07. Why Learn SEO in the AI Era: The New-Keyword Playbook for Beginners

A disclaimer before anything else: this is not a battle report.

I have not run one of these sites myself. What follows is part other people's retrospectives, part my own judgment, and part things that kept showing up, sounded right, and have been verified by exactly nobody on my behalf. I've tried not to blend the three. When I do run a site end to end, this page gets rewritten with real numbers.

So why write it at all? Because the more I dug, the more I thought this deserves attention now, not later.

1. While the window is still open

I have a feeling about SEO that I can't fully prove and can't shake either: it may be a skill you have to learn before the sun sets on it.

Think about it. Pretty soon everyone will just ask AI when they have a question. Searching moves inside the conversation. And then — which keywords are rising, what people are really looking for, where traffic gets divided up — all of that data sits in the model companies' hands. Why would they open it up for the rest of us to study?

Right now it's different. Google still leaks feedback to the outside, bit by bit, and if you're willing to do the work, you can still touch what real demand looks like. I don't expect that window to stay open forever.

Besides, SEO is not only SEO. What you're really training, once you go deep, is how to read trends, unmet needs, and the outline of a business out of a large volume of search data. Picking a topic uses it. Pricing uses it. Deciding whether a direction is worth your year uses it. If you build products, you will take this lesson sooner or later. Sooner is cheaper.

2. Beginners: touch new keywords first

What does a beginner lack most? Everything. There's no accumulated base at all.

The old keywords are already crowded, and the contest there is domain age, depth of content, backlinks — all things that only come with time. A beginner walks in with a fresh site and ends up ranked a few hundred places deep. The traffic belongs to somebody else. You're just there to make the results page look full.

New keywords are the one opening. The term just appeared. Searches are climbing, but barely any sites have claimed their places yet. Everyone's start line is close together for once, and a beginner gets to taste a little feedback early.

The work, though, is not smaller. Study what competitors wrote. Fill in your own content and pages. Chase backlinks. And then — wait. Wait for Google to release traffic to you a trickle at a time. It's an anxious, slow process, but trust is built exactly like that and there is no shortcut on offer.

3. Close the loop first. See actual money.

After years of building products, I hold to one rule: get the loop working at the lowest cost you can. Scale is a later conversation.

For a new-keyword site, I break the loop into four steps: find a new keyword, launch fast, fix problems, get paid out in dollars.

The first two test judgment and speed. The third is grinding against real feedback. The one people skip is the payout. Plenty of builders get a site up, watch some traffic arrive, and call it done. But if no money came in, the thing is still sitting at "I think this works." A loop earns its name by closing. Collect even one tiny payment and your understanding of the whole thing changes. After that, whether to copy it or speed it up is a decision you can actually make. And if the money never comes — admit it early and move on.

4. Speed is a hard requirement; weight is a consolation prize

With new keywords, slow means you don't get to play.

The window is that short. One day later and there are a few more sites in front of you. Traffic that gets taken is very hard to take back. So the domain, the template, the page structure — all of it has to be sitting ready beforehand. A keyword appears, you wrap it and ship. Anyone who starts building the site when the keyword shows up has already missed it.

Accept the other side of it, too: most new keywords are short-lived. The heat passes, traffic falls off a cliff, sometimes to nothing. Building your hopes on one trending term feeding you for years is not a plan.

What's left afterwards? Weight. Once pages are indexed, your site has a footprint with Google. The traffic can go and the footprint stays. It's the consolation prize of the trend site game, and it's also what makes the whole thing slightly less disposable than it looks.

One more claim showed up in the retrospectives: that even a keyword at its peak is worth chasing, and you can afford to be aggressive about volume. I'm parking that one here without endorsing it. I haven't run it, so I won't vouch for it.

5. Backlinks, and impulsive people

Backlinks: if you can get them, get more of them.

Content you can write yourself. Pages you can build yourself. Backlinks need somebody else to say yes. Of everything that goes into accumulated trust, this is the piece least in your own hands — which is why being passive about it makes no sense. Trade for them, ask for them. Politeness earns nothing here.

Trend site users have a trait that stuck with me: they're impulsive. Search, click, see that this is exactly the thing, and the purchase happens inside a few minutes. The retrospectives kept repeating one pattern — lower price, higher conversion. It sounds crude, but the logic holds: this person did not arrive to compare three vendors. The heat of the keyword pushed them through your door. Price greedily for margin and the first casualty is usually the conversion you were counting on.

6. Traffic that AI sends you is different

The biggest surprise of this whole research round was AI-referred traffic.

By several retrospectives' accounts, users who arrive because an AI recommended you buy noticeably more readily. When you think about it, it's not strange. This person had a conversation first. They explained what they needed, the AI listened, filtered, and only then sent them to you. Their intent was screened before they ever knocked.

A person like that is simply not the same as someone idly digging through a results page. The traffic is much more precise.

And this loops straight back to where I started. The more normal AI interaction becomes, the bigger this stream gets. Whoever works out first where it comes from and how to receive it takes the spot first.

7. A base: build once, use many times

Last, a tooling point — probably the part that comes most naturally to someone with my background.

A SaaS system can be reused. Login, payments, page templates, analytics: once one site has worn all of that smooth, you own a base. Next new keyword, you swap the content and the skin and the site is out the door. Nobody rebuilds from the first line of code every time. Most of the "speed" from section four is standing on this base.

A base is an engineering problem, and engineering problems are the friendly kind. You can prepare them early and polish them slowly. The genuinely hard part is upstream of it: which keyword deserves the chase, and where you stop. There's no template for that. It only comes from getting in, round after round, yourself.

No big finish

Having written all this down, my view of the playbook got simpler, not grander: it makes money out of a time window.

You spot it early, you move fast, you collect what there is to collect while the heat lasts. Why does the user pay? Because he found you, and because it was easy. Whether the thing itself is any good ranks, honestly, pretty far down the list in this particular business.

A long-term product is a different story entirely. That one lives on accumulation, on somebody still coming back six months or a year later. Neither is nobler. But mixing them is fatal. Do trend sites at a long-term product's pace and the window leaves without you. Build a long-term product with a trend chaser's mindset and, at the end, you're holding nothing.

These are my notes from before I've started. The next step is to run a site myself. Then each of these "reportedly"s gets tested — the ones that survive stay, and the rest get crossed out.