How I'd learn AI SEO if I starting today (2026)

Nathan Gotch's starting point is to build something and learn how to grow it. In this tutorial, he proposes a small software or service business as a way to practice AI SEO, rather than beginning with an exhaustive course of study.

The mechanism is just-in-time learning. A real obstacle creates a question; the learner finds an explanation, implements a response and sees what happens. Books and videos support that loop, but consuming them does not replace the implementation.

The project also combines two skills: building with an AI agent and acquiring customers. Gotch argues that working through both reveals connections among the website, checkout, delivery, analytics and marketing that isolated exercises can miss. He ends the video by emphasizing that action creates feedback, not a guarantee of success.

Gotch recommends selecting one category, one ideal customer profile and one specific problem. The goal is a small project that can produce a clear outcome, not a miniature version of a large company with a growing list of features.

His own early baseball blog illustrates a distraction: repeatedly changing the website felt controllable, but did not bring customers. Distribution required harder work on awareness, traffic and the reasons someone would buy.

That changes how he approaches content. A topic should connect to the thing being sold and to a stage in the buying process. The same topic can appear as a website article, a YouTube video or a shorter social post, but the format must fit the channel. His suggestion to reserve some content for experiments is a planning heuristic, not a performance formula.

For an opportunity, Gotch suggests looking at a broad software product and identifying one function that could be handled more deeply. His example is prompt research for AI search: a focused deliverable rather than an entire SEO platform.

An established competitor shows that a problem has a market context. It does not, by itself, validate demand for a new entrant. His proposed next step is therefore practical: create a simple page and sell the outcome as a service before investing in a polished interface.

The operator can use unfinished internal software to perform the work. Each delivery then exposes defects and produces customer feedback. If repeated demand develops, the same system may become a customer-facing product. The order is important: do the job for someone before assuming how they want to do it themselves.

Gotch proposes a low-priced initial service and collecting reviews after delivery. He also describes a rebate in exchange for an honest review, explicitly saying not to require a positive one.

That incentive is part of his proposal and should remain visible when interpreting the resulting social proof. Asking for honesty does not make an incentivized review independent, nor does the video establish that the arrangement is permitted by every platform or jurisdiction. The article does not turn his compliance assertion into a verified rule.

Customer feedback has another use regardless of whether a review is published: it shows where delivery failed or the software needs work. Gotch's service-first loop depends on that operational learning, not just accumulating favorable-looking evidence.

Gotch does not reject foundational study. He recommends learning terminology so that a learner can recognize a problem, investigate it and explain it to an AI tool.

He discusses AI SEO for Dummies, The Art of SEO and Product-Led SEO as complementary references, while acknowledging his bias toward the first book. These are the video's named references, rather than a freshly researched buying recommendation. His suggested pattern is to return to a source when a project exposes a gap, then implement what was learned.

Knowing a term is different from using the idea. The tutorial repeatedly asks the learner to connect concepts to an observable result instead of measuring progress by videos watched or pages completed.

In the on-screen walkthrough, Gotch creates a prompt-research project and asks the agent to establish context. The intended product generates prompts for tracking customer questions in AI search, grounded in actual data; that is distinct from merely generating instructions for an AI to conduct research.

He specifies the narrow problem, the evidence requirement and the service-first approach. There is no need for a polished customer interface at that stage. A maintained context file gives a later session something concrete to read rather than relying on an assumption that it remembers the project.

He also separates development work from AI SEO planning and discusses starting a fresh conversation when context becomes unwieldy. The reusable principle is to preserve a clear brief and decisions. Particular interfaces and memory behavior in the demonstration belong to the tool version he was using.

Gotch demonstrates plugins and MCP connections as ways to supply evidence. He promotes Rankability, his own software, and names other data services as lower-budget examples. Those commercial relationships matter when assessing the comparison; the video does not independently prove his product is superior.

He also discusses search and analytics accounts, including Google Search Console, Bing Webmaster Tools and Google Analytics. Their role in the workflow is to observe the site and its audience, rather than ask a model to invent a picture of performance.

Documentation can help an agent learn how a tool is intended to work. Gotch suggests building reusable instructions from help-center material. That can improve execution context, but an instruction file is not proof that every action or conclusion will be correct. The actual output still needs examination against the underlying evidence.

Gotch describes a campaign around his AI SEO book and shows a visibility report in which the book appears in answers after a short period of work. He uses it to illustrate category focus: choose an area where improved visibility would matter to the business over the next 90 days.

The screen is a demonstration of his experiment. It does not isolate which activity caused a mention, establish the stability of results or guarantee the same outcome for another site. He says the campaign involved more than one blog post and does not disclose every detail in this tutorial.

He also shows work distributed through Quora and LinkedIn, including agent-assisted responses and a list promoting his book. Those are brand-directed marketing activities. A self-published recommendation should not be confused with an independent consensus merely because it is visible on another domain.

The research sequence moves from a category to topics, entities and keywords, then to representative prompts. Gotch suggests beginning with manual runs rather than assuming a paid visibility tracker is necessary immediately.

Collecting the answer is only the first step. The learner can identify mentioned entities, inspect citations and check whether the brand appears on the cited pages. Competitor research then extends beyond the competitor's own site: reviews, distribution, links and other evidence help explain where visibility may be coming from.

For useful comparisons, the prompt and observation conditions need to remain explicit. A single answer is a sample, not a stable ranking. Gotch's emphasis is to prioritize work likely to affect visibility and credibility, rather than spend the entire project on small edits that are easy to finish but disconnected from the main goal.

39:25View full image

His example plan contains three data-driven studies, two free tools and five commercial assets. The studies could compare the outputs of different AI systems on the same prompt, creating first-party observations relevant to the service.

Free tools offer a useful deliverable and another building exercise. Commercial assets address comparison, alternatives and other middle- or lower-funnel questions. A brand's own list or comparison has an interested author, so its framing and evidence should be visible to the reader.

The topics can then be adapted to other channels and connected back to relevant pages, for example by embedding a related video. The ten-asset split is a manageable starting plan from the video, not a universal requirement or proof that a site will be cited.

Gotch adds accurate profiles, customer feedback and outreach to relevant retrieval sources. His sample outreach introduces the specific outcome and asks whether a resource accepts submissions. The broader idea is to investigate where relevant audiences and evidence already gather, rather than wait for publishing alone to create discovery.

He recommends observing visibility repeatedly, manually at first if necessary. Mentions, citations, referral behavior and actual customer outcomes answer different questions; activity counts alone do not show that a business has improved.

The learning loop returns to its beginning: build one useful thing, put it in front of the intended audience, examine the response and change the next action. The value of working through a problem is the new evidence and skill it produces, even when the first commercial attempt does not succeed.