Who are you still writing content for?
Last updated:
Your customer no longer googles a question. They ask ChatGPT, or read the AI answer at the top of Google. Answer delivered, no click to your help center. That is not a problem. It is a different playing field. And it changes exactly who you write for.
Not a fit for you? No problem. No is an answer too.
What is changing
The customer no longer searches at your place. The answer is already at their door.
The question is no longer “does the customer find my help center article?” but: does an AI know the right answer, because it pulls it from somewhere? Same question, a completely different path to the answer.
- Customer has a question
- Google — 10 blue links
- Your help center — click = you win
- Customer has a question
- ChatGPT / AI Overview — answer right away
- Your help center — no click
From “does the customer find my article?” to “does the AI know the right answer?”
The new content flow
From now on you write for three readers at once
One piece of content, three recipients, each with their own demands. Take that into account and you win on all three at the same time.
The customer searching for themselves
Classic self-service: clear, findable articles in your help center. This stays the base. But only if the content matches what customers actually ask (your C1-C2s), not what marketing thinks they ask.
The AI that searches before your customer
ChatGPT, Google AI Overview and Perplexity prefer to cite clear, self-contained answers in customer language. Corporate copy scores badly here. One question, one clear answer, no marketing wrapper. That wins the citation.
Your own AI agent
The most important one long term. Your support bot can only answer well, or actually resolve, if your knowledge base is current, structured and unambiguous. Messy or outdated content is invisible to an AI. Or worse: it makes something up.
The pitfall
Service is not marketing. One question deserves one answer.
“Marketing wants five variants. Service wants exactly one.”
Marketing often deliberately writes multiple versions of the same question: every variant its own landing page, every page a shot at a lead. That works for generating leads, not for service. In service, one question needs one answer: unambiguous, findable, not contradicting the five other versions sitting somewhere else on your site. Different intent, different writing rule.
From insight to execution
What this means in practice
Four rules that make your content ready for humans and AI. No theory, just things you can pick up on Monday.
Fix the foundation first (House of CX)
Know which questions actually come in (C1-C2) before you start writing. Otherwise you write beautiful answers to questions nobody asks.
Write in customer language, not company language
That works for the human reading it and for the AI citing it. The same words your customer uses.
One source of truth
The same content feeds your help center, the search AI and your own support agent. Right once, right everywhere.
Maintenance is not a side task
Content that is not kept up to date is, to an AI, the same as content that does not exist. Outdated is invisible.
- Help center — for the customer searching themselvessame answer
- Search AI — ChatGPT · AI Overview · Perplexitysame answer
- Your own support agent — answers and resolvessame answer
One source of truth, three channels, zero contradictions.
The question that matters now
If nobody visits your help center tomorrow, does an AI still know how your product works?
Evidence, not assumption
Everyone is claiming this now. We would rather check it in your data.
Zero-click support, AI search taking over the help desk: big statements, little evidence. Before we sell it as a lesson, we look at your own numbers. Four checks.
Question-pattern check
Are your C1-C2 volumes shifting? Fewer simple “where is my order” tickets can mean customers already find that answer somewhere else, with an AI.
Volume-versus-revenue check
Is your contact volume dropping while orders and revenue stay flat? That is a real signal, not an assumption.
AI-answer check
We test ourselves whether ChatGPT and Google AI Overview describe your product correctly. Or whether they guess.
Content-quality check
We link a declining volume on a C1 to the state of your knowledge base. So you know whether it is a trend or a content problem.
We always tie this to a CXReport scan of your own customer data. So it is evidence, not assumption.
Frequently asked questions
What is zero-click support?+
Zero-click support means the customer gets their answer before they ever click through to your site: through ChatGPT, Google AI Overview or Perplexity. The question is solved, but you see no visit and no ticket. Your content still works, just in a place you do not measure.
Who am I actually writing service content for?+
For three readers at once. One: the customer searching your help center. Two: the search AI that cites your answer. Three: your own AI support agent, which can only answer well if your knowledge base is current and unambiguous.
Why is service content different from marketing content?+
Marketing deliberately wants five variants of the same question, each with its own landing page and a shot at a lead. Service wants exactly one: unambiguous, findable and never contradicting another page. Different intent, different writing rule.
How do I get an AI to cite my answer?+
Write self-contained answers in customer language: one question, one clear answer, no marketing wrapper. Use the same words your customer uses, keep the content current and make sure only one version of the truth exists.
What does C1-C2 categorization have to do with content?+
Your C1-C2 structure shows which questions actually come in. Without that base you write great answers to questions nobody asks. Fix categorization first, then prioritize content by volume and impact.
How do I know whether this is happening in my organization?+
Four checks on your own data: are your C1-C2 volumes shifting, is contact volume dropping while revenue stays flat, do ChatGPT and AI Overview describe your product correctly, and what is the state of your knowledge base on the declining C1s. Evidence, not assumption.
Further reading
The full methodology
Want the full methodology? Start the free course.
The C1-C2 Masterclass teaches you step by step how to build a watertight ticket taxonomy. Five modules, about two hours.
Start the course