The gold rush is real. The measurement isn’t.
As AI search reshapes the buyer journey, the brands that win won’t just be the ones showing up in LLM answers. They’ll be the ones who can prove it. There’s no shortage of excitement around AI search right now. ChatGPT remains the dominant force, holding roughly 60% of the consumer AI chatbot market according to YouGov’s AI Index, while Gemini is surging fast on Similarweb, growing its share of Gen AI traffic to around 18% and tripling year on year. The shift isn’t coming. It’s here.
YouGov: Share of the consumer AI chatbot market
Source: YouGov AI Index, 2025.
And the queries themselves are changing shape. Traditional search averaged around three words. AI Mode queries run to ten. Informational prompts on ChatGPT can stretch to 31 words. Add context-heavy, multi-turn conversations and you’re looking at prompts of 60 words or more. People aren’t searching anymore. They’re having conversations. And those conversations are happening somewhere brands can’t easily see.
How a query refines inside AI search
| Step 1
Hair straighteners |
Step 2
Best hair straighteners |
Step 3
Best hair straighteners for thick and curly hair, £200–£300 that give beach waves |
A single traditional search evolves, inside an AI conversation, into a fully qualified purchase intent. The brand’s chance to appear sits at every step, not just the first.
The problem isn’t visibiliy, it’s measurement. According to SparkToro’s 2026 zero-click study, 69.5% of searches in the UK now end without a click (68% in the US), and LLMs refer less than 1% of prompts through to external websites. The decision about which brand to trust, which product to consider, which service to shortlist is being made inside the engine. Not on your site. Not in your funnel. Somewhere in the dark.
As Scott put it when we opened the webinar: “The question isn’t whether AI search is changing behaviour. It’s whether your measurement stack is keeping up.”
Why old measurement is breaking
For years, the performance marketing playbook ran on a simple chain: impressions led to clicks, clicks led to sessions, sessions led to conversions. Each step was trackable. Each step was attributable. The funnel had edges you could measure.
That chain is breaking. The middle of the funnel, the research, the comparison, the consideration, has gone dark inside LLMs. Users are getting answers without ever landing on a page. Impressions and sessions still exist, but they no longer tell the full story of how a brand is performing in the moments that matter most.
Ash, our GEO lead, framed it clearly: “We’re not losing the top of the funnel or the bottom. We’re losing the messy middle, and that’s exactly where brand preference is formed.”
The three-tier measurement framework
Emily, our Head of Data, presented a practical framework for thinking about AI search measurement. It doesn’t pretend to solve everything at once. Instead, it layers confidence levels, from what you can observe directly, through to what you can infer, through to what you need to model.
1. Observed: high confidence
This is your first-party foundation. GA4 referral data will show you traffic arriving from AI sources where it’s passed through. Google Search Console is now surfacing AI Overviews impressions data, giving you a direct signal of how often your content is appearing in AI-generated results. Bing Webmaster Tools, updated this week, now shows cited pages, intent topics and citation shares. As Ash noted, he hopes this pushes Google to follow suit with greater transparency.
These signals are limited in volume, but they’re reliable. Build them into your dashboard now.
2. Proxy: medium confidence
When direct data is thin, you borrow from adjacent signals. This is a tactic the industry learned when social platforms went dark, and it applies just as well here.
Post-purchase surveys: A simple “how did you hear about us?” question at checkout captures the AI-influenced journeys that never touched your analytics. Graphite’s study with n8n compared GA4 last-touch data against post-conversion surveys and found a 9x attribution gap: GA4 credited AI-driven answer engines with roughly 1% of conversions, while surveys showed the true figure was closer to 9%. Nine out of ten AI-driven conversions were invisible to standard analytics.A similar study was done with Otterly.AI where GA 4 attributed AI to 0.1% of sign ups vs 21% of self reported sign ups.
