GOOGLE · AI OVERVIEWS Aug 17, 2026 · 9 min read

Google quietly rewrote how AI Overviews pick citations, and page-one rankings barely matter anymore

MK Mara Kovač Editor-in-Chief · 6:00 AM ET 𝕏 in ✉

Editorial illustration: a stack of ranked bars dissolving while a single highlighted passage rises above it

For fifteen years, the deal was simple: rank on page one, get seen. That deal is over. Between May and August we analyzed 2.1 million queries that triggered AI Overviews, matching every citation back to its classic organic position. In May, 43% of citations came from page-one results. By August, that number was 18%.

Where AI Overview citations come from MAY 2026 AUG 2026 Page 1 (1-10) 43% 18% Pages 2-3 (11-30) 28% 31% Beyond page 3 17% 34% Not in top 100 12% 17% Share of AI Overview citations by classic organic position. Source: AEO Daily log analysis, n = 2.1M queries.

The shift isn't random. Google appears to have decoupled citation selection from document-level ranking almost entirely, moving to passage-level retrieval with its own quality gates. Pages ranking 40th are getting cited above pages ranking 3rd, when the 40th page contains a tight, self-contained answer passage.

How the four engines compare

Google isn't alone in this, but it's the most aggressive. Here's how citation behavior differs across the four engines we track daily:

ENGINE PAGE-ONE SHARE AVG CITATIONS FRESHNESS BIAS

GGoogle AI Overviews 18% 7.2 Very high

CChatGPT Search 31% 4.8 High

PPerplexity 38% 5.5 High

AClaude 27% 3.9 Moderate

90-day rolling averages across our 4,000-prompt tracking panel. Freshness bias = citation-rate premium for content updated in the last 30 days.

Old playbook vs. new playbook

✕THE OLD PLAYBOOK ·Chase page-one rankings with domain-level authority plays ·Refresh the publish date, leave the content untouched ·2,000-word pages built around one head keyword ·Measure success in blue-link clicks

WHAT WINS NOW ✓THE NEW PLAYBOOK →Build liftable answer passages: claim, evidence, caveat, all under 90 words →Date-stamp and genuinely update sections, not pages →One page per entity, structured for passage retrieval →Measure share of prompts where you're the cited answer

"Authority is a floor, not a ladder. Above the trust threshold, the passage does all the work."

What to change this week

Take your ten most valuable queries, find the exact paragraph that should answer each, and rebuild it as a liftable unit. Then declare your key pages to crawlers. A minimal llms.txt looks like this:

llms.txt COPY

# AEO Daily
> Daily news and analysis on answer engine optimization.

## Key pages
- [AEO audit checklist](/audit): 27 checks for AI search visibility
- [Citation study](/research/citations): 200-brand, 4-engine dataset

In our test set, pages restructured this way picked up their first AI Overview citations within three weeks. The window is open. It won't stay that way once the rest of the industry catches up.