The GEO industry has been selling visibility metrics that no longer mean what they did in SEO. It seems the builders, the analysts who cover them, the venture capital that funded them, and the buyers who subscribe to them are operating under a collective psychosis. Share of voice is worthless when the voice is not yours.
In SEO, share of voice mattered because your voice was the one that ultimately spoke. The search engine surfaced links to first party sources. The user clicked a result and arrived at your website, your messaging, your facts. Being visible meant being heard in your own words. The metric was a proxy for the actual outcome: a customer landed on something you wrote and saw exactly what you wanted them to see.
In AI, the search engine is no longer a switchboard. The AI is the voice. These platforms summarize, paraphrase, assert, and increasingly invent. The user often never clicks through. The user never leaves the LLM chat window.
That single shift breaks every visibility metric the GEO industry imported from SEO.
As a matter of fact, showing up more than your competitor is dangerous if what AI is saying about you is false. Visibility is a multiplier on whatever the AI happens to be saying. And the AI, today, is saying a lot of things that are not true.
Yes, brands can influence what AI says. That is exactly what serious AI Integrity work is. Shaping is not the same as speaking, which is why monitoring the actual output is the entire job.
Has anyone in the GEO industry asked this question out loud? We took a pretty close look at thirty different tools. The legacy SEO and social listening incumbents pivoted in: Conductor, Brandwatch, Cision, Talkwalker, Meltwater, Sprinklr, Yext, BrightEdge, Semrush, Ahrefs, HubSpot. Every one has bolted an LLM-querying module onto its existing platform in roughly the same window. The LLM piece is a feature, not a product. The GEO native startups raised over three hundred million dollars to ship the same shape. Profound at one hundred fifty five million. Bluefish AI at sixty eight million. Brandlight at thirty six million. Almost all counting mentions against competitors. Almost all pointless when the mentions are wrong. Bluefish AI’s accuracy module is the rare exception, and proves the point: the rest of the industry could have built this and did not.
I spent fifteen years inside the SEO industry. I know the playbook. Prompts replaced keywords. Share of voice stayed share of voice. Citation tracking replaced backlink tracking. The label changed, the product did not. None of these carry-forward metrics survive the move to AI. Backlinks were authority signals a search algorithm weighted. Citations in an AI answer are decoration the model generated to look sourced. There was a position one in a SERP and a click rate attached to it. There is no position one in a generative answer. A page change in SEO indexed in hours. A factual change in AI memory waits for the next training cycle, measured in quarters. The metric, the mechanism, and the feedback loop are all different. The category imported the metric and called it innovation.
The evidence that AI gets material business facts wrong is not disputed. Frontier reasoning models invent product launches that never happened. They credit acquisitions to the wrong owner years after the deal closed. They describe executives who left as still in their roles. They attribute products and trademarks to companies that never made them. They cite legal cases and regulatory actions that do not exist. This is documented, public, and reproducible across the major platforms. The GEO industry must know this deep down. So why is it selling the same tired SEO tools for an entirely new technology?
This is a category error performed at the scale of an industry.
Lipstick on a pig is still a pig. A business with real AI exposure does not need a competitive leaderboard. It needs answers to questions the leaderboard cannot reach. It needs to know whether the AI is right. It needs to know exactly which models and retrieval modes are incorrect, and the severity of the misinformation. It needs the verbatim quotes as evidence, not sentiment scores. It needs to know which training source the AI is repeating from, because that is the only place a correction can land. Until the industry can deliver that, every dollar spent on visibility is a dollar spent watching the wrong things.
Andrew David Linde
Founder and Principal
Craton Meridian™ | AI Integrity™