AI promotes feedback in the face of generic texts. The legitimacy of a site will also certainly be measured by the external validation of its ecosystem.
Danny Sullivan, Googlerecently spoke about the value of creating “unique, authentic and non-standardized” content. The latter must provide a unique vision, difficult to reproduce by others. It must address a specific situation or case study, avoiding global and theoretical rules. It must demonstrate real know-how and direct mastery of the subject.
This advice is in some way the emanation of gaining information. The principle of the latter is simple: assign a score between 0 and 1 to each content, depending on the quantity of new information it provides compared to what already exists. It measures their true added value based on the user’s browsing history. A way to respond to the production of generic content, which has become exponentially easier and less expensive with generative AI.
Precise and authentic experiences
Danny Sullivan anchored this with different examples. Among these, that of a running shoe store. He talks about the 10 things to consider when buying running shoes. It offers standard advice on sizing, arch support and cushioning. But this same store can also talk about “Why this customer’s shoes fell apart after 640 km.” With analysis of the wear pattern. With a video on lateral foam collapse. While the company’s first editorial approach refers to the creation of generic content, the second, put forward by Google, treats the subject in a non-generic way.
Another example proposed by the American giant, in the case of a real estate agent: the generic content could consist of “7 tips for first-time buyers.” With general advice on pre-approval, location and budget. Non-generic content could be “Why we waived the inspection (and saved $15,000).” With an overview of the sewer line and an explanation of the bidding war. “We offered $15,000 below asking price but waived the sewer line inspection because I personally crawled through the line and saw it was PVC, not concrete.”
Towards a battle to show specific examples?
Although these examples are not always very clear, for David GroultHead of SEO at Noiise, the advice from the Mountain View firm is in the right direction. “AIs seem more and more sensitive to “information gain”,” he points out. “Content must bring something really useful to the web. Like an experience, an angle, an analysis, a bias or original data.”
But isn’t there a risk of seeing an escalation of experiences appear, sometimes not really lived, to be displayed by Google? We know, for example, that there is a certain amount of abuse on Reddit to position yourself well. With, for example, the creation of fake expert subreddits or the purchase of positive votes. A game of cat and mouse could take place. “The risk of overbidding obviously exists, with players who will seek to artificially industrialize expertise or authenticity,” underlines David Groult. “But I think that AI will gradually become better at distinguishing real legitimacy from simulated expertise. This, thanks to the cross-analysis of reputation and coherence signals.”
So, moreover, for him, in this context, the role of Internet users remains very important. This is reminiscent of the function of backlinks in SEO. “This expertise can no longer be simply declarative,” analyzes David Groult. “To be recognized, it must be validated by third parties. What makes the difference today is not only the content produced, but the halo of signals that it generates around it. Namely, citations, mentions, discussions, reputation, engagement or consistency of speeches. This is what I often summarize with the CURL model, for Coherence, Uniqueness, Regularity and Legitimacy. AIs are indeed increasingly evaluating a brand ecosystem global, rather than just an isolated web page.”
Note that this distinction between generic and non-generic content affects certain queries more, whether in SEO or GEO. For example, in GEO, for high-volume informational queries, Google can summarize eight generic pages without citing any. For navigational queries, to search for a specific site or brand, the content of the home page or the product sheet may be sufficient. It’s up to the SEO to adapt their strategy.
Accessibility and user experience
For the moment, according to various feedback from the field, the “unique, specific and authentic” content rewarded can be varied. The shape can be important. For example, although the data distributed by Pappers (SIREN, reports, statuses) are accessible, the site has chosen accessibility. Thanks to structured data, the information becomes immediately intelligible to users and algorithms.
According to Similarwebbased on data estimated as of May 2026, Pappers is ranked second in France in its category, ahead of historical players such as societe.com, Infogreffe or Verif.com. While it has only existed since 2020. The gain of information via structuring is apparently reflected in the number of keywords. With 308,800 positioned keywords, Pappers captures long-tail queries for different French companies. The GEO dimension is also visible. The fact that 91% of traffic comes from France, with a bounce rate of 52% and 4 pages per visit on average, perhaps indicates that users and AI agents are finding a direct and precise response on the site.
Another example, according to Mathieu Chapon, founder of Peak Ace, optimizing the content of product sheets allows certain e-commerce sites to see their visibility increase in AI responses. Among the measures, theintroduction of user reviews to bring real use cases.
