The Loop That Ranks in Google AI: From Keyword Gap to Dedicated Page

The story opens with a site that climbed from zero to over 1,300 daily clicks and a screenshot claiming number one in Google AI for best GEO expert . The speaker says the result did not come from one viral post but from a loop that feeds itself every day. Generative Engine Optimization is presented not as ranking in classic blue links but as getting cited inside generative answers from ChatGPT, Perplexity, Gemini and Google AI Overviews. The same thesis lives on Juliangoldie.com's Best GEO Expert page: GEO means placing the brand inside AI answers. The appeal is that the system spreads across a portfolio, not just one domain.

Gap hunting: demand whispering in 24 hours

The first step is demand discovery. The speaker scans the last 24 hours of Google Search Console data and lists keywords that have started to earn clicks and impressions but have no dedicated page. Examples are concrete: Hermes 3D office shows clicks and impressions while the site only has a nearby Hermes Agent Office page, no exact match. Jev desktop has two clicks and three impressions, Jev open source shows similar signals. According to the Performance report documentation on Support.google.com, an impression counts how often a page appeared in results, a click counts actual visits, read together with CTR and average position. An impression is therefore evidence of demand you have not yet fully served.

Instead of manually pulling data site by site, the speaker pulls the whole portfolio through an Agent Operating System called Search OS. With a filter for the last seven days or last 24 hours, typing a stem like Jev surfaces all related new queries. That is faster than the manual tour of separate Search Console panels. The shortlist goes into a Google Doc, kept to three to five candidates at a time. With a broad portfolio the same skeleton works across verticals: SEO niche, AI niche, same gap logic.

Each candidate gets a second filter of on-site verification. A site: search checks whether a page already dedicated to that keyword exists. Hermes on Omachi, already covered, is skipped, while a comparison intent falling between two existing posts like DeepSeek harness versus Claude Code is deemed worth filling with a new page. Keeping a won query like Best AI SEO community is less valuable than answering a high-intent comparison with a clear table, which reduces cannibalization and matches intent more precisely.

The production layer then kicks in: a skill built on Claude Opus 5.5 . The speaker enters the target keyword and a case study into a Generate tab and creates a page in one click. According to Anthropic, Opus 5.5 was announced on September 22, 2026, runs about 40% cheaper than Opus 5 for typical workloads, offers a 1M context window for long-running agents, and is priced at $4 per million input tokens and $20 per million output tokens. That economics makes it feasible to distribute the same skeleton as distinct pages across five different sites. The skill runs either directly inside Claude or in the Agent OS interface with the same template.

Anatomy of the automated page and why it works

He walks through a resulting page hosted on Netlify. The structure is familiar but complete: the keyword in the title and first paragraph, immediately below a case study grounded in his own test, then conversion CTAs scattered through the copy and a pinned CTA at the bottom. He credits Kazer Dash for suggesting the pinned CTA that lifted conversion. In the middle a nicely formatted comparison table, near the bottom two internal links to related posts on the same topic, then a FAQ block, testimonials, and an author bio for authority. Internal links and FAQ give a chance to appear for additional queries via Google's FAQ rich result, and as SearchEngineJournal notes, the push of ads and modular blocks in AI Overviews makes that structured layer more valuable.

Why the page ranks is answered with content uniqueness. Each page rests on an experiment never published before, so it carries the uniqueness and experience signals Google rewards. As described in the Generative AI in Search launch post on Blog.google, AI Overviews became the default layer of Search since 2024 and earning a citation in the answer requires the page to present both intent and evidence clearly. The speaker provides that evidence with case data: tables, screen recordings, numbers, all verifiable on camera. That aligns with the core of GEO.

Repeatability is backed by the multi-site proof. The same find-gap, add-case, publish, measure loop delivers similar momentum across verticals. In addition to the documentation on Support.google.com, the portfolio view showing the same filters working for several sites strengthens the scale argument. Because new impressions are born every day in a volatile space, the loop never closes: a new page brings new impressions, those impressions breed new gaps.

Measurement is the fuel of the loop. After the new page is indexed, its impression and click curve for that query is tracked in Search Console, noting position and CTR shifts. The speaker feeds that feedback into the next 24-hour scan, seeing which template stuck, which title format alongside FAQ earned visibility on additional queries, and sharpens the template.

Of course the loop is not magic. Creating a dedicated page for every query brings the risk of thin content and template fatigue from automation. The speaker implies he manages that risk by running a site search for each candidate, filtering overlap with existing pages, and requiring a real case study for every page. Analyses on Blog.google and SearchEngineJournal whisper the same caution: durable citation in AI answers demands both original experience and clear structure, otherwise visibility is fleeting.