The old SEO playbook is dying in slow motion. Google’s AI Overviews answer 40% of queries without a click. ChatGPT search and Perplexity siphon away the rest. “Rank #1” doesn’t mean what it meant in 2022. Here’s what SEO actually is in 2026 — and what to do about it.

What changed

Three quiet shifts:

  1. AI Overviews ate the top of the page. Even when you rank, your click-through dropped 30–60% on informational queries.
  2. LLMs became search engines. ChatGPT, Claude, Perplexity, Gemini all surface answers from web content. They cite, sometimes; they often don’t.
  3. The user model changed. Younger users start in an LLM and only fall back to Google for transactional queries.

What “SEO” means in 2026

It splits into two disciplines:

  • Classic SEO — still real, still valuable, still rewards quality content with structured data, fast pages, and earned links. Google isn’t dying. The traffic is.
  • LLM-O / GEO — getting cited by AI search engines. Different optimization surface, different signals, different metric set.

The two overlap, but optimising only for one leaves money on the table.

What LLMs actually look for

From watching my own content get cited (and not cited):

  • Structured, scan-friendly content — H2/H3 hierarchies, lists, definitions up front
  • Specific numerical claims with sources — LLMs love quotable, verifiable facts
  • Strong entity signals — clear schema, consistent naming, presence on authoritative profiles (Wikidata, Wikipedia, IMDb, Crunchbase)
  • Recency — most LLM search products weight fresher content heavily
  • Citation-graph density — being cited by other authoritative pages compounds inside LLM training and retrieval

The llms.txt convention

The new /llms.txt file at your domain root is a structured site summary written for LLMs. It’s the robots.txt of the AI era. Mine is here — copy the format, adapt it for your site, ship it this week.

The practical 2026 SEO checklist

  1. Schema everywhere — Person, Organization, Article, Product, FAQPage
  2. llms.txt published
  3. Content with named entities, specific numbers, citable claims
  4. Reverse-engineer LLM citations: search your topic in Perplexity, see who gets cited, study why
  5. Don’t game it. The age of keyword stuffing is over twice over.

The honest part

The brands winning in 2026 are the ones treating LLMs as the new front page of the internet. The brands losing are still optimising for Google rankings nobody clicks. Both still matter; the relative weights have flipped.

Want a 30-minute LLM-O audit on your site? Email me.