What to remember
- Good content in 2026 is useful, signed, lived and multi-format — not just well written.
- Video is no longer optional: it has become the ultimate trusted format.
- Being cited and mentioned by others matters just as much — sometimes more — than being well ranked.
- What worked just a year ago is now largely outdated. Here’s why.
Let’s be frank: the majority of content published on the web right now is poor. Not technically — it is well written, well structured, well optimised. But it reminds me of the scene from the film 'The Spanish Apartment' where Romain Duris plays the pipe... These contents are hollow. It’s just hot air !
Whose fault is it ? A race to produce that has forgotten the essential : content is only valuable if it provides something that cannot be found elsewhere. Right now, this obviousness has become an algorithmic requirement. Google measures it. AIs detect it. Readers feel it.
What has changed since last summer (June 2025)
Just a year ago, one could still be satisfied with content well-optimised text to rank correctly, multiply articles on close keyword variants, and build visibility primarily through Google. This model is collapsing for three combined reasons.
The explosion of AI content has killed the average. According to a study reported by Siècle Digital by the end of 2025, AI-generated content officially surpassed human content in November 2024 — peaking at 55% of the web in January 2025. Google is literally flooded with correct, factual, structured texts — and absolutely undifferentiated. The bar for "acceptable content" is now so low that it is no longer sufficient for anything.
The generative AIs have changed the way search is used.An increasing share of internet users no longer search on Google — they query ChatGPT, Perplexity, or Gemini. These tools do not return ten links: they cite one or two sources. Being the cited source has become a visibility issue in its own right.
Google has tightened its qualitative criteria. The successive updates of 2025 continued to penalise thin, generic content or content without a real signal of expertise. The E-E-A-T is no longer a recommendation — it is a filter.
Bias No. 1: without field experience, your content is worthless.
An article explaining "how to choose your hiking gear" can be perfectly written, well-sourced, and pleasant to read. If it is written by someone who has never set foot on a mountain trail, it will always be inferior to an article written by a mountain guide sharing their real recommendations, mistakes, and discoveries.
Why? Because experience produces information that cannot be found elsewhere: nuances, warnings, unexpected angles. This is exactly what Google seeks to identify with the first "E" of E-E-A-T — and it is exactly what generative AIs cannot authentically create.
The practical consequence:have your content written by people who know the subject from experience, or interview them and use that as the raw material for your articles. Co-constructed content with a field expert is worth ten articles written from other articles.
Bias No. 2: video is the new trust signal
The video is no longer a complementary format that we add "if we have time". In 2026, it is the format that proves that behind the content, there are real people with real expertise.
Immediate nuance, and it is significant: what applies to human video does not apply to AI video. Automatic video generation tools (synthetic avatars, cloned voices, generated edits) have experienced the same explosion as textual content — with the same effects. The web is now flooded with smooth, technically well-produced videos that are completely hollow in substance. Audiences are identifying them more and more quickly. So are the algorithms.
What creates trust in a video is not the production quality. It is the real presence: a face that reacts, a voice that sometimes hesitates, an authentic environment, a strong opinion that an AI would not have dared to express. A video shot with a smartphone by someone who truly knows their subject will almost always outperform a polished production driven by a synthetic avatar.
This is true for your readers — and it is also a signal that Google and generative AIs are increasingly incorporating into their assessment of a source's credibility. An active YouTube channel, Reels signed by real people, video testimonials from actual customers: all of these are proofs of existence that generated content cannot sustainably imitate.
Short video (Reels, YouTube Shorts, TikTok) generates audience and notoriety. Long video (tutorials, interviews, demonstrations) builds authority and trust. Both are useful — but for different objectives.
Our bias is : if you cannot invest in both, prioritise long video. It has a more lasting impact on your perceived credibility, it feeds generative AIs more effectively, and it can be more easily repurposed into other formats. But regardless of length: put real people in front of the camera. This is non-negotiable.
Bias No. 3: being mentioned matters as much as being clicked.
This is the least understood paradigm shift among content teams.
In a world where generative AIs answer questions directly, clicks are no longer the only indicator of visibility. What matters now is to be cited, mentioned, referenced— whether in a ChatGPT response, in a newspaper article, in a specialised forum or in a podcast.
These mentions feed what is called brand authority: the recognition of your expertise by third-party sources independent of you. And this authority is precisely what theAIuse to decide which sources they will cite in their responses.
Specifically : good content in 2026 must be designed to be citable. This means original data, clear positions, memorable formulations. Content that people want to share, cite, mention in their own productions.
What constitutes good content in June 2026
- Signed : an identifiable author, with documented legitimacy on the subject
- Grounded in reality : data, examples, lived experience — not a synthesis of other articles
- Multiforma t: text + video at a minimum, to reach different consumption contexts
- Citable : a position statement, an original statistic, a strong formulation that one wants to reuse
- Updated : dated content without revision is content that degrades
What is no longer sufficient: being well-written, well-structured, and properly optimised. That was the standard in 2022. In 2026, it is the minimum requirement to exist.
FAQ
Can AI produce good content ? Yes, if it is guided by real human expertise, reviewed and enriched with field experience. No, if it is used alone to produce volume.
Should we delete our old, overly generic content ? Not necessarily delete — but audit, consolidate or enrich. Updated content with real data and a genuine angle can regain value.
Does the length of an article still matter ? Not in itself. What matters is that the content fully meets the search intent — whether it is 600 or 3,000 words.