Every AI answer is a referral

When someone asks an AI tool "who can help me with X?", the tool is effectively making a referral. It is doing what a trusted friend or colleague used to do: weighing up who seems to understand the problem and pointing the person towards them.

Referrals have always gone to whoever understands our needs best. What has changed is who does the referring, and the evidence they use. An AI engine can only recommend you for a need if your content clearly, credibly and consistently shows that you meet it.

Search has moved from keywords to questions

For twenty years, getting found online meant guessing the words people typed into a search box. Now people ask whole questions, in their own words, and increasingly get a written answer back instead of a list of links.

The scale is hard to ignore. Google says AI Overviews now has over 2.5 billion monthly active users, and AI Mode has passed 1 billion. OpenAI says ChatGPT now has 1.2 billion weekly users. Ofcom's 2026 research found that 54% of adults now use AI tools, up from 31% a year earlier, and three-quarters of online adults read AI-generated search summaries at least sometimes.

Two findings from research on these answers matter most. First, the shape of the question matters: a study of 68,879 Google searches found that longer queries, and queries beginning with a question word, are more likely to trigger an AI Overview. Second, people click less. Pew Research Center found that users clicked a traditional result in 8% of visits when an AI summary appeared, against 15% when it didn't.

So the prize has moved. It's no longer just ranking on page one; it's being the source the answer is built from.

Guessing at keywords loses to knowing needs

Here's the opportunity most organisations are sitting on. The questions people ask AI tools are, very often, the same needs your customers already express to you every day: in reviews, complaints, support tickets, survey comments, sales calls and chat logs.

Customers don't say "omnichannel resolution optimisation". They say "I just want someone to sort it out the first time." They don't ask for "flexible billing solutions". They ask "can I change my payment date?" Those words, in that order, are what they will type or say to an AI assistant.

If you know those needs, you can write the answers that AI engines quote and cite. If you don't, you're guessing.

A candid example from my own business

I tested this on Customer Needs Solution itself. In a day-zero baseline on 3 November 2026, CNS appeared in 5 of 20 AI answers we could test (25%). Every one of those mentions came from branded questions, where someone already knew to ask about CNS. On the 9 general category questions, the ones a new prospect would ask, CNS appeared in none. The answers recommended familiar alternatives instead.

Worse, some answers got us wrong, treating our name as a generic phrase or misnaming our metric. That told me something important: if you don't publish a clear definition, an engine will invent one for you. So we published one, and we'll keep measuring.

The rest of this article is the approach we're now following, and the one I'd suggest to anyone in the same position.

Needs and wants: why the difference matters

In the Customer Needs Solution, the distinction is precise. In my own words from earlier articles: "Customer needs can be defined as the benefits the customer absolutely must be getting from the organisation if that organisation is going to be successfully in business at all. Customer wants, on the other hand, are benefits that will enhance the experience but are not fundamental needs."

The example I always use: "Customers of an airline, for example, need a seat in an aeroplane that will take them safely to their destination on time. A haute cuisine meal on the flight, by contrast, is a want, not a need." And a practical tip: customers understand the phrase "nice to haves" better than "wants".

This matters for AI search in two ways.

  • Needs drive the "can you help me?" questions. When a need isn't met, people go looking for someone who will meet it. These are the questions closest to a decision, and the ones where an AI answer becomes a recommendation.
  • Wants drive the "who does it better?" questions. They are where you differentiate, and where comparisons are won.

The measurement side matters too. CNS measures how well needs are met with the Net Customer Needs Score (NetCNS): positive experiences netted against negative ones, giving a score between −100 and +100. When you know which needs are met and which aren't, you know which questions you can honestly claim to answer well, and which you need to fix before you publish anything.

How to find the needs your customers are asking about

You almost certainly have the raw material already.

  1. Mine your open-text feedback. Survey comments, reviews and complaints are needs spoken in the customer's own words. Read what people say, not just how they score. Group the statements into needs and wants rather than counting keywords.
  2. Listen to support contacts. Every repeat call or ticket is a question your website hasn't answered yet. In CNS terms, an unmet core need doesn't produce a complaint; it produces more demand.
  3. Ask your sales team. The questions prospects ask before they buy are exactly the questions they are now asking AI tools.
  4. Don't wait for volume. In CNS we count statements, not customers, and saturation can be reached with as few as 300 statements, depending on the context. You only need to hear something once.

SEO, AEO and GEO, in plain English

Before you publish, it helps to know the three terms you'll hear, because they all point the same way.

  • SEO (search engine optimisation) is making your pages easy for search engines to find, understand and rank. It still matters: Google says the best practices for SEO remain relevant for its AI features.
  • AEO (answer engine optimisation) is shaping your content so it directly answers the questions people ask, in a form an engine can lift and quote: a clear question, a short direct answer, then the detail.
  • GEO (generative engine optimisation) is about being included, accurately, in the answers that AI tools like ChatGPT, Claude, Grok and Perplexity write. The term comes from academic research presented at KDD 2024, which found that adding citations, quotations and statistics could lift a source's visibility in generative answers by up to 40%, while traditional keyword stuffing did little or nothing.

Hold on to that last point. Guessed keywords don't impress engines that read for meaning. Being the clearest, most credible answer to a real question does.

Turning needs into answers AI tools can cite

Once you know the needs, publish answers to them.

  • One need, one clear answer. Use the customer's question as the heading. Start with a 40 to 60 word direct answer, then add the detail.
  • Build real FAQ pages. Give each important question its own anchor or page, so it can be cited directly.
  • Define your terms. If you use a term, define it in one sentence and repeat that sentence word for word everywhere: your website, profiles, directory listings and author pages.
  • Keep facts consistent across the web. Names, prices, credentials and descriptions should match wherever they appear. Engines cross-check, and contradictions cost you trust.
  • Use structured data honestly. Schema markup (for example Organization, FAQPage or Book) helps machines read your facts. Google is clear that no special markup is needed for its AI features, and that structured data must match the visible text on the page. Treat it as a fact sheet, not a trick.
  • Back claims with evidence. Cite sources and use real statistics. That is what the GEO research found helps, and it is what customers deserve anyway.

Where to start

Begin with what your customers already tell you. Find the needs that matter most, answer the questions behind them as clearly as you can, and make every fact about you consistent wherever it appears.

That is the thinking behind the Customer Needs Solution. As we define it: "Customer Needs Solution (CNS) is a customer experience (CX) framework, created by Stephen R Hewett, for understanding what customers actually need and measuring how well those needs are being met. It analyses what customers say, not just how they score. Its metric, the Net Customer Needs Score (NetCNS), complements NPS: NetCNS is to operations what NPS is to marketing." If you'd like to explore it, you'll find more at customerneedssolution.com.

The businesses that know their customers' needs have always earned the most referrals. AI just made the rule harder to ignore.