Measurement continued to be a topic at Ahrefs Evolve after BrightonSEO. Everyone has heard of and is talking about the decreasing line in search console. Organic is down. People have said it’s disappearance has been tucked away in direct. The attribution is messier than ever, but that doesn’t mean outcomes are.
At Ahrefs Evolve, we interviewed 12 in-house marketers about how they're handling AI search. When we got to measurement, a surprising number of them told a version of the same story. Their traffic is down, but the business is fine. In several cases it's better.
- Jordan Frey, SEO Manager at U.S. Career Institute: traffic is down about 30%, but leads are up, enrollments are up, and leads convert to enrollments at a higher rate than before.
- James Shaw, Head of Content Marketing and SEO at The Alliance: referral traffic from AI tools converts at four times the rate of traditional organic traffic.
- Alexandre Teulade, Head of Growth at ClearPoint Strategy: fewer MQLs, but a bigger average contract value. "We kind of got rid of the tire kickers."
- Eden Wallace, Senior Manager of Web and AI Discoverability at NRG: "Our sales are all up. Everything's all up."
The pattern makes sense. AI tools answer a lot of top-of-funnel questions on the spot. The people who still click through have usually already done their research. There are fewer of them, and they're closer to buying.
The problem is proving it. As Dave Kimball, marketing director at Invite Networks, put it: "Attribution is the billion-dollar whale. Everyone's trying to figure that out."
Why the old attribution model breaks
The people we talked to named a couple reasons AI search is so hard to attribute, outside of the typical conversation of traffic hiding in direct.
Sales cycles outlast the data. Dave's IT business has sales cycles of one to three years. Tying a first touch in ChatGPT to a deal that closes two years later is close to impossible with today's tools.
The systems don't talk to each other. Eden used the shift in search to push NRG's CMS from on-premises to the cloud, so she could finally connect analytics, journey data and AI visibility in one place. Until that's done, "I can see clicks went down, clicks are now up. I don't know why."
What leadership actually wants to see
We asked every marketer what their leadership looks at when they report on AI search. Seven of the 12 gave some version of the same answer: revenue, pipeline, leads or customers.
Irina Jordan, VP of Marketing at Checkmate, carries a pipeline number herself, so she didn't hesitate: "It is all about SQLs." Leadership asks where deals come from, "but they really don't care how or what. They just want to see the dollar signs at the end." Jonathan Rundle, head of marketing at ChemPoint, said his leadership cares about "revenue, margin impact, leads." Eden's leadership at NRG asks one question: does it add customers or help keep them?
Visibility metrics still have a place, but mostly as supporting evidence. Stefanie Maio at Evans Hotels reports mentions, citations and share of voice, framed explicitly as trends rather than precise numbers. James's leadership watches prompt visibility as a brand-health signal, "but ultimately it's where are the leads coming from and what's the conversion rate."
How in-house teams are connecting the dots
No one we talked to has solved attribution. But the most confident teams are combining several imperfect signals into one story:
Pipeline as the north star. Alexandre's team stopped arguing about which channel gets credit and made "pipeline generated and influenced by marketing" the main goal. They also track new signals, like time spent on money pages and conversion rates on those pages.
AI referral traffic, judged by conversion. James calls this "the metric I can really hang my hat on." The sessions are small. The conversion rate isn't. Jonathan's team is seeing ChatGPT referrals convert too, enough that they're now testing ChatGPT ads.
Self-reported attribution. Irina reads her form fields closely. Leads now name AI search as a source alongside direct and organic. James sees more people saying, "I found you through ChatGPT or Claude or a Google AI Overview."
Context with every number. This might be the biggest takeaway when reporting your success in IA Search to leadership who might understand the full picture. Eden's tip for reporting up: never show an AI metric on its own. When she compared NRG to its top five competitors, she pointed out that AI only recommended a provider in about 100 of 40,000 answers, and a specific mention of NRG was only a handful of times. "Every time you say a number, there's like four lines of context you have to give around it."
One trusted source of truth. Will Heany, SEO manager at Snowflake, has seen two managers send leadership AI numbers that say completely different things. His team is now building prompt tracking inside Snowflake's own data warehouse, which they find "a lot more trustworthy" than outside tools.
Rand Fishkin, who spoke at Ahrefs Evolve on zero-click marketing, recently pointed people to Eduard Blacquière's AI KPI model as "possibly the only thing I've seen comprehensive enough" to show how you're actually doing in AI visibility. It's worth a look if you're building a framework from scratch.
How we'd tell the story to leadership
If your traffic is down and your leads aren't, here's how we'd help you prove it:
- Report traffic, leads and revenue together. Put organic sessions, AI referrals, direct traffic and pipeline on the same page. Tracked separately, each looks small. Together, they tell the real story.
- Make conversion rate the headline. If AI referrals convert at several times the rate of organic, lead with that, not with session counts.
- Add a "How did you hear about us?" field. Make it open text on your high-intent forms. It catches AI influence your analytics can't.
- Use first-party AI data where it exists. Google Search Console now has a generative AI performance report for AI Overviews and AI Mode. Start there before buying more tools.
- Give every number its context. Explain what a metric can and can't tell you, the way Eden does. It builds more trust than a confident number that later falls apart.
For more on the measurement gap, read Measurement is AI search's biggest unanswered question, and listen to Rand Fishkin and Paxton Gray on Attribution Is Dead. Now What?
Fewer visitors, better buyers
A falling traffic chart used to mean something was broken. Today it often means AI is answering the easy questions, and the people who still show up are the ones ready to buy.
That's not a problem to explain away. It's a story to tell, with the numbers to back it up.
Need help proving what AI search is doing for your pipeline? See our GEO/AEO services or let's talk.
