Four studies show AI visibility scores do not predict buyer behaviour

Marketers now spend real money on AI search visibility tools. Greg Jarboe, co-founder of SEO-PR, argues in Search Engine Journal that most of those dashboards fail a basic test. The number they produce does not predict what customers do next.

The four studies he lines up

Jarboe cites Marie Haynes, who looked at twelve months of Google Analytics data across 100 sites. ChatGPT sent 0.62 percent of the traffic Google organic sent. All large language models together sent 0.73 percent. Google organic drove 27 times more conversions than every language model combined.

Jarboe then cites SEO Works, which measured 95 web properties over a twelve-month window. Organic search sent 39 times more visits than AI assistants. A 5 percent lift in organic search delivered 10.3 extra conversions per site per month. Doubling AI referral traffic delivered 6.6.

He cites Kevin Indig, who ran 632 correlation tests across 12 Google Analytics properties. Indig checked weekly AI visibility against direct traffic and branded search, at delays of zero to six weeks. Indig told readers he could not isolate an effect. When he ran placebo tests with deliberately wrong dates, some of them correlated as strongly as the real ones.

Finally Jarboe cites an Ahrefs study of 75,000 brands. Branded search volume correlated with AI Overview visibility at 0.392. Amanda Natividad reads that as weak: if AI visibility reliably drove branded search, the number would be higher.

What Jarboe wants marketers to use instead

Jarboe points at AMEC, the International Association for Measurement and Evaluation of Communication. AMEC has published a set of GEO Principles. The framework asks four questions of any measurement. Is the data from the expected sources? Is the metadata faithful? Does the data match real behaviour? Do stakeholders trust it?

Jarboe builds a budget defence from that. Stop setting targets on AI traffic and mark it as a diagnostic signal. Document every tool, every prompt and every bot filter in an AMEC audit. Run Indig's placebo test on your own data. Build a library of prompts worded as real buyers ask. Add a "how did you hear about us" field to every form. Spend on the earned authority that both channels share. Jarboe gives that site overlap a correlation of 0.557.

Natividad, quoted in the piece, writes that familiarity itself is a marketing outcome. She then warns that nobody should trust an attribution just because it feels right. Jarboe reads AI visibility as a real brand channel being measured with the wrong metric.

Source

AI Visibility Isn't A Sham, Most Ways Of Measuring It Are, Greg Jarboe, Search Engine Journal