After the popularization of generative AI through ChatGPT in 2023, we witnessed the rise of conversational AI search engines in 2024, such as Perplexity and ChatGPT Search.
At the same time, Google is not far behind. The Mountain View firm is gradually rolling out its AI Overviews in Europe (after an initial launch in the US market), a new engine combining AI-generated responses and a classic results page.
This new type of engine meets new user demands: faster, simpler results, which now compile different content into a final personalized response.
However, one can wonder about the emerging uses of these engines: are the results generated by AI really reliable? Will conversational engines replace traditional search engines, signaling the end of Google? Or is it rather a new use case that complements current habits? Several studies are already revealing nuances to the tired refrain of “Google is dead.”
Conversational engines considerably simplify the way we search
Yes, AI search engines are the easy solution compared to traditional engines, making Google, Bing and others look like web archivists.
For 25 years, the ritual has been the same: the user asks questions, reformulates their problem in the form of keywords, types them into the online search engine, and is returned a list of articles deemed “relevant” to their topic.
A revolution in itself, but one that requires involvement in the selection of resources, and, depending on the topic, first and foremost looking for the right way to search for said information (have you ever asked yourself “How could I write this so that Google can find it?”).
Conversational consoles, on the other hand, accept any type of message, in both written and oral format, and fully manage the reformulation to return a documented, personalized response, which theoretically compiles the best results for the problem posed. Unparalleled practicality.
Reliability questioned by several studies
Technically, conversational search engines operate on the databases of traditional engines (i.e. Google, Bing, Brave, etc.).
This is essential to be able to process queries related to current events in particular.
As a result, the reliability of AI engines essentially depends on the quality of the source data. However, the web is not exempt from content containing erroneous information!
According to a study by the Columbia Journalism Review published in March 2025, nearly 60% of AI-generated responses contain errors.
AI engines are still little used
“AI Revolution”, “AI Bubble”… Beyond the enthusiastic media rhetoric, the reality is more measured. Yes, the use of AI engines is increasing, but Google is also experiencing notable growth: +21.64% monthly searches in 2024 vs. 2023 according to SparkToro.
For comparison, AI engines represent only 2% of the global market. Let's face it: ChatGPT, despite being the leader in this segment, generates 37.5 million searches per day, or 373 times less than Google.
A new complementary search usage
In reality, conversational engines are a new search tool that complements the mosaic of current uses. The same search path is now carried out on several platforms in parallel (note, for example, the rise of local queries on Instagram and TikTok). And AI consoles meet a relatively different demand: only 30% of interactions with ChatGPT are similar to traditional information searches.
Paradoxically, Google's recent launch of AI Overviews has increased user engagement on its engine, while reducing the click-through rate by 70% on organic results.
These figures put a damper on the passionate discourse surrounding conversational engines.
These new tools are developing rapidly, of course, but they fill a gap that "classic" search engines are not intended to address. They make it possible to offer a wider range of solutions in search journeys.
Rather than a revolution, we are therefore witnessing a gradual evolution in search practices. Tomorrow, search will be multi-modal, thanks to an ecosystem of traditional and conversational tools adapting to the needs of different types of searches.
But to be part of this ecosystem, AI engines still need to resolve their current reliability issues.
0 Comments