Search has changed. And if you haven’t rethought your SEO approach in the past year, the distance between you and the brands that have is growing by the week.
People aren’t just typing two-word queries into Google any more. They’re sitting down with ChatGPT, Gemini or Perplexity and having what amounts to a conversation: one question, then a follow-up, then another follow-up that’s more specific still. We’ve been watching this pattern with our own clients for months. When Adobe polled 1,000 people last July, 770 of them said they’d already used ChatGPT the way they’d use Google. That figure stopped us mid-meeting. It confirmed what we’d suspected: something fundamental has shifted in how people look for information online. And it’s given a name to a discipline we’ve been quietly building around at AIWIZ for the better part of a year: prompt research.
What Is Prompt Research?
Put simply, prompt research is the practice of studying how real users phrase questions to AI systems, then using those patterns to shape your content so it shows up in the answers those systems generate.
If you’ve spent any time doing SEO, keyword research is probably muscle memory by now. Prompt research is what happens when that same instinct meets a completely different kind of search behaviour. You’re no longer chasing two- or three-word phrases. You’re dealing with full sentences, sometimes rambling ones, that carry genuine context about what somebody wants to know. Casey Nifong wrote a useful piece on this for Search Engine Land in March 2026. He drew a sharp line between the two disciplines: keyword research maps search queries and ranking opportunities; prompt research maps the questions that lead AI systems to explain topics, compare options or recommend specific brands. That distinction matters, because it means your team has to get good at spotting recurring question patterns, grouping them by topic, and thinking through how someone’s line of enquiry is likely to develop over three or four follow-up prompts.
We keep coming back to the same example internally. Say you’re an agency trying to get cited. A prospective client might open ChatGPT and type something vague:
‘Which digital marketing agency should I use?’
Not satisfied with the generic answer, they add a constraint: ‘Which agency specialises in AI-driven SEO in the UK?’
And tighter still: ‘Which Manchester digital marketing agency has AI expertise?’
Three prompts, three different answers from the AI. If your content doesn’t speak to each stage of that narrowing process, you won’t feature in any of them.
Why This Matters Right Now
Ranking on page one of Google used to be the finish line. It isn’t any more. The real prize now is being cited, recommended or summarised inside an AI-generated answer.
Generative engines don’t hand you a list of ten blue links. They pick the sources they consider most credible for a given prompt, pull out the relevant bits, and present them as a single answer. Your content isn’t fighting for a ranking position. It’s fighting to be the answer. Yoast’s Alex Moss and Carolyn Shelby said much the same in their 2026 predictions: the metrics that matter now are inclusion, citation and trust, not just traffic numbers.
We’ve watched this shift play out across the AIWIZ client portfolio. The brands producing well-structured, expertise-led content are turning up in AI responses. The ones still coasting on thin, keyword-stuffed pages? They’re nowhere.
How to Build a Prompt Research Strategy
You don’t need to start from scratch. Most of the raw material is already sitting in your existing keyword data and content. What you need is a different lens.
Begin with scope. What services actually make you money? Which postcodes or regions do you serve? What does your ideal buyer need to know before they pick up the phone? If the research isn’t tied to what the business sells and where it sells it, you’ll burn hours chasing prompts that bring traffic but zero enquiries. We learned that one the hard way on an early client project.
Then go digging for the questions people are genuinely asking. Google Search Console is the obvious starting point for long-tail data. But don’t sleep on Bing Webmaster Tools. Bing’s index now powers Microsoft Copilot, ChatGPT’s search functionality and DuckDuckGo, which means the data sitting in your Bing Webmaster Tools dashboard is far more commercially relevant than it was two years ago. A Seer Interactive study found that 87% of SearchGPT’s citations matched Bing’s top organic results. If that doesn’t shift where you spend your indexing effort, nothing will. Ahrefs is useful for pulling out the longer, more conversational queries that people tend to feed into AI tools, and People Also Ask boxes are still a reliable source of real questions worth answering.
With the raw queries in hand, sort them by intent. A single topic will throw up informational prompts, comparative ones, transactional ones and recommendation-style ones. Each type needs different content. Transactional and comparative prompts are where the money is, so start there.
After that, think in clusters rather than individual keywords. Take ‘AI SEO agency UK’. Instead of one thin page built around that phrase, you’d create a single in-depth piece covering how to pick an agency, what AI-powered SEO actually looks like day to day, and where it parts company with the traditional approach. That’s what AI engines want to pull from: thorough, connected content on a topic. Not a dozen shallow pages each targeting one variation of the same phrase.
There’s one more thing worth mentioning, and it’s easy to overlook: how your content reads to the AI itself, not just to a human visitor. We’ve noticed a clear pattern in the pages that get cited in generative results. They’re not fancy. They’re written in straightforward language, broken up with sensible headings, and clearly authored by someone who’s done the work rather than summarised someone else’s. Waffle gets filtered out. If your page buries the answer three quarters of the way down, an AI engine will skip it entirely and cite someone who got to the point.
What to Avoid
The mistakes we see most often are all variations on the same theme: treating prompt research as if it’s just keyword research with longer phrases. It isn’t. Sticking to classic keyword logic, ignoring multi-step prompts, churning out generic content with no particular point of view. All of it leads to the same outcome: invisibility in AI answers.
The single most damaging habit, though, is never testing. Open ChatGPT or Perplexity right now. Type in the questions your prospective clients would ask. Look at what comes back. If you’re not in the answer, that’s your gap. Prompts change fast, and your content strategy has to keep up.
Our Approach: Built on Strong SEO Foundations
We don’t treat prompt research as a novelty at AIWIZ, or some separate workstream bolted onto the side of an SEO retainer. For us it’s the natural next step in a job we’ve been doing for years: working out what your audience needs to hear, then making sure you’re the one saying it best.
The fundamentals haven’t gone anywhere. Technical SEO still matters. Authoritative content still matters. Brand credibility, the kind you can’t shortcut, still matters. What’s new is that you also need to account for how AI systems read, interpret and surface your pages. Envisionit’s senior SEO team put their finger on it in a late-2025 outlook: the brands earning visibility write in plain, precise language and structure their information so that both crawlers and real people can scan it, digest it and act on it quickly. The winners we’re seeing right now aren’t just ranking well on Google. They’re showing up in Perplexity answers, Copilot summaries and ChatGPT citations too.
The Bottom Line
This is a proper inflection point. The logic that ran SEO for two decades (find the keyword, rank for the keyword, collect the click) is giving way to something less tidy but more honest: understanding how people actually talk to AI when they want answers, and making sure your content is part of that conversation.
The businesses that grasp this now will be very difficult to catch later.
Want to find out how your brand is performing in AI-generated search? Get in touch with the AIWIZ team.