SEO Recap covering September 30, 2026: AI Overviews on branded queries and Google's publisher payouts

Built by Stephanie Chung·6 min read·7 stories

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)

Wikipedia tops AI citation studies: Muck Rack's analysis of 25 million AI-cited links makes it the most cited domain in ChatGPT, Ahrefs ranks it sixth in Google AI Mode, and Profound found ~1 in 6 cited ChatGPT conversations pull from it.
Muck Rack found 42% of Wikipedia citations in AI answers come from pages about processes, technologies, or methods, not brands or people, so prioritize a clean, well-sourced page for your category and methodology.
Reddit citations from Google AI Overviews reached 5.8 million via Ahrefs Brand Radar, ahead of ChatGPT (1.3 million), Gemini (477,000), and Copilot (202,000), and Reddit is the fourth most visible domain in U.S. Google Search per SISTRIX.
Treat a podcast booking as a multi-format asset: confirm the transcript and show notes, request a backlink, clip quotable moments for YouTube and LinkedIn, and add it to a crawlable press page, since YouTube ranks third in Contently's 2026 LLM citation analysis.
Run a Wikipedia footprint audit, fold your PR team's Reddit subreddit list into AI visibility tracking, and lean on comms colleagues who already understand Wikipedia's sourcing rules.

Brand presence rate: run 5 buyer-intent prompts across 4 AI assistants weekly and record whether your company is named, linked, or absent across the 20 answers.
Answer accuracy: check product description, buyer, price tier, and competitors in each mention. One Series A client had only 4 of 11 mentions accurate until a cleanup brought it to 10 of 11.
AI referral conversion: group GA4 sessions from ChatGPT, Perplexity, Gemini, Copilot, and Claude by session source. These converted at 3-5 times the rate of organic sessions.
Branded search and self-reported attribution: track Search Console branded impressions and clicks, and add an AI assistant option to "How did you hear about us?" One client saw 9 of 54 new opportunities pick AI, and those deals closed 30% faster.
Expect timing in stages: brand presence improves in 2-6 weeks, AI referral conversions in weeks 4-8, and branded search in weeks 8-12, with self-reported attribution lagging a sales cycle.

Chris Long found AI Overviews appearing for 93% of ~100 brands he tested, including Reddit, Salesforce, Amazon, and Adobe, with Google's own name and news publications as exceptions.
Most branded AI Overviews appear mid-page below the main company result rather than at the top, though Adobe's shows at the top for its name.
DemandSphere data says AI Overviews on branded queries tripled in September 2026.

The Information reports ~100 publishers in Google's AI contribution pilot are earning less than one-tenth of 1% of their advertising revenue from content used in AI Overviews, AI Mode, and Gemini.
Payouts vary widely. One publisher on a $1 million deal earned less than $1,000 over several months, another joined recently and earned $50,000-60,000, and one early participant makes more than $1 million a year.
Payments are said to depend on how much each source contributes to an AI answer, but participants say they do not understand how Google calculates the amounts, which fluctuate month to month.
Publishers can accept terms, track accrued earnings, and manage participation inside Search Console settings.

Dots can check connected sources for updates before being asked (proactive research) and run scheduled recurring checks, with the no-prompt mode limited to reading sources and saving private notes.
Proactive research cannot send messages, change content through plugins, or control a browser or computer, and Custom Rules cannot override those limits.
Dots are rolling out to Pro users (excluding the EEA, Switzerland, and UK) and Business Premium users, with a beta for Enterprise, Edu, and Healthcare, and the plugin catalog now exceeds 4,000 apps.
As of September 30 none of the Dots help pages reference an analytics or SEO platform, so the open question is which monitoring tools a dot can read through ChatGPT's plugins.

LiveRamp's RampID now lets advertisers activate CRM, loyalty, website, and app audiences in ChatGPT Ads to target or suppress customers, extending a partnership that began with measurement in June.
The integration is live in 11 markets, with expansion planned as ChatGPT Ads reaches more regions.
Advertisers can also upload Custom Audiences directly in OpenAI Ads Manager via CSV or TXT, include or exclude them at the campaign level, and apply bid multipliers from 0.1x to 10x at the ad group level.
There are no public benchmarks for how first-party audiences perform or price versus ChatGPT's other targeting, so suppression and small tests make more sense than a budget shift for now.

Chris Green built a technical SEO Chrome extension around Gemini Nano, Chrome's small on-device model, to handle light interpretation without API calls, credit cards, or sending data off-device.
Nano proved unreliable at final judgments like deciding whether a raw-versus-rendered HTML difference is a real problem. A stronger API model (ChatGPT or Gemini) handled the same structured evidence far better.
The resulting architecture uses 3 layers: code for exact tasks (fetching URLs, comparing HTML, checking responses), a local model for human-readable summaries, and a larger model only when genuine reasoning is needed.
- Code handles exact work like URL fetching, HTML comparison, and canonical checks
- Nano turns established evidence into readable passages
- A larger model is called only for ambiguous, complex judgment
Designing tools around a replaceable local model means future improvements in models, quantization, and hardware can arrive without a redesign, and forcing support for a small model also sharpened the deterministic code.

Published on October 1, 2026