Wil Reynolds didn't mince words. Quoted in a Seer Interactive recap of his and Alisa Scharf's comments to Search Engine Journal, published July 20, 2026, the Seer Interactive founder said: "AI visibility only matters if it's tied to an outcome." Read that as a direct shot at every marketing team currently screenshotting a mention-rate dashboard for a leadership deck. He's not wrong. A raw count of how often a brand shows up in an AI answer, with no connection to pipeline, is exactly the kind of AI search visibility reporting that gets cut in the next budget review. But "tie it to an outcome" is a destination, not a starting instrumentation plan — and skipping straight there is its own mistake.
What counts as AI search visibility reporting right now
Walk into most quarterly reviews right now and the AI search visibility slide looks the same everywhere: one number, usually labeled something like "AI mention rate" or "AI visibility score," trending up and to the right. It's built by a vendor tool that checks whether a brand name shows up somewhere in an AI-generated answer to a set of prompts, then rolls that up into a single percentage. It photographs well. It also tells you almost nothing about whether the work behind it is producing anything a CFO would recognize as a result.
That's the gap Wil Reynolds is calling out, and it's a gap Something Inc. has been writing about in our own client work, from a different angle. Both critiques land on the same target from different directions — one from a working practitioner pushing back on vanity reporting in front of clients, one from an agency that had already found the same problem in its own client data. The disagreement isn't about whether the single blended number is a problem. It's about what to do instead, and that's a real, substantive fight worth having in the open rather than smoothing over.
Wil Reynolds' case: AI visibility only matters if it's tied to an outcome
In Seer Interactive's recap of Wil Reynolds' comments, the Seer Interactive founder pushes back on treating AI mention tracking as a standalone vanity metric, one that exists to justify a program rather than to inform a decision. His argument is straightforward: a number that doesn't move a business outcome isn't a metric worth defending in a room with finance in it.
“"AI visibility only matters if it's tied to an outcome." — Wil Reynolds, founder of Seer Interactive, in Search Engine Journal, as recapped by Seer Interactive, July 20, 2026.”
Reynolds doesn't stop at the critique. He flags a specific, commonly-missed gap in how teams currently track AI mentions: most measurement is binary, present or absent, checked once and reported as a static rate. What actually matters, in his framing, is how a brand's mentions shift across different AI models and across different prompts over time — a trend line across ChatGPT, Perplexity, Gemini, and whatever comes next, not a single snapshot check of whether the brand showed up once. A brand can be present in one model's answers and absent from another's for reasons that have nothing to do with the quality of its content and everything to do with how that specific model retrieves information. Averaging that into one number erases the exact signal a team needs to act on.
Something Inc.'s case: you can't skip straight to outcomes in AI search visibility
Something Inc.'s own published position, laid out in our guide to measuring GEO grounding, citation, and mention as separate market-share numbers, starts from a related but distinct premise. Grounding, citation, and mention are three separate events. Grounding is whether an engine's retrieval layer pulls a brand's content into the context it's reasoning from at all. Citation is whether the engine links that content as a named source. Mention is whether the brand's name appears in the generated answer, cited or not. A brand can be grounded without being cited. It can be mentioned without either. Collapsing all three into one "AI visibility score" isn't simplification — it's a category error, because different engines run genuinely different retrieval mechanics, and a change in the blended number doesn't tell you which of the three events moved, on which engine, or why.
That's also the core argument of our article on why the mention-rate dashboard is lying to you: a single blended AI visibility number can go up while the thing that actually matters, say citation rate on the one engine your buyers actually use, goes down, and a team reporting only the blended figure would never know. The fix isn't to abandon measurement. It's to measure the three events separately, per engine, before compressing anything into a summary metric for a slide.
Our verdict: instrument first, correlate second, cut third
Reynolds is right, plainly, that a raw mention count with no link to pipeline is a vanity metric. It will not survive a serious budget conversation, and it shouldn't. Agree with him on that without qualification. Where the outcome-only framing runs into trouble is in what it implies about sequencing: you cannot tie AI visibility to an outcome if you've never separated grounding from citation from mention in the first place, because you don't yet know which of those three events, on which engine, is the one actually predictive of a downstream result. Skipping straight to "just show me the ROI number" assumes you already know which input matters. Most teams don't, because they've never measured the inputs as separate variables.
| WIL REYNOLDS (SEER INTERACTIVE) | SOMETHING INC.'S PUBLISHED POSITION | |
|---|---|---|
| Core claim | "AI visibility only matters if it's tied to an outcome" — mention tracking with no business result attached is a vanity metric. | Grounding, citation, and mention are three distinct, independently trackable events; a blended 'AI visibility score' is a category error. |
| What it's reacting to | Teams reporting a single AI mention rate with no connection to pipeline or revenue. | Vendor dashboards that average different engines' retrieval mechanics into one composite score. |
| What it recommends measuring | How brand mentions shift across different AI models and prompts over time, not a one-time present/absent check. | Each event — grounded, cited, mentioned — tracked separately, per engine, before deciding which one to tie to an outcome. |
The synthesis, and the actual sequence we'd run with a client: instrument the granular, per-engine numbers first. That is not vanity tracking — it's the necessary intermediate step Reynolds' own advice quietly depends on. Then correlate each one, separately, against pipeline and revenue over a real measurement window, the way we've laid out in our framework for attributing pipeline to organic and AI channels. Only then do you drop the ones that don't correlate. Reynolds' "tie it to an outcome" bar is exactly where a team should end up. It's just not where a team can start, because you can't correlate what you haven't separated.
| CAMP | GETS RIGHT | MISSES |
|---|---|---|
| Reynolds' outcome-first camp | A number with no business result attached will not survive a finance conversation — full stop. | "Tie it to an outcome" assumes you already know which of several distinct AI events is the predictive one. You don't, until you've measured them apart. |
| Something Inc.'s granular-instrumentation camp | Grounding, citation, and mention behave differently on different engines. Blending them hides which lever is actually moving. | Granular tracking on its own isn't an ROI story. It's the input to one, and a team can stall there if it never correlates the numbers to pipeline. |
There's a real cost to skipping the granular step, and it's worth naming plainly. Teams that jump straight to "just show me the ROI number" tend to land in one of two places: no instrumentation at all, because nobody built the tracking underneath the outcome metric, or a black-box vendor score they can't interrogate when a board member asks why it moved. Both are arguably worse than the over-tracking Reynolds is criticizing, because at least a granular, over-built dashboard can be pruned. A number you can't decompose can only be trusted or distrusted wholesale.
How to build AI search visibility reporting that survives a budget conversation
This isn't an abstract debate for a lot of the teams we work with. It comes up constantly with clients in professional services — law, accounting, consulting firms — where a partner asks a straightforward question in a quarterly meeting: are we showing up when a prospect asks ChatGPT who handles this kind of work, and does it matter? The honest answer requires the granular numbers, not a blended AI visibility score, because the firms in that category live or die on referral-adjacent trust signals, and a single averaged number can't tell you whether it's citation on the engine buyers actually use, or just mention volume across engines nobody in their market touches.
Practically, that means building the tracking in the right order: separate grounding, citation, and AI mentions per engine before you build the outcome layer on top. Run that against pipeline for a real window, long enough to see a correlation rather than noise, then report the two or three numbers that actually move revenue and retire the rest. That's the sequencing work we run inside our reporting and analytics service — not because instrumentation is the interesting part of the story, but because it's the part nobody can skip if they want the outcome-tied number Reynolds is asking for to mean anything when someone in finance pulls on it.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.