AI Overviews: The Impact on Social Media in 2026

AI Overviews and Social Media: How to Adapt Your Content Production?

Social Media Enters the Research Arena

July 22, 2026, marks the arrival of Google’s AI Overviews in the French market, placing France among the last major countries to benefit from this update. This introduction has raised concerns among SEO professionals, worried about the potential impact on traffic and brand visibility. Social media teams may soon share these concerns, as artificial intelligence models increasingly incorporate content from social media into their responses. To explore these implications, a discussion was held with Sophie Gonçalves, Head of Content and SEO at Iconosquare.

Just like SEO professionals, those in social media feel a certain apprehension regarding the arrival of AI Overviews. Sophie Gonçalves acknowledges this concern, although she puts it into perspective. She explains that any new development can be intimidating, as it introduces new behaviors among users. However, since France was one of the last countries to receive this update, it has allowed for observation of effects elsewhere before experiencing them directly. While a decrease in traffic is a reality, the arrival of AI Overviews does not mean the end of social media.

Social media is not dying: it is becoming a source from which AIs draw.
Sophie Gonçalves, Content Manager, Iconosquare

The question then arises as to whether this transition requires a complete relearning. According to Sophie Gonçalves, this is not the case. She asserts that while the way of thinking about content is evolving, there is no need to reinvent the wheel. The questions to consider beforehand are more important:

• Does my caption convey information independently?
• Is my video accompanied by a transcription?
• Is my expertise clearly expressed?

Well-designed social content for a human audience is already suitable for AI, provided it is clear, useful, and expert.

On the side of French brands, Sophie Gonçalves observes a phase of observation rather than a rush to optimization. Teams are testing queries, analyzing what AIs say about them, and comparing themselves to competitors, but few have structured a strategy yet. This caution is deemed healthy by the expert, as it is better to understand the starting point before modifying the entire editorial line.

At Iconosquare and more broadly within the Groupe Positive, visibility in search engines is a constant concern for marketing teams.

What Should Content Look Like to Be Cited by AIs?

Good Writing Practices

The basic principle is simple: AI answers questions. Beyond merely occupying a space in the feed, the challenge is to provide the best possible answer to a specific question to be cited. Sophie Gonçalves explains that the first reflex to adopt is to start from the real needs of the audience and the questions they are asking, in order to respond clearly in the content. She suggests three concrete best practices:

• Provide the answer as high up as possible: whether it is a statement, an answer to a question, or a definition, it is crucial to provide a clear and "autonomous" answer right from the first sentence, similar to a well-thought-out title. AIs place great importance on what opens the content, so information should no longer be hidden after several lines of teasing.

• Ensure that the caption stands on its own: the caption is often underestimated, as it is thought that everything has already been said in the visual, carousel, or video. However, if the information exists only in the visual or an untranscribed video, the AI will not perceive it. On the other hand, the AI reads the caption, which must therefore contain precise facts, figures, and examples, without overloading with keywords, while maintaining the brand's authenticity.

• Provide real expertise: AIs increasingly value demonstrated expertise. A reasoned opinion, a demonstration, or a concrete feedback is much more likely to be cited than a generality, even if it is inspiring.

Platforms to Favor

In terms of formats, educational content and a question-and-answer logic are favored, as they use the natural vocabulary of a subject and respond to concrete needs. Again, the caption becomes a crucial strategic space.

Whether on Google or in LLMs, users ask a lot of questions, such as "what is the best social media management tool." Thus, a question-and-answer format, which uses the natural vocabulary of the subject, clearly favors citation.

Sophie Gonçalves, Content Manager, Iconosquare

Regarding platforms, the shift is quantifiable. Sophie Gonçalves shares figures from a survey by Adweek published in January 2026. YouTube is cited in 16% of AI responses over the past six months, compared to 10% for Reddit. When examining only citations from social media, the trend is even more pronounced: YouTube's share rose from 18.9% in August to 39.2% in December, while Reddit's share fell from 44.2% to 20.3%.

However, this does not mean that AIs are cut off from Instagram or TikTok. Technically, they can access them. The real challenge is that they find much less readable content there. An AI does not watch a video; it reads the surrounding text. YouTube, for example, naturally offers this possibility (not to mention its owner, Google). Conversely, a TikTok video without transcription remains almost a black box.

Sophie Gonçalves also notes that even though Reddit is in decline, discussions from this platform remain highly cited: "it's pure text, with depth of discussion. In other words, it's exactly what models know how to exploit."

What Mistakes Should Be Avoided?

The first mistake, according to Sophie Gonçalves, is reversing the order of priorities by writing for AIs and bots before writing for the audience. "People are not fooled; they can sense when we write for machines," she warns. The risk is losing authenticity, whereas it is precisely the brand voice that makes the difference, even in the eyes of generative AIs.

Next, more technical mistakes can occur. Stuffing captions with keywords is unnecessary. The valued signal is expertise, not the density of keywords in a single post. Moreover, optimizing content without checking that the platform allows the AI to read it is akin to working in vain.

Finally, the social media expert highlights a common error in prioritization: focusing all energy on community content and earned media, at the expense of one's own content; a territory that the brand knows inside and out, and which AIs often cite the most, provided it is clear and structured. "Before trying to influence what others say about you, make sure your own content is readable enough to be cited."

Where to Start?

For a team looking to adapt without completely overhauling everything, Sophie Gonçalves advises starting with an audit. By typing key queries into AIs, observing who is cited, in what position, and what is said about the brand, it is possible to identify gaps and develop an action plan.

In practice, the focus should be on writing: captions, alternative text for images, video transcriptions. It is not about starting from scratch and building a new editorial line, but rather rethinking how to make information readable.