Branded search lift: If AI is driving awareness, you should see it in branded search volume over time. Track it as a proxy for share of voice. Similarweb research found that users who receive an AI recommendation are 2.5x more likely to visit that brand’s website within seven days, with around 56% of AI-influenced visitors arriving via branded search. Profound research showed a similar effect when AI recommends brands users visit that brands website at 1.5- 2.25x the forecasted baseline over the next 7 days.
Bot and crawler visit logs: AI systems crawl your content before they cite it. Monitoring these logs gives you an early signal of which pages are being indexed by LLMs.Just make sure to know how often and the exact time you run them, in order to understand that spikes in crawling are your AI visibility tool and not performance. AI share of voice vs competitors: Are you being cited more or less than your closest rivals? Even rough benchmarking here is valuable.
3. Modelled: lower confidence, higher context
For the 80% of AI search activity concentrated in ChatGPT and Gemini, where first-party data is almost non-existent, you need modelled visibility. A growing set of tools now surfaces this: Profound, Peec AI, Semrush’s AI Visibility Toolkit, and Ahrefs Brand Radar all provide visibility scores, citation tracking, mention frequency and sentiment analysis across the major LLMs.
These aren’t perfect signals.They are directional at best until the AI infrastructure is built out using by ads, which is what happened previously with Google ads and Google keyword planner. SEO’s were actively using Google ads to understand volumes. But they’re the best lens available for understanding how your brand is performing in the spaces you can’t directly observe. Emily’s view: treat them as directional, not definitive, and triangulate against your proxy data wherever possible.
Start with a North Star prompt set
Before you invest in any tooling, you need a prompt set, a defined collection of queries that represent how your target audience is likely to discover your brand through AI search.
The best place to start is Search Console. Pull your top queries, categorise them and use them as anchors. These are real searches, from real users, that your content is already appearing for. From there, layer in audience and persona context to build out the long tail: what would a first-time buyer ask? What would a category researcher want to know six weeks before purchase?
Once you have your prompt set, you can start tracking manually. The ChatGPT Finance Capture plugin is a low-cost way to bulk-run prompts and log whether your brand appears, how it’s described, and whether competitors are being cited alongside you. It’s not glamorous, but it works, and it gives you a baseline before you scale into a paid visibility tool.
The MMM moment is here
Emily made a point in the webinar that deserves more attention than it usually gets: as agentic commerce arrives and on-site tracking continues to diminish, brands will increasingly lean on Marketing Mix Modelling to understand what’s actually driving growth.
Commerce data from Shopify sales, CRM pipeline and revenue by channel becomes the source of truth. GA4 shifts from being the primary attribution tool to being a behavioural lens: useful for understanding what people do on your site, less useful for explaining why they arrived. And SEO, as a discipline, starts to look less like a performance channel and more like a brand measurement task.
That’s a significant shift in how teams are structured, how budgets are justified, and how success is reported. The brands that get ahead of it now, building MMM capability, cleaning up their first-party data, and treating AI visibility as a brand metric, will be better placed when the measurement landscape settles.
Three things to do this month
The framework is clear. The tools exist. Here’s where to start:
- Build AI referrals and Search Console AI impressions into your dashboard.
Even if the numbers are small today, establishing the baseline now means you’ll have trend data when it matters.
- Add a post-purchase ‘how did you hear about us?’ survey.
Keep it simple. One question. The qualitative signal it provides will surprise you. If you are not an e-commerce or lead generation site, you can still get a steer by adding a on site survey pop up product pages using something like HotJar.
- Pick one North Star prompt per product or category and start tracking visibility manually.
Run it weekly on ChatGPT using ChatGPT Search & fan-outs capture. Log the results. Before you invest in a visibility tool, understand what you’re trying to measure.
Measurement has always been the thing that separates brands that grow from brands that guess. In the age of AI search, that’s more true than ever.
Want to talk through your AI search measurement strategy? Get in touch with the team at hello@26pmx.com.