AI search has gone from curiosity to revenue channel. Use our AI search maturity model to see where your brand sits, the questions to ask at each stage, and how to move up.
Tom LightIn November 2025, we asked more than 300 marketers at our From curiosity to clarity webinar where they were on the AI search journey. 90% placed themselves in the first two stages. They were aware something was changing, perhaps checking ChatGPT now and then, but had no real way to measure it and no plan to act on it.
A month later, we followed up with GEO search is already here: is your website ready?, which set out a readiness checklist and a first set of steps.
Ten months on, the picture looks very different for the brands we work with. We've taken most of our clients from exploration into integration, and some into optimisation. Along the way, AI search stopped being a curiosity and became a revenue channel.
The most useful thing we've learned is simple: you can tell exactly where a brand is by the questions it's asking. In this article, we'll walk through what happened at each stage, the questions our clients asked, and what it took to move them to the next one. By the end, you should know which stage you're at and what to do next.
Our maturity model has five stages. The first two are about observing AI search. The last three are about influencing it.
Awareness: recognising AI search as a new traffic and visibility channel.
Exploration: starting to monitor AI traffic and test prompts.
Integration: embedding AI search into SEO and content workflows, and measuring its impact on traffic and conversions.
Optimisation: actively shaping how your brand appears in AI answers, and testing what works.
Leadership: becoming the source AI trusts for your category.
The jump from stage 2 to stage 3 is where most brands stall. It's rarely a technical problem. It's usually about ownership, because no single team holds AI visibility, and measurement, because there's no way yet to connect visibility to revenue.
In the first two months of the year, we moved the majority of our clients from exploration into integration. Almost every conversation started with the same three questions.
Most analytics setups bury AI referrals inside referral or direct traffic. We set up dedicated AI channel groupings in GA4, so traffic from ChatGPT, Perplexity, Gemini and others could be seen and reported on as its own channel.
We built a baseline set of prompts for each client, based on their organic search data and the questions their customers actually ask. We then ran those prompts across the main AI engines to see where the brand appeared, and where competitors appeared instead.
Manual checks don't scale, and a single check can mislead. We set up AI visibility tracking, agreed which metrics to report on, and made sure the data sat alongside the rest of each client's search reporting.
None of this was optimisation. Stage 3 starts with making AI search visible and measurable, so it can be judged alongside every other channel. If you can't answer these three questions yet, this is your starting point.
Two changes to ChatGPT turned AI search from something our clients monitored into something that moved their numbers.
The first came on 7 May, when ChatGPT started turning brand names in its answers into clickable links to brand homepages. Across the brands Profound monitors, OpenAI referral traffic roughly doubled overnight (for sites that already held visibility). A mention in ChatGPT was no longer just a perception metric. It had become a referral.
The lift wasn't even across sectors. It was strongest where ChatGPT recommends companies, such as financial services, and almost flat where it recommends products, such as retail. For service, insurance and lead generation brands, this mattered a lot.
The second came in August, when OpenAI removed the daily limit on free text chats. By then ChatGPT had passed one billion monthly users, and people could now use it for longer and for more of their research.
For clients who had built their stage 3 foundations earlier in the year, the effect was immediately visible:
For one client, AI search went from the 10th-largest revenue channel at the start of the year to the 5th-largest in under six months
Another client saw AI-attributed revenue more than double in a single month, with ChatGPT driving almost all of it.
Because these clients had set up AI tracking in January, they could see exactly what was happening and respond. Without that tracking, the same growth would have shown up as an unexplained rise in direct or referral traffic.
There's one practical lesson from the May update too. With ChatGPT now linking straight to homepages, your homepage is a landing page for AI traffic again. It needs to work for a visitor who arrived because ChatGPT recommended you, and who may know nothing else about your brand.
Once clients could see AI search working, their questions moved from "what's happening?" to "how do we make more of it happen?" Three came up again and again.
The first step was diagnosis. When we segmented prompts by intent and by engine, patterns appeared that a single blended visibility score had hidden.
Brands often win the advice prompts and lose the money prompts. One client averaged 92% visibility on advice-led prompts, but just 18% on cost and coverage prompts. That's a 74-point gap at the most commercial stage of the journey.
The weakest engine can be where the money is. The same brand had 27% visibility in Google's AI Overviews and 24% in Gemini, but only 4% in ChatGPT. ChatGPT was where almost all of its AI revenue came from.
AI repeats whoever states the number. Third-party pages that publish clear figures, such as average costs, get pulled directly into answers and repeated over time. Brands that deflect those questions to "get a quote" hand the answer to someone else.
The honest answer is that nobody knows for certain yet. That's why we started running controlled experiments rather than making changes and hoping.
