How AI search is rewriting the SEO playbook | Leveret Drive

How AI search is rewriting the SEO playbook

Half of what you read about this is panic and half is denial. The useful version sits in between and is fairly boring.

Jon Usborne
· 10 min read
A dim workspace lit by a single screen

There are two confident positions on this and both are wrong.

One says search is over, organic is dead, start again. The other says nothing has really changed, keep doing what you were doing. Neither describes what we are actually seeing in client accounts.

What genuinely changed

The easy queries stopped paying

Definitions, comparisons, how-does-this-work, what-are-the-steps. Those used to be reliable traffic. A summary answers them at the top of the page now and the click does not happen.

If a meaningful share of your content was built to capture that kind of query, that content is not going to recover. Not because it got worse. Because the job it did is being done somewhere else.

Being the source became a separate outcome from ranking

You can rank well and not be quoted. Ranking makes you eligible. Whether a specific sentence of yours ends up inside an answer depends on whether that sentence works on its own, without the paragraph around it.

This is the change most teams have not internalized. It is not a new channel. It is a second, quieter competition happening on top of the one you already understood.

Off-site presence became load-bearing

Models assemble a view of you from everywhere. Directories, reviews, roundups, forums, transcripts. An unlinked mention used to be a nice bonus. Now it is a source, which means your visibility depends partly on rooms you do not own.

What did not change

More than the panic suggests.

  • Technical health. A site that is slow, broken, or hard to crawl loses in both worlds.
  • Genuine expertise. Thin content failed before and fails now, just faster.
  • Intent. Somebody ready to hire still behaves like somebody ready to hire.
  • Proof. Case studies, reviews, and specifics still do the persuading.

If you had a solid program, most of it still works. You are adding a layer, not replacing a foundation.

What we changed in our own work

We stopped writing for questions a summary can finish

The topic filter now is simple. Can a two-sentence answer satisfy this. If yes, we might still cover it briefly for completeness, but we do not build a page around it and we do not expect traffic from it.

What we write instead: judgment calls, tradeoffs, opinions with a position, and anything that needs specific experience to answer. A summary can define a term. It cannot tell you which of two approaches to pick when your situation is awkward.

We rewrote how paragraphs open

Every load-bearing statement now stands alone. No "as mentioned above." No "this means that." If a sentence carries a fact worth quoting, it names its own subject.

This felt unnatural for about a week and then stopped being noticeable. It also made the writing clearer for people, which was not the goal but is a nice outcome.

We put sources inside sentences

Numbers travel with their source and their date in the same sentence rather than in a footnote or a chart caption. A statistic that arrives without attribution is much less useful to anything trying to quote it responsibly.

We started measuring differently

Sessions dropped down the report. Qualified conversations, branded search volume, and direct traffic moved up. We also check by hand, monthly, what a few assistants say when asked the obvious questions in our clients' categories. It is manual and slightly tedious and it is the most useful thing on the list.

Ranking is eligibility. Being quoted is the outcome. Treat them as two scoreboards.

What we would tell someone starting now

Do the fundamentals, because they still determine whether you are eligible at all. Then spend your remaining effort on two things: making your best pages extractable, and getting mentioned in the places your industry actually talks.

Skip the tooling debate for a while. Most of the platforms selling AI visibility dashboards are measuring something you can check yourself in twenty minutes, and the checking is more useful than the dashboard because you read the actual answers.

The honest uncertainty

Nobody running a marketing program has complete visibility into this. The systems change, they disagree with each other, and there is no report that tells you why you were left out of an answer.

What we are confident about is narrower: clear writing, real expertise, structural discipline, and a presence beyond your own domain have never been the wrong bet. That is what we are running, and we adjust as things move.

What to take from this
  • Informational queries stopped producing clicks. Content built for them will not recover.
  • Ranking and being quoted are separate outcomes with separate causes. Track both.
  • Write for judgment calls and tradeoffs, not for questions a two-sentence summary can finish.
  • Make load-bearing sentences stand alone and keep sources inside the sentence.
  • Check by hand what assistants say about your category each month. It beats most dashboards.

Questions people ask

Is SEO still worth investing in with AI search?

Yes. Search visibility is what makes a page eligible to be used as a source, and AI features largely run on the same ranking and quality systems. What changes is which queries produce clicks and how content should be structured.

Which content lost the most traffic to AI answers?

Definitional and explanatory content. Queries asking what something is, how a process works, or how two things compare are now commonly resolved in a summary without a click.

What kind of content still earns traffic?

Content requiring judgment or specific experience: tradeoffs between approaches, opinions with a clear position, pricing realities, and anything where the answer depends on the reader situation.

Do we need an AI visibility tracking tool?

Not to start. Asking several assistants the questions your buyers ask, monthly, and recording who gets named gives you the same signal plus the actual wording, which is more useful than a score.

Jon Usborne, founder of Leveret Drive
Written by

Jon Usborne, Founder

Jon runs Leveret Drive, where the team builds strategy, brand, websites, video, and demand systems for businesses that want to be the obvious choice.

Keep reading
Follow in Google

Want this kind of thinking pointed at your business?

Tell us what you are working on. We will tell you what we would do first.

Chat with us →