Optimizing for Google AI

The one-sentence version

Google runs two AI surfaces — AI Overviews (inline in Search, appearing on ~25.8% of US queries with static answers) and AI Mode (a separate conversational tab at google.com/aimode, powered by Gemini 3.5 Flash and running at 1B+ monthly users) — and both are trained to reflect Google’s ranking signals, which means strong organic SEO is the prerequisite that ChatGPT and Perplexity don’t require (NetConnect Digital, SEMrush, Google Search blog).

AI Overviews vs. AI Mode — different products

Feature	AI Overviews	AI Mode
Access	Inline in regular Search results	Separate tab at `google.com/aimode`, or “Show More” from mobile AI Overview
Interaction	Static answer with links	Conversational, multi-turn, retains context
Coverage	~25.8% of US searches, ~50% of informational, 47%+ of commercial	Optional destination; 1B+ MAU
Input types	Text only	Text, voice, images, files, Chrome tabs
Model	Custom Gemini integrated with Search	Gemini 3.5 Flash (default globally) with agentic RL
Update visibility	May 2026 rolled out inline citations	Follow-up queries reuse initial retrieval set

The May 2026 AI Overviews update

On May 6, 2026, Google’s VP of Product Management for Search Hema Budaraju rolled out five structural changes (NetConnect Digital):

Inline source links — citations now appear next to specific sentences they support, closing part of the CTR-collapse problem.
Expanded coverage — 50% of informational queries, 47%+ of commercial.
Deeper reasoning via Gemini upgrades — multi-step synthesis, context retention.
Multimodal input — images, files, videos, Chrome tabs as retrieval inputs.
Search agents — long-running tasks that hold context across multiple retrieval-and-answer loops.

The traffic-impact reality behind the update:

Publisher size	Referral traffic decline
Large publishers	22%
Medium publishers	47%
Small publishers	60%

The May 2026 Google I/O AI Mode announcement

From Elizabeth Reid, VP of Search, at I/O 2026 (blog.google):

AI Mode passed one billion monthly users one year after debut.
Queries have more than doubled every quarter since launch.
Gemini 3.5 Flash is the new default AI Mode model globally.
Search box redesigned with multimodal input — “the biggest upgrade in over 25 years.”
Personal Intelligence expanding to nearly 200 countries and territories, 98 languages.

What actually moves the needle

Get the Google fundamentals right (both surfaces)

Google’s AI surfaces are trained to reflect Google’s ranking signals. That means the traditional-SEO fundamentals matter here more than on any other AI engine:

Core Web Vitals under Google’s thresholds (LCP < 2.5s, CLS < 0.1, INP < 200 ms).
62% of AI Overview citations reference content updated within 90 days (xSeek). Freshness is a first-order signal.
Complete schema: Organization, Article, FAQPage, HowTo. Product / Offer / Reservation for commercial pages.
Clean canonical URLs, semantic HTML, coherent internal linking.

Structure content for sub-query citations

Both AI Overviews and AI Mode decompose complex questions into sub-queries and cite whichever page best answers each sub-query. One idea per page, one direct answer per idea. Sprawling pillar pages are for humans, not for AI Mode.

FAQ blocks with 4–6 pairs, 40–80 words per answer.
H2/H3 sub-questions with self-contained answers underneath.
Statistics with inline attribution (+37% citation probability lift).
Own content cites authoritative sources (+40% lift).

Own the multimodal surface (AI Mode specifically)

Since AI Mode accepts images and files, image and chart metadata are first-class retrieval signals:

Every chart caption restates the key numbers in visible body text.
Alt text describes image content, not just its subject.
PDF titles, headings and metadata treated like page-level SEO.

Win first-turn category citations (AI Mode specifically)

Because AI Mode retains context across follow-ups, being cited in the initial answer elevates your chance of being cited in follow-ups on the same topic. Focus content investment on top-of-funnel category-definition pages — “What is X?”, “How does X work?” — because they answer sub-questions rather than triggering more of them.

Retrofit checklist from Fuelonline’s 2,400-query analysis

Based on Fuelonline’s May 2026 AI Overviews analysis:

Direct Answer Block (60–100 words) in first 100 words.
3+ referring domains to the specific URL — 3× citation-probability increase.
FAQPage JSON-LD, 4–6 pairs, 40–80 words each.
Content freshness under 12 months.
1,200–3,000 word optimum length.
Mobile page load under 2 seconds.

Sites implementing the full retrofit have been observed earning citations within 14–21 days of Google’s next crawl cycle.

The Google AI checklist

[ ] Core Web Vitals green on priority URLs.
[ ] All priority pages indexed in Google Search Console; no coverage errors.
[ ] Content last-updated within 12 months on every priority URL.
[ ] Direct-answer paragraph (60–100 words) at the top of every page.
[ ] FAQPage JSON-LD with 4–6 pairs, 40–80 words per answer.
[ ] Article / HowTo / Product / Organization schema where applicable.
[ ] Named-expert bylines with Person schema.
[ ] 3+ referring external domains to priority URLs.
[ ] Chart captions restate key numbers in body text.
[ ] Alt text describes image content, not just subject.
[ ] Category-definition pages exist for every major concept in your domain.
[ ] GA4 custom channel group separates AI Mode traffic (referrer google.com/aimode).

Deep dives

Structured data that actually moves Google AI Overviews citations — the three schema stacks (commercial, informational, local) with real JSON-LD markup, and the controlled test where only complete-schema pages appeared in AI Overviews.
Google’s May 2026 AI Overviews update — five structural changes and the retrofit playbook.
Optimizing for Google AI Mode — the 1B-user conversational surface with query decomposition and multimodal input.
The AI crawler decision framework — why Google-Extended doesn’t block AI Overviews, and the full four-currency ledger for every robots.txt rule.
Digital PR for GEO — the 25% earned-media citation surface across every AI engine, including AI Overviews.
Content freshness for AI search — the 13-week rule and the freshness signal hierarchy that AI Overviews inherits from Google Search.
Answer-first writing — the passage-level editorial mechanics.
Schema for AI search — the JSON-LD patterns Google’s structured-data-first surfaces reward most.
Entity-first SEO — Wikidata + schema.org sameAs for entity consolidation.
Measuring GEO in 2026 — the four-metric framework and how to bucket AI Mode traffic in GA4.
Reverse engineering competitor AI citations — the eight-step manual audit that diagnoses why competitors appear in AI Overviews before you commit to paid tracking.

How Google AI fits the wider GEO picture

Google’s AI surfaces are the least disruptive to well-run SEO programs. If your organic rankings are strong, your fundamentals in place, and your content genuinely useful, you’re already most of the way to AI Overview and AI Mode citations. The specific work above — sub-query focus, multimodal metadata, first-turn category pages — is the delta between “Google AI friendly” and “Google AI optimized.”

But that same property caps the upside. Unlike Perplexity, where a rank-12 page can still be cited if its passages are best-in-class, Google AI inherits Google’s ranking prior. If you don’t already compete organically, neither AI Overviews nor AI Mode is a shortcut.

Start with the GEO Guide for the universal foundations, cross-reference with Perplexity and ChatGPT, and treat Google AI as the surface where traditional-SEO investment pays out most directly.