Several datasets published in October 2026 show that AI assistants send a very small share of the traffic that Google sends to sites. They also show that none of these datasets can isolate the hidden influence that AI would exert on direct traffic and brand search. The columnist specifies that this does not make AI visibility a lure, but that most measurement methods would not pass a basic quality check.
The debate was launched on October 22, 2026 by Rand Fishkin, who published a short video on the SparkToro blog titled ‘Is AI Visibility a Marketing Boondoggle?’. The next day, Amanda Natividad, chief evangelist at SparkToro, asked in her newsletter Audience Research: ‘What if AI visibility is a lure?’. Both authors are co-authors of the book Zero Click Marketing.
The data underpinning these doubts come from other sources. Kevin Indig, from Growth Memo, analyzed 12 GA4 properties, i.e., Google Analytics 4, an audience measurement tool. Marie Haynes studied over 145 million sessions across 100 sites over twelve months. According to her, ChatGPT sent 0.62% of the volume of Google organic, and all LLMs, i.e., large language models, 0.73%. The SEO Works study, covering 95 GA4 properties between September 2025 and August 2026, indicates that organic search sent 39 times more traffic than AI assistants.
Low traffic, contrasting conversion
Haynes notes that visitors from ChatGPT converted at 13.8%, compared to 3.02% for Google organic, i.e., more than 4.5 times better. However, Google still accounted for 96.5% of total conversions in her sample, and organic search generated 27 times more than all LLMs combined. She also observes that the strong growth in ChatGPT sessions stopped in August 2025, around the release of GPT-5.
The SEO Works study shows a smaller gap: 7.18% for AI traffic versus 6.08% for organic, a ratio of 1.18. On 23 of the 63 sites with sufficient volume, however, AI traffic converts less well. For B2B product companies, AI conversion reaches 0.62 times that of organic. In the same report, a 5% increase in organic search generated 10.3 additional conversions per site per month, versus 6.6 for a doubling of AI-originated traffic.
A hidden influence not yet visible
The usual argument is that the click is not essential: a user sees a brand in a ChatGPT response, does not click, then searches for it later. Indig looked for this trace. Eight of his 12 properties showed an increase in direct traffic between August and September, but after filtering engaged sessions, total direct traffic increased by 140.9% over January-September, while engaged direct traffic fell by 1.0%. On one site, GA4 recorded 409,000 direct sessions, of which only 198 were engaged.
Indig then conducted 632 tests across four brands, comparing weekly AI visibility to direct traffic and brand search, with lags of zero to six weeks. Using deliberately false dates, used as a control, several erroneous correlations proved as strong as the real ones. His conclusion: ‘An effect could still exist. I just couldn’t isolate it.’ Natividad recalls that an Ahrefs study covering 75,000 brands, published on May 26, 2025, gives a correlation of 0.392 between brand search volume and visibility in AI Overviews, versus 0.664 for web brand mentions.
A brand channel, rather than a search engine
Neither Fishkin nor Natividad deny the role of AI. Fishkin believes it is ‘probably quite influential on some buyers in some sectors’ and retains Mike King’s formulation, according to which AI is more akin to ‘a brand channel, a digital billboard if you will.’ Natividad summarizes: ‘Familiarity itself is a marketing outcome.’ She adds: ‘If we can’t prove it, we shouldn’t trust attribution simply because it feels right.’
The measurement sector has already voiced these reservations. On May 20, 2026, AMEC published its GEO Principles at its global summit in Dublin, followed by a white paper titled ‘AMEC Data Quality Principles,’ whose lead author is Rob Key, founder and CEO of Converseon. The document proposes four questions to ask of any data before using it: provenance, accuracy, alignment with reality, and trust. On the third, it warns that a metric can be ‘internally consistent but externally false.’ The columnist notes, however, that some figures in the white paper, including an accuracy of 60% to 65% for sentiment models, come from a benchmark funded by Converseon, which sells solutions in this field.
Fishkin asked if AI visibility was a marketing waste. According to the columnist, it becomes one as soon as a figure is reported without having been confronted with what customers actually do.
