This was one of those weeks where data contradicted assumptions. And that might be the most valuable thing a set of numbers can do: stop you in your tracks and make you look again at what you thought you knew.
I picked four topics that deserve more than a quick scroll. Not because they are new, but because they bring evidence where we previously had only opinions. From AI Overviews to local SEO, from search intent to Google's regional rules, each topic challenges an assumption that many of us treat as fact.
Le Monde and AI Overviews: the catastrophe that never arrived
The dominant industry narrative goes something like this: AI Overviews are eating publisher traffic. It is a compelling story, but Le Monde brings numbers that say otherwise. Since AI Overviews launched in France, the publication has reported no abrupt audience decline. In fact, in July and August they added 24,000 new subscribers per month, historic records.
Context matters, of course. Le Monde generates 36% of its traffic through mobile apps, which are unaffected by AI Overviews. And news content holds up better than service content (health, recipes), where declines reach up to 50%.
But the most interesting number is this: visitors coming from ChatGPT convert to subscriptions at 260 times the rate of those from Google Discover. 260 times. That is not a statistical error, it is a signal that search behavior is changing, and those who adapt do not just survive, they grow.
What this means for the Romanian market: if you have an editorial product or a content-driven business, do not rush to mourn lost traffic. Look at direct channels (apps, newsletters, community) and at the quality of AI-driven traffic. Volume may drop, but value per visitor can rise.
There is another thing worth noticing: Le Monde did not react passively. They invested in the mobile app, in the subscription experience, in exclusive content. The lesson is not that AI Overviews does no harm. The lesson is that publishers who diversify their traffic sources are harder to displace.
Local SEO in the AI era: reviews beat ratings
A study of 120,000 mentions across 5 AI models (ChatGPT, Gemini, Claude, Perplexity, Grok) showed something many suspected but few had measured: review volume matters more than review score for AI visibility.
Each AI model has distinct preferences. Claude favors local businesses with a conservative approach. Gemini pulls from Google Maps. ChatGPT generates concentrated lists. Grok looks at Instagram and credentials. Perplexity cites live sources. There is no universal recipe.
The study proposes a simple framework: BARS (Business data, Authority, Reviews, Social). And the strongest visibility predictor in the restaurant category? Photo count. Restaurants frequently mentioned by AI have 3 times more photos than those ignored.
Good news for businesses in Romania: brand size does not guarantee AI visibility. An independent restaurant in Cluj with 500 reviews and 200 photos can outperform an international chain with a 4.8 rating but only 50 reviews. It is a matter of volume, not vanity. And that is a real opportunity if you have a consistent, well-documented brand.
In practical terms, the local SEO strategy for AI comes down to: complete your Google Business profile with every available data point, actively request reviews from customers, upload quality photos (many, not just 5), and maintain a social media presence. Sounds basic? Perhaps. But the difference is that now we have data from 120,000 mentions confirming it works.
Search intent does not look the way you think
68% of Google searches in the US end without a single click. In 2019, the figure was 49%. And with AI Mode, queries are 3 times longer, using action verbs like "explain," "compare," "summarize." People are no longer looking for links. They are looking for answers.
The data gets more interesting when you look at what triggers AI Overviews. Four-word searches activate AI Overviews in 48% of cases. But questions containing "why" activate them in 92.3% of cases. That means your content structure needs to change. FAQs and paragraphs that directly answer questions are no longer nice to have. They are infrastructure.
Google itself has redefined search intent through 5 pillars: explore, decide, learn, create, act. It is no longer about informational vs. transactional. It is about what the user wants to do with the information.
An eye-tracking study showed that images and product listings capture 100% of visual attention but generate 0% click-through. AI Overview text? Converts at 86.4%. If you are measuring visibility, make sure you are measuring what matters.
For content teams, the practical implication is clear: every article should answer at least 2-3 specific questions in short, direct paragraphs. Not because it is trendy, but because that is the structure AI cites. And if you are not in AI Overviews, you are invisible to an increasingly large share of your potential audience.
Google documents regional search: what it means for the EEA
Google published a guide documenting search features specific to the European Economic Area (EEA), Turkey, and South Africa. For Romania, this means we finally have clear documentation on what types of results are available and how to qualify for them.
The EEA has the most options: aggregator units for comparison sites and supplier units for direct providers. Requirements vary significantly: from structured data markup to feed/API integration, down to simple enrollment forms.
If you are in tourism, e-commerce, or financial services, complexity increases considerably due to data requirements. But if you have a local business or a services site, the supplier unit requires only a crawlable website with no additional technical setup.
My recommendation: do not wait for the dust to settle. Check Google's guide now, see which category you fall into, and start the eligibility process. Those who qualify first will have a visible advantage in regional SERPs.
What we have seen at difrnt. is that many businesses in Romania do not even know these options exist. Hotels, for example, can appear in comparison units alongside Booking or Trivago, but they need to go through an eligibility process. Doing this now, when competition is low, creates an advantage that latecomers cannot easily recover.
Four topics, one simple takeaway: assumptions in digital marketing do not last long. Numbers replace them. And those who move on data rather than opinions save time and money.
