ChatGPT Feeds Google. Google Rewards Real Assets. Here's the Play. | Soaring Above Search #2

Everyone keeps framing AI search and Google as a battle. One eats the other. Traffic collapses. The SEO industry dies. It's a clean narrative and it's wrong.

A Semrush study just dropped showing that 21.6% of all ChatGPT referral traffic goes directly to Google. That's not a rounding error. ChatGPT is Google's single largest referral source. The platforms aren't competing for the same moment. They're different stops on the same journey. Someone asks ChatGPT a question, gets a direction, then goes to Google to go deeper.

But here's where it gets interesting. The Zyppy data published this week shows that what actually wins in Google isn't better content, it's proprietary assets and task completion. 92.9% of ranking winners have something nobody can copy. Generic SEO content is being squeezed from both ends: AI pulls the research traffic, and Google deprioritises sites with nothing real behind the words.

The consideration era doesn't replace search. It precedes it. Build something real, structure it so machines can read it, and you show up everywhere.

What Moved This Week

• ChatGPT's biggest referral destination is Google: The Semrush study tracking 170,000 domains found that 21.6% of ChatGPT referrals go straight to Google, with 206% YoY growth in ChatGPT referral traffic overall. AI search is expanding the pie, not stealing slices. The implication for brands is clear: you need to show up in AI search to be discovered, and in Google to be chosen.
• Google's ranking winners have products, not just content: Zyppy analysed what separates ranking winners from losers and found product/service presence (0.391 correlation) and task completion (0.381) are the strongest signals, not content quality. 83.7% of winning sites let users complete the full task on-site. Sites that exist to rank rather than to serve are being filtered out. This has been coming for years. The data just confirmed it.
• Structure drives AI citation share, not just authority: Hostinger ran a three-month experiment optimising 100 pages for machine readability, using answer-first introductions, question-based headings, schema markup, and an llms.txt file. Citation share jumped from 3.1% to 4.7%, a 52% increase. The lesson: authority gets you on the shortlist. Structure gets you cited.

Claude Skill of the Week: AI Citability Audit Prompt

What it does: Audits any page against the five structural signals that drive AI citation share, based on Hostinger's published methodology.

The prompt:

You are an AI citability auditor. Review the following page content and score it across five dimensions (1-5 each):

1. Answer-first structure: Does the introduction directly answer the primary question without preamble?
2. Question-based headings: Are H2/H3 tags phrased as questions or direct informational labels?
3. Schema markup signals: Are there FAQ, HowTo, or Author schema patterns present in the content?
4. Data tables: Are claims backed by HTML tables with quantifiable comparisons?
5. Topical completeness: Does the page cover the full question space without deferring to other pages?

For each dimension, give a score, one sentence of diagnosis, and one specific fix.

Page content:
[paste your page content here]

What I found: Running this against three of my own pages surfaced the same pattern: answer-first scores were low despite strong topical coverage. The headings were descriptive, not interrogative. Small structural shifts, big citability difference.

Watch This Week

Harry Sanders walks through what he'd actually do if he were learning SEO from scratch in 2026. The Consideration Era framing, earning topics instead of chasing keywords, and why doing real SEO on real sites beats any course. After a week of data confirming proprietary value wins, this is the clearest breakdown of what that actually looks like to build.

From the AI Search Tracker

Akamai State of the Internet Report, April 8, 2026: AI bot traffic surged 300% in 2025. AI chatbots now deliver approximately 96% less referral traffic than traditional Google search, with users clicking cited sources roughly 1% of the time.

The Semrush study showing ChatGPT referral growth and the Akamai bot data aren't contradictory. Referral clicks from AI search are growing fast from a tiny base, while bot-driven content scraping is accelerating even faster. The traffic gap is real and worth watching. Separate your AI referral traffic from organic in Analytics so you can actually see what's moving.

This Week's Takeaway

ChatGPT and Google are not competing for the same moment. AI search is where decisions start. Google is where they get confirmed. The brands winning both have something real, structured for machines to read, and deep enough to complete the full task on-site. That's the same play. It always was.

Frequently Asked Questions

Why is ChatGPT sending traffic to Google instead of away from it?

Because AI search and traditional search serve different moments in the same journey. ChatGPT is most useful for exploration, synthesis, and framing a problem. Google is where people go once they know what they're looking for. The Semrush data showing 21.6% of ChatGPT referrals flowing to Google reflects this: AI answers the "what is this" question, then people use Google to find the specific site, product, or review that closes the decision.

What counts as a proprietary asset for SEO purposes?

Anything on your site that can't be replicated by someone writing content about your topic. Zyppy's analysis includes: original data and research, user-generated content platforms (reviews, forums, Q&A), tools that complete a task on-site (calculators, comparison engines, booking systems), exclusive product or service inventory, and genuine expertise tied to named individuals with verifiable credentials. If a content farm could publish the same page tomorrow, it's not proprietary.

How do I make my pages more citable by AI search engines?

Structure is the lever. Hostinger's experiment showed that answer-first introductions (where the first sentence directly answers the primary question), question-format headings, FAQ schema, and HTML tables for data all drive measurable citation share increases. An llms.txt file helps AI crawlers understand your site architecture. Authority matters too, but a well-structured page from a mid-authority domain often outperforms an unstructured page from a high-authority one in AI citation.

Want more? Follow the full AI Search Tracker or explore the AI SEO hub.