Discover and AI Mode Now Read You as an Entity, Not as Text

If you've been watching your traffic graph these past months and seeing not a dip but a cliff — don't panic just yet. It's not only about you. Google is expanding the number of links inside AI Overviews and AI Mode across the board, and it's already hit most publishers. That's a fact.

And right behind it comes a second piece of news that looks minor at first glance but actually matters more: Discover now picks articles for users not by keywords and not by site topic, but by Knowledge Graph entities the person has already encountered in their reading history. In other words, the system isn't asking "what is this article about" — it's asking "about whom, about what specifically" — and matching that against what the user has already read.

Why this isn't a small thing

You used to be able to write vaguely and still rank — the algorithm forgave fuzziness because it worked with keywords and general topics. That doesn't work anymore. The Knowledge Graph is an entity graph: companies, products, people, brands, teams. If your text doesn't name things directly, the system can't tie it to a specific entity — and it simply won't connect you to a reader who's already interested in that topic.

Isn't this a complete reversal of how content logic used to work? You used to write "for people, and a little for the robots." Now you need to write so that both the human and the knowledge graph understand, unambiguously, exactly who and what the text is about.

What to actually do about it

Call things by their real names inside the text. The product, the company, the team, the specific term — directly, not vaguely. "One of the platforms" instead of an actual name is a signal going nowhere for the Knowledge Graph. It has nothing to latch onto.

Next: competition for the click is rising at the same time. If AI Overviews and AI Mode show more links inside the answer, that means within those links you're not competing with ten sites on a results page — you're competing with ten sites inside one generated block that the reader sees for three seconds and either clicks or doesn't. Your content has to grab attention from the first line and immediately make clear why someone should click through to you instead of reading the AI's summary and closing the tab.

And third — get ready for AI Mode to stop being an experiment and become the primary results page. That means optimizing "for the traditional results page" is already yesterday's approach. You need to think about how your content gets quoted and retold inside an AI answer: clear phrasing, explicit facts, unambiguous entities, no filler.

My take on this

I've said it for a long time: search isn't about "tricking the algorithm," it's about making it unmistakably clear what you're actually saying. Google just raised the bar on clarity. The people who've spent years writing vaguely to chase keyword reach are feeling it first. The ones who wrote with substance and called things by their names — they dropped less, or didn't drop at all.

For publishers and content projects, this is a signal to rework editorial policy: fewer generic phrases, more concrete entities in the text. For everyone else, it's a reason to check how clearly your site actually explains who you are, what you do, and who you work with. If Google struggles with that, a real reader struggles with it even more.

If you want to figure out how ready your project is for this AI-search logic, and how to restructure content around entities and the Knowledge Graph, let's talk it over in an audit and consultation.