Many marketing managers are still writing content with one goal: get the page
ranked. That works for traditional search, but AI answer engines work
differently. ChatGPT, Perplexity, Claude, and Gemini don't rank your site, they
extract passages from it to answer specific questions. If your content doesn't
contain a passage that directly answers the question someone is asking, it
doesn't matter how well you rank. The AI has nothing to cite.
The problem isn't that keywords are irrelevant. It's that keyword thinking
optimises for presence rather than directness. A page can contain every
relevant term for a topic and still fail to answer the implicit question a user is
asking. When someone searches "best physio for lower back pain Auckland,"
they're not looking for a page that mentions those words frequently. They want
to know who can help with their specific problem. Keyword-focused content
treats that as a semantic match problem. AI treats it as a question that needs
a direct answer.
This matters because of how AI answer engines actually retrieve information.
Unlike traditional search, which ranks pages and sends users to them, AI
systems use Retrieval Augmented Generation to pull specific passages from
across the web and synthesise them into responses. RAG doesn't retrieve
pages. It retrieves the most relevant chunks of text that directly address the
query.
That changes what good content looks like. Domain strength and backlink
profiles matter less than whether your content directly answers the question. A
single paragraph that directly answers "What conditions do you treat?" will get
retrieved over a 2,000-word guide that covers the same ground without ever
stating a clear answer in plain language.
The change, when it comes, happens all at once rather than in small steps. AI
systems don't reward incremental improvement the way traditional search
does. They're making a simple assessment: is this passage clear and specific
enough to stake a recommendation on, or isn't it? If it clears that bar, it gets
cited. If it doesn't, it gets passed over entirely, regardless of how close it came.
You can do seventy percent of the right things and see no change at all,
because you haven't cleared the threshold. Then you fix the one remaining
structural issue, usually the directness of the opening paragraph, or the
coherence of the heading structure, and suddenly you're appearing across
multiple queries at once. The system isn't rewarding effort incrementally. It's
rewarding clarity absolutely.
The fix is simpler than it sounds. Every section of your content should answer a
specific question. The simplest diagnostic is to audit your headings. If they're
keyword phrases ("Auckland Physiotherapy," "Lower Back Pain Treatment,"
"Sports Injury Specialists"), they're not doing the work AI needs. Those headings
signal topics, not answers. Compare that to "What conditions do you treat?" or
"How do I book a first appointment?" Those headings tell the AI exactly what
the following passage addresses, which makes the content retrievable when
someone asks that question.
The same principle applies to opening sentences. The first sentence of each
section should directly answer the question the heading poses. Context and
nuance follow, but the answer comes first.
Testing this doesn't require sophisticated tools. Open ChatGPT, Perplexity,
Claude, or Gemini, type the questions your content is meant to answer, and
see whether your pages get cited. Run it at least three times across separate
sessions, because AI responses vary and a single result tells you very little.
What you're looking for is threefold. First, do you appear at all. Second, if you do
appear, what passage is the AI pulling, is it something specific and accurate,
or is it a vague description that could apply to anyone in your category. Third,
who else appears, and what does their content seem to be doing that yours
isn't.
If you don't appear, that's the finding. It doesn't mean your business isn't good
enough, it means your content isn't structured clearly enough for the AI to
extract anything citable. The ideal result is appearing with a specific, accurate
description that you'd actually be happy for a prospective client to read.
If you want this checked systematically, ADMATIC offers a free AI visibility check
for Marketing Association members. It runs across ChatGPT, Perplexity, Claude,
and Gemini to show how each one currently describes your business and
where you're absent, and checks whether the AI crawlers can reach your site in
the first place.
The broader implication is that the mental model most marketing managers
inherited from traditional SEO, build comprehensive pages, optimise for
keyword clusters, establish topical authority, doesn't translate to AI search. AI
visibility isn't about ranking pages. It's about having retrievable passages that
directly answer the questions people ask. That requires different content
structure, different editorial thinking, and a willingness to prioritise clarity over
comprehensiveness.
For New Zealand businesses, this creates an unusual opportunity. The US and
UK markets are about eighteen months ahead of NZ in terms of how seriously
businesses are taking AI search visibility. What we're seeing there is that the
early movers, businesses that restructured their content and built their entity
presence in 2024 and early 2025, are now sitting in consideration sets that are
proving surprisingly stable. Research indicates that once a brand appears in
the top positions of an AI consideration set, it appears in roughly 55 to 77 percent of relevant queries, even though the exact composition of the answer changes every time.
Right now the consideration sets for most NZ category queries are genuinely
open. Nobody has done enough work yet to earn that settled view. A business
that moves now can establish itself in that consideration set whilst the field is
still forming. A business that waits another twelve to eighteen months may find
the set has formed without them, and getting in becomes a much longer
project.
The pattern we're seeing internationally is that the businesses that move first
are the ones that are hardest to displace later. Early movers are building a
signal history that compounds over time, making it progressively harder for
latecomers to catch up. NZ businesses are still largely writing for the search
paradigm they know. That gap is the opportunity, and it won't stay open.
Businesses that start now, audit their content for question-answer structure,
and test in AI engines have a window to establish visibility before the market
catches up. That window won't stay open, but right now it's there.
Frequently Asked Questions
Should we rewrite our existing pages or start fresh?
Rewrite your highest-value pages first. Identify the three to five pages where AI
visibility would matter most commercially, then restructure those to answer
specific questions directly. Most pages can be fixed by rewriting section
openings and changing headings from keyword phrases to questions. Starting
fresh rarely makes sense unless the content has no clear purpose.
Will restructuring content for AI search hurt our normal Google rankings?
No. The changes that make content extractable for ChatGPT, Perplexity,
Claude, and Gemini, direct answers, question-format headings, clear structure,
improve readability and relevance for traditional search too. The pattern we're
seeing is that pages restructured for AI citation tend to hold or improve their
organic performance rather than lose it. You're not trading one for the other.
How long before restructured content starts getting cited?
Perplexity typically reflects updated content within four to six weeks. ChatGPT
and Claude take longer depending on whether they're using live search or
training data. Gemini sits somewhere in between. The change tends to
happen all at once rather than in small steps. Once content clears the
threshold, it often appears across multiple queries simultaneously.
Is this just adding an FAQ section to every page?
No. The principle applies to your entire page structure, not just an FAQ block.
Every section should answer a specific question: the heading poses it, the first
sentence answers it, the rest provides context. Adding an FAQ at the bottom
whilst leaving the rest structured for keywords misses the point entirely.