The keyword loses its central role. What decides visibility is the identity machines can verify.

The number that says it all

A page ranking first on Google now loses 58% of its clicks the moment an AI Overview appears above it. That is Ahrefs' measurement across 300,000 keywords, published in February 2026 using December 2025 data, benchmarked against December 2023, before AI Overviews existed. The figure doubled in a year: it stood at 34.5% in April 2025.

This is not an isolated case. Seer Interactive measured a 61% collapse in organic click-through rate, and 68% on paid. Pew Research Center found that users clicked a traditional result in only 8% of sessions where an AI summary appeared, versus 15% when it did not.

Semrush's data shows the same trajectory: the share of queries triggering an AI Overview rose from 6.49% in January 2025 to a peak near 25% in July, before settling around 15.69% in November. Other firms, including Advanced Web Ranking, place that figure at 48% of queries by March 2026. The trajectory is not reversing.

What this means, technically

These numbers do not describe a traffic slowdown. They describe a change in mechanism. A Google search no longer compares a keyword against millions of pages to rank ten of them. It compares an intent against a set of recognized entities capable of answering, and synthesizes a response citing an average of three or more sources.

This shift built up in layers. BERT, launched by Google in October 2019, taught the engine to understand the meaning of a sentence. MUM, announced in May 2021, let a query connect to multiple sources and formats. AI Overviews, rolled out broadly from 2024 onward, are the visible outcome of that shift.

A page is still a container. An entity is an identity. The container can change, a redesign, a migration, a domain change, without affecting the identity, as long as it was built correctly. It is that identity, not the page, that the system now queries first.

The new marketing edge: being cited beats being clicked

According to Seer Interactive, a brand cited inside an AI Overview earns 35% more organic clicks and 91% more paid clicks than an uncited brand, on the same queries. Ranking first is no longer enough, but citation now pays more than it used to.

That changes the question a marketing director needs to ask. It is no longer just how much organic traffic this content generates. It is: does the machine cite this brand when it answers on the user's behalf?

HubSpot saw its organic traffic decline 70 to 80% between late 2024 and early 2025. Chegg sued Google over this in February 2025, with revenue down 24% year over year. These are marketing budgets that failed to anticipate the shift in value, from the page to the entity.

What machines check instead of the keyword

Four things, in this order, before deciding who to cite.

Identification: does this entity have a stable name, a clear type, an unambiguous description. Disambiguation: does this entity get confused with a homonym, a neighboring brand, a close concept. Connection: is this entity linked to its creators, sources, proof and external profiles. Validation: do independent signals confirm this entity actually exists and deserves to be cited.

A company that checks these four boxes exists for a machine. A company that checks none of them stays a collection of pages, however well written.

How this gets measured

This is exactly what the SFT method quantifies, across three pillars: Structure, Flow, Trust. A score of 0 to 20 signals an invisible or incoherent entity. A score of 81 to 100 signals a strong, connected, trusted entity. Entity Confidence synthesizes these three pillars into a single number, measured live.

What this means for a business, this week

Three concrete actions, in priority order.

First, check whether the company is cited in AI Overviews and in ChatGPT or Perplexity answers on its brand and product queries. Then fix structure before content: incomplete schema or a missing sameAs prevents an otherwise solid entity from being recognized. Finally, reallocate part of the content budget toward building verifiable external proof: coherent profiles, third-party sources, independent mentions.

The delay does not get fixed with a link campaign. A bought link no longer fools a system that can check identity coherence across multiple sources in a single query.

What still holds true

Content does not disappear. Its role changes. It is no longer the entry point, it becomes the proof. A strong entity still needs solid content to exist. But content alone, without a clear identity behind it, is no longer enough to exist in an AI-generated answer.

2026 does not bury SEO. It buries the version of SEO that still thought in pages.

See how the SFT method builds this identity step by step.