What actually died
It is overstated to say SEO is dead. It is precise to say keyword-density SEO is dead. The specific tactics that no longer work are: targeting a single 'primary keyword' per URL, optimizing for 1–3% body density, building exact-match anchor text portfolios, and writing for an algorithm that no longer parses content this way.
Why now? Three concurrent shifts. First, Google's transformer-based ranking systems (BERT in 2019, MUM in 2021, and successive updates) parse semantically rather than lexically. Second, AI Overviews and LLM-based assistants ingest content in chunks and care about answer quality, not keyword presence. Third, the rise of Information Gain as a differentiating signal (see /insights/information-gain-scoring) makes 'I targeted this keyword harder' a losing strategy when your competitor said something genuinely new.
Replacement 1: Entity coverage
An entity is a named thing — a person, place, brand, regulation, technique. Entity coverage is the practice of writing pages that treat a topic as a connected graph of named things. A page about 'SEO for law firms' covers the LSO, the SEO industry, named directories (Lawyers.com, FindLaw, Justia), specific schema types (Attorney, LegalService), regulators in each jurisdiction, and the specific practice areas that change SEO meaningfully.
The operational test: can a reader who knew nothing about your topic before reading the page now name 8–10 specific entities related to it? If yes, your entity coverage is strong. If your page is a string of generic claims with no named entities, it will lose to one that has them, regardless of keyword density.
Replacement 2: Query baskets
Single-keyword tracking is brittle. The query that drives revenue today may not exist in three months. Query basket modelling tracks rank across a cluster of 10–50 related queries representing the user's full information need. A page is winning if it is in the top 5 for the basket on average, not necessarily #1 for any single query.
This also matches how AI Overviews trigger. They appear for clusters of related questions, not single keywords. A page that wins the basket is far more likely to be cited in the AI Overview than a page that wins a single-query battle but fails the cluster.
Replacement 3: Answer-shaped content
Answer-shaped content is content where the answer is the structure. A page that opens with the literal user question as H1, follows immediately with a 4-bullet TL;DR that answers it, then expands with proof, sub-questions, and counter-positions, is answer-shaped. A page that opens with brand fluff and buries the answer in section 4 is not.
This is also where E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) operationalises. Experience shows up as 'we tested' markers. Expertise shows up as named entities used correctly. Authoritativeness shows up as inbound citation patterns. Trust shows up as version stamps, author bylines, and verifiable sources cited inline.
What to stop doing this week
- •Stop using 'target keyword' as a single field in your content brief. Replace with 'query basket' (a list).
- •Stop optimizing for body keyword density. Optimize for entity coverage instead.
- •Stop building exact-match anchor portfolios. Vary anchors naturally; trust your topical authority.
- •Stop writing the answer in section 3. Move it to the TL;DR immediately under H1.
- •Stop assuming a page can win 'a keyword.' Pages win clusters, not single queries.