SEO Recap covering August 6, 2026: Google ranking volatility and an AI leadership shakeup
Daily summary of what matters in SEO, GEO, AEO, and AI search generated with Claude Code (beware of hallucinations)
Generative Engine Optimization (GEO) & Answer Engine Optimization (AEO)
How I'd Build an AI Search Strategy Today
Source: Search Engine Journal
- Semrush found that ~90% of the pages ChatGPT cites rank at position 21 or lower for related queries, so the pages assistants trust are mostly not the ones winning page one.
- Six signals explain most AI visibility results:
- Brand authority from third-party mentions, not your own blog
- Content freshness with visible, dated updates
- Entity recognition of your company and its named experts
- Extraction-ready structure with direct answers and schema
- Cross-platform consensus across Reddit, YouTube, reviews, and press
- Earned media you did not pay for or write yourself
- Run a 20-prompt drill: write 5 buyer prompts and test them across ChatGPT, Perplexity, Gemini, and Google AI Mode, logging whether you are named, cited, or absent. Under 12 of 20 signals a citation problem.
- A 90-day sprint moves from audit to scoped channel tests to systematizing what works. One B2B client went from 4 of 20 to 13 of 20 drill runs in a quarter, with the biggest jump after Reddit answers went live.
- Track citation share over time with tools like Profound, Otterly.AI, and Peec AI, and fund experiments by cutting a budget line surviving on habit.
Why AI Authority Takes Years and Comes From Other People
Source: Search Engine Journal
- Parametric standing, what a model says about you before it retrieves anything, comes from many independent parties describing your company in varied phrasing, not from publishing more of your own content.
- Kandpal and colleagues showed at ICML that a model's accuracy on a fact tracks the number of relevant pretraining documents, and that models would need scaling by many orders of magnitude to answer well about thinly covered subjects.
- Allen-Zhu and Li found knowledge only becomes reliably extractable when it appears in sufficiently varied phrasing during pretraining. Repetition from one source can leave a fact present but returning zero accuracy.
- Elazar and colleagues' review of 10 training corpora found C4 so diverse that its single most common domain is under 0.05% of documents, so any one owned property is a vanishingly small share.
- Cheng and colleagues at Johns Hopkins found a model's knowledge often concentrates earlier than its published cutoff, meaning much of a company's current standing was deposited by PR, reviews, and press work done years before AI optimization existed.
AI in Search / AI Overviews
Google Search Console Shows AI Mode Queries in the Main Performance Report
Source: Search Engine Roundtable
- John Mueller confirmed Search Console includes AI Overviews and AI Mode data in the general performance report, not the separate generative AI reports, which carry no query data.
- Follow-up questions in AI Mode count as new queries, so short conversational strings like "yes", "yes go on", and "yes, pricing" surface in the query data.
- The SEO community has long inferred AI-driven queries by spotting unusually long, conversational phrasing in performance reports.
Organic Search & Algorithm Updates
Google Search Ranking Volatility Heats Up Into August 5th and 6th
Source: Search Engine Roundtable
- Chatter across WebmasterWorld, Black Hat World, and Search Engine Roundtable spiked over 24 hours, with reports of big swings and reversals from an earlier unconfirmed update. Google has not confirmed it.
- Site owners reported swings including traffic down more than 50%, rankings improving while traffic fell 14%, and drops in Google Discover and Google News.
- Multiple third-party trackers (AccuRanker, Mozcast, SEMrush, Sistrix, Algoroo and others) spiked, matching the community reports.
- Some operators said low-authority sites were wiped out in favor of high-authority sites, with Google showing 3 to 4 pages per high-authority site plus extra YouTube presence in the SERPs.
Google AI Leadership Changes: Jeff Dean Leaves, Hassabis Shifts to Chair
Source: Search Engine Roundtable
- Google announced a shakeup in its AI leadership, and Alphabet stock fell ~4% on the news.
- The changes include:
- Jeff Dean, employee number 30 and a 27-year veteran instrumental in early Google Search, leaving to start Discovery Loop
- Demis Hassabis moving from Google DeepMind CEO to chairman and Alphabet Chief Scientist
- Koray Kavukcuoglu becoming senior VP of DeepMind, reporting to Sundar Pichai
- Sanjay Ghemawat, Oriol Vinyals, and Quoc Le departing with Dean
- Some observers questioned whether the moves relate to delays in Gemini releases, though no link is confirmed.
Published on