Rand Fishkin posted a wheel chart on LinkedIn last week that should make every CMO uncomfortable. The inner ring shows where audiences say their opinions get shaped: podcasts, SubReddits, industry blogs, YouTube channels, even ChatGPT, Gemini, and Claude. The outer ring shows where the budget goes. The two rings barely rhyme. Fishkin’s diagnosis was blunt: pouring the majority of a marketing budget into AI visibility while only a sliver of the audience says AI tools actually influence them is, in his words, a “great way to set money on fire.”
He’s right about the gap. He’s less specific about how you close it, which is fair, because closing it isn’t a posting problem, it’s a measurement problem. And measurement people have fought this exact war before.
We’ve Seen This Fight Before
In 2010, the PR industry’s own standards body, AMEC, sat in Barcelona and formally killed Advertising Value Equivalency, the practice of converting a press mention into a dollar figure by pricing it as if it were an ad. AVE had survived for decades not because it measured anything real, but because it produced a number executives could put in a slide. The Barcelona Principles that replaced it insisted on outcomes over outputs: don’t tell me how many clips you got, tell me what changed because of them.
Fifteen years later, the AI-visibility gold rush is repeating the AVE mistake in a new costume. Dollars are flowing toward “AI visibility” as a category because it produces a number, a citation count, a share-of-answer percentage, that feels measurable. Meanwhile the channels an audience reports trusting, industry blogs at 10 percent, podcasts at 15 percent, YouTube at 6 percent, SubReddits and comparison sites in the low single digits, get whatever’s left over, partly because they’re harder to track.
That’s the real problem underneath Fishkin’s chart. It isn’t that marketers are choosing badly. It’s that budget follows measurability, and for 20 years the tools have made search and paid social trivially easy to measure while making earned attention, the podcast mention, the Reddit thread, the trade press quote, expensive to track manually. Fix the measurability gap and the budget gap starts to close on its own.
Tagging Links Is Table Stakes, Not a Strategy
Before going further, one piece of housekeeping. Every clickable asset a brand controls, newsletters, guest posts, conference landing pages, social ads, should carry parameters through Google’s Campaign URL Builder so the resulting traffic shows up cleanly in GA4. That’s baseline hygiene, and plenty of PR teams still skip it.
But it’s a small piece of this puzzle, and treating it as the fix is exactly the mistake Fishkin is pointing at. A UTM parameter only exists where three things are true: there’s a link, someone clicks it, and it points at a property you control. Look at the inner ring of that chart again. Podcasts are audio, there’s nothing to click mid-episode. Reddit threads and Slack groups strip or mangle parameters the moment a link gets copy-pasted. A mention inside a ChatGPT or Claude answer isn’t a link at all, it’s a paraphrase with no click event behind it. Three of the four biggest slices of audience influence in that chart — podcasts, industry blogs, and AI tools– combined, are largely invisible to a tool built to track clicks. Clean UTMs on your press releases or newsletters don’t tell you anything about whether the other 80 percent of that wheel is working.
The actual fix requires eight other ways of tracking results, and the trick is matching each one to what it’s actually built to measure, because none of them does everything.
Eight Ways to Measure What UTMs Can’t
1. The gap audit itself, matched to prioritization. Before touching a budget line, put audience-attention data next to spend data the way Fishkin’s own chart does. SparkToro and similar audience-research tools exist for this comparison. If podcasts represent 15 percent of influence and 0.5 percent of spend, that’s not a footnote, that’s the headline of your next planning meeting.
2. Unlinked share-of-voice tracking, matched to awareness. Media monitoring platforms can track brand and topic mentions across news, forums, and social conversation whether or not a link exists. This is the modern, evidence-based descendant of what AVE was trying and failing to do in 1995: put a real number on earned attention without pretending it’s an ad impression.
3. AI citation tracking, matched to consideration during the research phase. Whether a brand gets named, and how favorably, inside an AI-generated answer is a distinct metric from search rank, and it deserves its own reporting line rather than living inside an SEO dashboard. Treat the resulting score as a leading indicator that someone is being influenced during their research, since it usually produces no click for a UTM to catch.
4. Podcast-specific proxies, matched to traffic and leads from unlinkable audio. Vanity URLs read aloud on air, unique promo codes, and post-episode listener surveys are the industry’s workaround for the fact that no link exists to tag in the first place. Some ad networks now support dynamic insertion with pixel-based tracking, but plain old vanity domains still catch more of the audience than any parameter ever will.
5. Share of search, matched to consideration and intent. Tracking branded search volume over time, through Google Trends or Search Console, works as a proxy for whether unlinked exposure, a podcast mention, a conference talk, a Reddit recommendation, is actually moving someone toward branded intent. If a podcast sponsorship correlates with a bump in branded search the following week, that’s a signal no click-tracking model would ever surface.
6. Brand lift and awareness surveys, matched to awareness and consideration directly. Rather than inferring whether an audience knows or trusts a brand from indirect click data, tools like YouGov BrandIndex Lite or the native Brand Lift products from Google and Meta ask people directly. When the KPI is awareness or trust rather than a session on a website, a survey answers the question a pixel was never designed to answer.
7. Post-conversion attribution, matched to leads and sales. A single required field at the point of lead capture or purchase, how did you hear about us, tied into the CRM, catches the podcast listener and the Reddit lurker that no tracking parameter would ever see, because by the time they convert, they’ve often searched the brand name directly rather than clicked through anything.
8. Incrementality testing and marketing mix modeling, matched to validating the reallocation itself. Geo-holdout tests, turning a channel off in one region and comparing outcomes against a matched market, sidestep the attribution-modeling argument entirely. At the portfolio level, marketing mix modeling can statistically weigh spend against outcomes across every channel at once, which is the only method built to answer Fishkin’s actual question: is the money going where the audience’s attention already lives.
None of these eight is exotic, and none of them existed because of generative AI. What’s changed is the size of the gap they now need to close, and the fact that AI added an entirely new unlinkable channel, the answer engine, at the exact moment marketers most need proof that unlinkable attention still counts.
The Uncomfortable Part
Fishkin ended his post promising to fight this battle “day by day, conference by conference, post by post.” I’d argue the fight isn’t won on LinkedIn. It’s won in the budget meeting where someone has to defend moving money away from a channel that produces a clean UTM report toward one that only produces a share-of-voice estimate and a survey response.
That’s a harder pitch to make to a CFO than a dashboard screenshot. It’s also the only pitch that’s honest about where the audience actually is.
