AI Recommendation Stability

Why this snapshot

In June we measured how AI engines recommend brands. This snapshot asks the follow-up question: when a buyer types the same question into Google, which pages do they see first, and are those the same sources AI engines draw on?

How it was collected

Google organic search, US English, desktop, collected July 13, 2026. Los Angeles and New York City location settings for the two local legal questions, United States for the rest. The first five organic results for each of the 26 questions that produced usable AI results in the June study: 15 regulated, 6 consumer and 5 B2B. The set is not balanced by segment. No new AI calls were made; the comparison uses the sources AI engines cited in June.

Finding 1: Google usually shows a gatekeeper before the provider

Only 5 of 26 first organic results were the provider's own page: Salesforce, ProProfs, Fidelity, Walgreens and Wildes Law. The other 21 were publishers, rankings, communities, reference sites or government sources. For recommendation questions, being found is mostly about being included in the pages Google already trusts.

Finding 2: An AI layer already sits above the organic results

Google showed an AI Overview for 24 of 26 questions. In each of those, the first organic result was the second thing on the page. For this kind of question, search results and AI answers are no longer separate channels.

Finding 3: Google and AI answers share much of the same source layer

68 of the 130 Google results came from domains that AI engines also cited somewhere in the June study, across 24 of the 26 questions. The first organic domain was among them for 19 of 26. This is shared sources, matched across the whole June dataset rather than question by question. It does not show that a Google ranking causes an AI recommendation. Recurring names include Vault, Best Law Firms, Forbes, NerdWallet, Healthline, Newsweek, Fidelity, the SBA, Bon Appetit and Runner's World.

Finding 4: Reddit is a major discovery surface

Reddit appeared in the top five for 14 of 26 questions: all 5 B2B questions, 5 of 6 consumer questions and 4 of 15 regulated ones. It ranked first for the VPN question. Peer discussion is part of how buyers compare options in B2B and consumer categories.

Finding 5: Many sources, a few recurring gatekeepers

The 130 results came from 103 different domains. A handful recurred: Reddit (14 appearances), Forbes (4), Vault, Wikipedia and YouTube (3 each), and Amazon, GoodRx, PCMag, the Wall Street Journal and LinkedIn (2 each). The work is twofold: earn a presence on the recurring platforms, then find the specialist gatekeeper for your own category.

What marketers should do next

• Find the pages that rank for your buyer questions, and check which brands those pages mention.
• Prioritize sources that show up in both Google and AI answers.
• Separate work on your own site from earned coverage; most first results were not provider pages.
• Treat Reddit as a real channel in B2B and consumer categories.
• Track AI Overviews alongside organic position.

Question-level results

Limits

A point-in-time snapshot, not a stability study. Google results change with location, device, personalization, index updates and news. The source overlap shows shared domains, matched across the whole June AI dataset; it does not prove that a ranking caused an AI recommendation. Only the top five organic results were kept. The segments are not balanced.