On the occasion of SMX Paris, an event taking place on March 14 and 15, 2024, Pierre Calvet, customer support and advice at babbar.tech, will participate in a conference on the potential replacement of classic search engines by conversational agents. We asked him our questions about the future of web search.
Pierre Calvet, SEO and customer support at babbar.tech
Pierre Calvet comes from a background in tourism and new technologies. In particular, he held the position of international SEO manager at Club Med and worked to optimize SEO at Govoyages and for various freelance clients. He is now in SEO and customer support at Babbar.tech.
With the democratization of conversational agents and the upcoming arrival of Google's SGE, what could Internet search look like in a few months?
In the majority of cases, the search on an engine will change. There are several categories of queries for which we can see that conversational agents will be perceived as very relevant. This is already somewhat the case, particularly in the information sector. Users will move their habits outside of sites to consume information in search engines, therefore Google, Bing, etc. This is a development that is quite logical in a way because it stems from the changes that Google has made to results pages in recent years. It will have an impact on websites that will be less visited since people will consume information on the search engine and in the search experience. Paid ads will always be present because they are part of the search engine business model. But on the other hand, we are going to enter into a discussion system with the engine. This will allow us to refine the need a little and improve the perceived quality of the response a little.
Furthermore, we will observe that the organic results will differ quite significantly between the classic results page and the SGE pages, for example. The links highlighted by SGE and Bing are fewer (up to three on average). It still changes quite a few things compared to the formats that we knew well until then. Users will get a response more quickly, and they will have less need to visit the sites, so site editors have every interest in experimenting with authority and expertise criteria.
From the point of view of the research experience, we can add, in addition to classic intentions, the possibility of creating texts or images to meet a specific need for illustration or synthesis. Which is something that did not exist—or very little—until now in search engines.
Finally, the impact on the actual quality of the results is really important. This is what explains the changes in the results. The expertise criteria are much more important because, as SGE gives less voice to the various existing sites, it cannot afford for the conversational agent to say anything, which we can see quite frequently at the 'actual hour. Expertise, authority, and trustworthiness—the trust that can be placed in a website—will therefore be particularly important.
Are we seeing behaviors ready to evolve on the user side?
Technically, users are already changing their operating mode a little. There is a part of search habits that has evolved because users discovered what they could do in particular with ChatGPT. It met some of their needs. Someone who wants to learn to code, instead of looking for an explanation for their error code on Stack Overflow, for example, goes to ChatGPT. The explanation is more personalized, and it allows you to better understand the problem. We will refine the subject or seek to synthesize it.
These uses had already started to change with answer snippets, position zero, “people also ask”... All these elements are already installed in our search habits. SGE is a bit like the next step that really allows us to change this web consumption habit. And I think that overall, the world is already changing on this, particularly through ChatGPT.
Are conversational agents as convincing as traditional search engines in the eyes of Internet users?
It depends, since there will be several types of EMS, precisely depending on the needs of the users. The engine adapts to needs, and with a conversational mode, it becomes a bit like a search assistant. In reality, there are already several types of search, with different actions accordingly, such as opening a Google Maps or Wikipedia insert depending on whether you are looking for a restaurant or a personality, for example.
What emerges from a few use cases studied is that the results of conversational agents are apparently good, and that is sufficient for a large part of the public.
In some cases, even if the answer is wrong, it seems right. That's what can be a little scary. And it should be noted that for the moment, certain intentions are not responded to properly by the conversational agents, so the result is not completely satisfactory for part of the public or for the actors.
Is Google the only player that can claim to transform the search experience with an AI-assisted chatbot?
No, we can notably cite Bing (now Copilot, editor's note), which has a good lead, moreover, in this area. But you shouldn't think that this makes Bing the new Google. Already, because that is probably not their intention. But also because users have their own consumption habits. For example, users appreciated having sources, which Bard (Google's conversational agent, which became Gemini, editor's note) did not really allow. However, a good number of them preferred the Google ecosystem because they were used to it.
The advantage of Google is that if it offers a new tool tomorrow, everyone will test it.
We can also add another actor: OpenAI (publisher of ChatGPT, editor's note), even if it is supported by Microsoft and therefore close to Bing, remains quite independent in its way of doing things. If they want to release a more reliable, more efficient, or more efficient model, they already have the user base that would serve them.
For which sectors is the arrival of conversational agents within and alongside search engines a good prospect? Who will suffer?
It's hard to read a crystal ball. Those who will likely succeed are those who will put their expertise forward. Presumably, the big players, the e-retailers, and the companies that put in the means to give a voice deemed expert will put their content forward. And so, conversely, bloggers and small players will most certainly be negatively impacted. When a response from a powerful e-retailer masks theirs because it seems more legitimate, this is where a small player risks losing its say, at least in the response generated.
The EEAT criteria should be used to identify who is an expert. It will therefore be extremely important to produce quality content—this was already the case, but it will be even more so—by, for example, bringing in an expert whose words can be a real support on which the generative tool will be used as a basis for its conversations with users. For the robot, user signals will also be used.