For one client, we designed a test to see whether showing an indicative price range on product pages improves visibility on cost-related prompts. Three test pages get a new pricing block, based on the client's own claims data. Two matched control pages stay as they are. We'll track visibility and share of voice weekly for at least six weeks.
AI tracking tools produce a lot of data. The value comes from feeding it into the rest of the programme:
The third-party sources AI cites become digital PR targets.
The objections AI raises about a brand become the content roadmap.
The prompts where competitors appear and you don't become content and keyword priorities.
As clients asked sharper questions, their prompt sets grew. Several expanded their tracked prompts by three to four times. A baseline prompt set tells you whether you're visible. Only a segmented one tells you where you're losing.
If you're asking these questions, you're at stage 3 and ready to start testing.
The clients now in the optimisation stage have made two shifts.
AI data has become routine. It's reported monthly and feeds directly into ongoing recommendations. AI search is no longer a side project. It's part of the organic search programme, with shared goals across SEO, content and digital PR.
Campaigns are built AI-first. These clients are creating campaigns designed specifically to improve their coverage in AI answers. Visibility in AI search is the primary goal. Because the same signals drive both, the work also brings cumulative gains in traditional search as it rolls out.
Campaign example: For one client, a research-led organisation with a large library of older articles, we updated a set of existing pages rather than creating new ones. Before the updates, Google showed AI Overviews for a growing number of the related keywords, but almost never cited the client's pages. After the updates, the client's pages began to be cited, and by the end of the period they were cited in around half of the AI Overviews shown for those keywords.
At this stage, the questions get more specific:
"Why does AI describe us like that?" For one brand, we analysed the objections AI engines raised most often, such as hidden fees and delayed timelines, and turned each into a piece of direct, citable content.
"What content actually moves visibility?" Our current answer is original, structured, fact-based content, especially content built on data only you have.
"How do we test changes in a regulated sector?" For regulated brands, compliance is often the real bottleneck. Framing changes carefully helps: a range rather than a quote, figures anchored to historical data, and existing approved copy reused wherever possible.
Being visible in AI search isn't the same as being chosen. The brands pulling ahead aren't the ones that appear most often. They're the ones AI trusts enough to recommend at the moment a customer decides. That trust comes from the same place it always has: clear answers, real data, and other people vouching for you.
Few brands are at stage 5 yet, but we can already see the questions coming:
"How do we protect our lead as competitors catch up?"
"How do we become the source AI cites for our whole category?"
"Are we ready for AI agents that compare, quote and buy on our customers' behalf?"
The answers will come from two places. The first is first-party data. Claims data, transaction history and local pricing are things competitors can't copy, and AI rewards unique, verifiable information.
The second is readiness for AI agents. AI assistants answer questions; AI agents complete tasks. As agents start comparing providers, requesting quotes and managing renewals, brands will need clear products, consistent information, and journeys and structured data that an agent can navigate reliably. Emerging standards such as WebMCP and the new agentic commerce protocols are the early signs of this shift.
The quickest way to find your stage is to look at the questions your team is asking right now.
If you're asking… | You're probably at… | Your next step |
|---|---|---|
Is AI search real? Does SEO still matter? | Stage 1: Awareness | Set up AI channel tracking in GA4 and check AI crawlers aren't blocked |
How much AI traffic do we get? Are we visible? How do we track it? | Stage 2: Exploration | Build a structured prompt set and set up AI visibility tracking |
How do we grow visibility? How long does it take? | Stage 3: Integration | Segment prompts by intent and engine, and start controlled tests |
Why does AI describe us like that? What content moves visibility? | Stage 4: Optimisation | Make AI data part of monthly reporting and build AI-first campaigns |
How do we protect our lead? Are we ready for AI agents? | Stage 5: Leadership | Build first-party data assets and agent-ready journeys |
Ten months ago, most marketers were watching AI search from the side-lines. Today, for the brands that started early, it's a measurable, growing revenue channel. The gap between the two groups is widening.
Join us later this year. We'll be running a follow-up event on GEO and AI search in Q4, keep an eye on our events page for details.
Our GEO discovery gives you a snapshot of how AI engines currently see your brand: where you're visible, where competitors appear instead, what AI says about you, and what to do next. Talk to us to get started.
Is Zero Click marketing dead? The branded link update, Profound, 19 May 2026. Note: Profound sells AI tracking tools, so attribute the data to them clearly.
OpenAI removes the daily limit on free ChatGPT chats, The Next Web, 6 August 2026.
From curiosity to clarity: Act now to get ahead in AI search, Fresh Egg webinar, November 2025.
GEO search is already here: is your website ready?, Fresh Egg, 10 December 2025.
