asking users to change their habits, and that’s important.
Where ChatCPT has to create a use case, Google is reinforcing its own. When ChatGPT asks us to use a new interface, Google infiltrates the ones we use every day. This allows Google to experiment on a large scale, refine, adjust, and cut the ground from under OpenAI before the agent becomes second nature to the general public.
Added to this is something new: a form of reverse disintermediation. Today, it is wrappers, businesses that encapsulate OpenAI’s models in business solutions that capture usage and value (Wrappers, deeptechs, and generative AI: a profitable but fragile house of cards). OpenAI feeds but does not distribute, is invisible to the customer and end user, and provides but does not build loyalty. As long as it remains at this level of the chain, it is interchangeable and therefore fragile.
But there is above all an economic and structural aspect: the marginal cost of an AI query remains much higher than that of a traditional search.
A text search on Google costs 1 to 5 microcents ($0.0000001), which can be explained by optimized architecture, proprietary servers, and an engine that leverages pre-existing indexes. Conversely, a query by GPT 4 involves real-time calculation of several hundred tokens, which can involve web browsing, complex rewriting, and even multimodal interaction. The cost of each query is estimated to be between $0.01 and $0.1.
This difference is enormous: OpenAI pays thousands of times more for each response than Google, and 100 million active users generating a few queries each day would represent several billion dollars per month in infrastructure costs.
Without a technological breakthrough (lighter models, optimized hardware, etc.) or a radically different business model (value per task, not volume), OpenAI cannot grow without putting itself at risk.
What about advertising?
Some suggest that OpenAI should follow Google’s lead and integrate advertising into its responses, but this would be to deny what is essential to the agentic approach: trust, neutrality, and transparency (What do the AI giants really want?).
An AI that recommends a sponsored link is no longer an advisor but a salesperson in disguise, and the risk of losing credibility is too great.
Moreover, while he says he is pragmatic on the subject and not closed to advertising, Altman admits that he prefers to make money in other ways, through the use and value of his products (An Interview with OpenAI CEO Sam Altman About Building a Consumer Tech Company).
OpenAI knows this, which is why it is exploring other avenues: a differentiated freemium model, transactional uses (booking, purchasing, task execution), or B2B positioning in high-value verticals. These models are based on value created, not attention captured.
What remains to be done
Even with cutting-edge technology, OpenAI will face two other major challenges.
The first is regulatory. The AI Act in Europe (The European AI Act for dummies), appeals by news publishers, and intellectual property issues are all converging to limit free access to web data. Generative AI would no longer be able to train and synthesize at will without royalties or restrictions, and the dismissive tone with which Altman sidestepped the issue at the latest TED conference, saying, “I don’t care, you don’t understand” (Sam Altman at TED 2025: Inside the most uncomfortable — and important — AI interview of the year) may not be tenable in front of regulators.
The second is epistemological. If ChatGPT becomes an entry point to information, it must make its answers verifiable, traceable, and auditable. A synthesis without links to sources is a cognitive blind spot. OpenAI will have to invent a pedagogy of transparency to maintain trust.
Bottom line
OpenAI does not want to become Google. It wants to bring about a new way of interacting with information, services, and tools, but this ambition paradoxically pushes it to replace Google in our everyday lives.
To do so, it will have to invent an advertising-free business model, an OS-free platform, native distribution-free usage, and trust without asymmetry.
In short: it doesn’t really want to, it can’t yet, but it doesn’t really have a choice. The next Google may not be a search engine, but an agent, but it remains to be seen whether it will speak with the voice of OpenAI.
Image credit: Image generated by artificial intelligence via ChatGPT (OpenAI)
