SEO in 2026: How I'd Rank in Google in the AI Era

TL;DR

When an AI Overview appears, the #1 organic page now loses ~58% of its clicks (up from 35% in April). But last year's turmoil revealed three shifts that tell you what to do: (1) discovery is multi-platform — AI assistants research the same places people do (YouTube, Reddit, forums, reviews), and 86% of top cited sources differ per assistant; (2) brand mentions win via query fanout — AI explodes one query into dozens of subqueries and stitches an answer, so being mentioned on the pages that rank for those subqueries matters more than backlinks; (3) organic clicks flow to action — AI can answer but can't do, so action/transactional queries still belong to traditional SEO. Split your effort across all three.

Key takeaways

The click loss is real: AI Overview present → the #1 page loses ~58% of clicks (Ahrefs re-ran their April study).
Discovery is everywhere: people (and AI) research across Google → Reddit → YouTube → Amazon. 86% of top cited sources are unique to each AI assistant — Google AI Overviews lean YouTube/Reddit/Quora, ChatGPT prefers publishers/news, Perplexity cites niche/regional.
Reverse-engineer citations: Ahrefs Brand Radar → cited domains report shows exactly where AI pulls from in your niche.
Query fanout is the mechanism: a query fans into subqueries → AI pulls from pages ranking for each → stitches one answer. Branded mentions have the strongest correlation with AI-Overview visibility — more than backlinks, referring domains, or domain rating.
Get on the right pages: Brand Radar → cited pages, filter for best / top / versus / review / alternative → the roundups/comparisons/reviews worth getting mentioned in. Then pitch authors (send samples).
Action queries still win the click: if AI can't fully satisfy the query (backlink checker, snow-removal service, mortgage calculator), traditional SEO (content, links, technical) still works. Informational queries (what is a mortgage) go to AI Overviews.
The one test before targeting any query: "Can AI fully satisfy the user here?" If no + it has business value → target it the traditional way.

The modern buying journey: Google → Reddit → YouTube → Amazon reviews

Shift 1 · Discovery is multi-platform

People don't research on one site, and neither do AI assistants — they read YouTube, Reddit, forums, and review sites to decide who to recommend. Since the cited sources differ by assistant, don't guess — reverse-engineer what AI cites in your niche:

Ahrefs Brand Radar — the "cited domains" report shows where AI pulls from

Shift 2 · Brand mentions win, via query fanout

AI doesn't answer from memory — it fans one query into dozens of subqueries, pulls from the pages ranking for each, and stitches them together. So the more the pages behind those subqueries mention your brand, the more likely you're in the final answer.

Query fanout: one query explodes into many subqueries, each pulling from ranking pages

flowchart TD
    Q["'best smartphone for photography'"] --> S1["iPhone 17 Pro vs S25 Ultra camera"]
    Q --> S2["best phone for low-light photos"]
    Q --> S3["smartphone with best zoom 2026"]
    S1 & S2 & S3 --> P["pull from pages ranking for each"]
    P --> A["stitched AI answer<br/>(brands mentioned across those pages win)"]

Move: Brand Radar → cited pages, filter query for best/top/versus/review/alternative, then get your product into those roundups and comparisons.

Shift 3 · Organic clicks flow to action

AI is good at answering and summarizing, but it can't do things yet. So action/transactional queries still get the click — and no AI Overview appears for them, because Google knows the user needs to go somewhere.

flowchart LR
    Q["A query"] --> T{"Can AI fully<br/>satisfy the user?"}
    T -->|"Yes — informational<br/>(what is a mortgage?)"| AIO["AI Overview → zero-click"]
    T -->|"No — action/transactional<br/>(mortgage calculator, snow removal)"| SEO["✅ Traditional SEO still wins the click"]

Action queries (e.g. "snow removal service") still need a click — AI can't do the job

Move: find these in bulk in Keywords Explorer — matching-terms + calculator / checker / generator modifiers (free-tool intent), or just filter for transactional intent.

Keyword research for action-oriented, click-worthy queries

The reframe

SEO didn't end — it split into three buckets: be discoverable everywhere, build a brand that gets mentioned on the lists/comparisons/reviews AI cites, and keep winning the action queries with traditional fundamentals. The old version isn't dead; it's now one-third of the job.