If I Were Starting SEO Again in 2026

If I were starting SEO again in 2026, the first thing I would do is find a mentor.

Not someone to give me a list of ranking factors. I would look for someone who can explain SEO at a high level, break it into blocks, and show how those blocks connect.

A beginner cannot deliberately pursue a field they cannot yet see. If they do not know which blocks exist, every tutorial, tool, algorithm update, and tactic can look equally important. They may work very hard while still chasing SEO aimlessly.

The mentor's first job is to provide a map.

Start with a map of SEO

There is more than one useful way to divide the field.

At the simplest level, SEO can be explained through two large blocks:

• On-page SEO covers what happens on the websites we control. This includes technical SEO and content SEO: crawling, rendering, indexation, site structure, templates, content quality, intent, accuracy, and how information is organised.
• Off-page SEO covers what happens beyond the websites we control. This includes brand building, links, media coverage, and user-generated discussions that create references, attention, and evidence around a brand or subject.

This map is simple, but it gives a beginner somewhere to place each new idea.

I also use my CAKE framework to explain what good SEO is trying to achieve:

• Clarity
• Authority
• Knowledge demonstrated
• Experience

On-page and off-page describe where much of the work happens. CAKE describes qualities the work should help create. The two maps answer different questions, and together they make a large discipline easier to understand.

A map does not remove the need to go deep. It gives the depth a direction.

Once I knew which blocks existed, I would pursue them deliberately. I would learn how information moves through a business, how websites produce it, and how people make decisions with it.

Tools will change. Interfaces will change. Search results will change. The ability to understand a system, identify a constraint, and improve the quality of an answer will keep compounding.

I would learn the web properly

I would learn enough HTML, HTTP, JavaScript, rendering, routing, structured data, and analytics to understand what a website actually sends and how a platform produces it.

Not everyone needs to become a software engineer. Everyone working in SEO should be able to inspect a page, explain a technical problem clearly, work with engineers, and distinguish a platform defect from a content issue.

AI can generate code and explanations. Technical literacy is what lets you judge whether they are correct and safe.

I would learn how a business makes money

Traffic is not a strategy.

I would learn the products, margins, customer journey, markets, sales model, regulatory constraints, and economics of acquisition. I would ask which questions appear before a valuable decision and which information reduces uncertainty.

This changes prioritisation. A technically interesting project may matter less than improving a small set of pages that support a major product or market. A high-volume topic may attract people who will never become useful customers.

SEO becomes much clearer when the business consequence is visible.

I would learn research and evidence

Generative AI makes producing plausible text easy. That increases the value of knowing how to find primary sources, verify claims, distinguish direct evidence from inference, and state uncertainty honestly.

I would practise turning expert knowledge into clear public information. I would learn when first-hand experience strengthens an answer and when it is merely anecdotal. I would learn to cite the source behind a claim and to update or remove information when it becomes stale.

The future advantage is not producing more words. It is producing information that deserves to be used.

I would learn AI as a system, not a trick

Prompting matters, but prompts are only one component.

I would learn structured inputs, retrieval, tool use, evaluations, model limitations, privacy, cost, and human review. I would build small workflows that classify data, accelerate research, explain anomalies, or make recurring checks more reliable.

I would also observe how AI search products discover sources and construct answers. Google's current guidance explicitly says established SEO fundamentals still apply to its generative search features. That is a good reminder not to abandon durable skills whenever a new interface appears. Read Google's guidance.

I would learn to communicate across functions

The biggest SEO opportunities usually depend on people who do not report to SEO.

I would learn how to write a useful engineering requirement, present a business case, explain a trade-off to legal, give local teams room to disagree, and report to executives without hiding behind jargon.

Being correct is not enough if the organisation cannot act on the recommendation.

I would ignore several things

I would spend less time on:

• daily arguments about unconfirmed ranking factors;
• chasing every tool launch;
• producing generic content because a competitor has a page;
• treating a single AI answer as stable evidence;
• reports that celebrate activity without consequence;
• personal branding that grows faster than actual capability.

None of these creates a durable advantage.

The skills that compound

The most valuable skills reinforce one another:

1. Curiosity helps you find the real question.
2. Technical literacy helps you see how the system behaves.
3. Business judgement helps you choose what matters.
4. Evidence discipline makes the answer credible.
5. Communication helps other people act.
6. Automation lets good methods scale.
7. Leadership makes the capability survive beyond you.

I would still learn keywords, links, crawling, content, and measurement. I would simply learn them as parts of a larger operating system.

If I were starting again, I would not prepare for the search interface that exists today.

I would prepare to keep understanding whatever system comes next.